{"@context":"https://schema.org","@type":"DataCatalog","@id":"https://lunartuliplab.com/research.json#catalog","name":"Lunartulip Canonical Research Objects","publisher":{"@type":"Organization","@id":"https://lunartuliplab.com/#organization","name":"Lunartulip Lab","url":"https://lunartuliplab.com/"},"dateModified":"2026-09-14","inLanguage":["zh-CN","en"],"usageNotice":"Public metadata is provided for discovery and citation. Publication does not grant a commercial data license. Research is not investment advice.","citationFormat":"Lunartulip Lab, [title], [research object ID], version [version], published [publishedAt], as of [asOf], [canonical URL].","dataset":[{"id":"RO-COMP-PLTR-001","slug":"palantir-ai-application-commercialization-2026q2","kind":"company-deep-dive","tickers":["PLTR"],"publishedAt":"2026-08-05","asOf":"2026-09-14","version":"3.0.0","sourceLabel":"Palantir commercial budget migration · 2026-09-14 edition","claims":[{"id":"CL-PLTR-001","type":"Fact","text":{"zh-CN":"Palantir 2026 年第二季度收入 19.35464 亿美元，同比增长 93%；其中美国商业收入 7.64 亿美元，同比增长 149%。","en":"Palantir reported Q2 2026 revenue of $1.935464 billion, up 93% year over year, including U.S. commercial revenue of $764 million, up 149%."},"evidenceIds":["EV-PLTR-001"]},{"id":"CL-PLTR-002","type":"Fact","text":{"zh-CN":"美国商业 TCV bookings 为 21.32 亿美元，同比增长 153%；公司同时把 2026 年收入指引上调至 81.50–81.58 亿美元。TCV 计量的是合同承诺总额。","en":"U.S. commercial TCV bookings reached $2.132 billion, up 153%, and FY2026 revenue guidance was raised to $8.150–8.158 billion. TCV measures total contracted commitment."},"evidenceIds":["EV-PLTR-001"]},{"id":"CL-PLTR-003","type":"Derived","text":{"zh-CN":"按当季披露数计算，美国商业收入相当于美国政府收入的约 94%（7.64 亿 / 8.09 亿），商业预算已经构成第二条规模化收入来源。","en":"On the quarter's disclosed figures, U.S. commercial revenue equals roughly 94% of U.S. government revenue ($764m / $809m), making commercial budgets a second revenue base at scale."},"evidenceIds":["EV-PLTR-001"]},{"id":"CL-PLTR-004","type":"Hypothesis","text":{"zh-CN":"如果 bookings、TCV 与确认收入继续同步，说明企业客户已把 AIP 从试验预算迁进核心经营预算，高增长的久期会比单季增速所暗示的更长。公司没有按预算类别披露收入，这一条目前只能作为假设。","en":"If bookings, TCV and recognized revenue keep moving together, enterprise customers have shifted AIP from pilot spend into the core operating plan, and the duration of elevated growth would run longer than a single quarter's rate implies. Palantir discloses no revenue split by budget category, so this remains a hypothesis."},"evidenceIds":["EV-PLTR-001","EV-PLTR-002"]},{"id":"CL-PLTR-005","type":"Inference","text":{"zh-CN":"TTM P/S 约 65x、对 FY26 指引中点 forward P/S 约 49x 的定价把数年不降档写进前提；增速从 93% 降到 60% 仍属顶级增长，但倍数可能先于业绩重定价——9 月回撤已示范这一路径。","en":"Pricing at roughly 65× TTM sales and about 49× forward sales on the FY2026 guidance midpoint embeds years of non-decelerating growth; revenue growth falling from 93% to 60% would still be elite, yet the multiple can reprice ahead of earnings, as the September pullback demonstrated."},"evidenceIds":["EV-PLTR-001","EV-PLTR-003"]},{"id":"CL-PLTR-006","type":"Inference","text":{"zh-CN":"2026 年 9 月 11 日收盘 Palantir 股价 $167.23，市值约 $401.87B，较 52 周高点 $207.52 回撤约 19%；9 月 2 日单日收跌 5.8%，当日无负面公司公告。","en":"Palantir closed at $167.23 on September 11, 2026, with a market cap of about $401.87B, roughly 19% below the 52-week high of $207.52; on September 2 the stock fell 5.8% in a session with no negative company announcement."},"evidenceIds":["EV-PLTR-004"]},{"id":"CL-PLTR-007","type":"Fact","text":{"zh-CN":"2026 年 9 月 1 日美国陆军公告 TITAN 下一代移动情报节点量产合同总额 $192M，Palantir 作为主承包商获 $127M、Anduril 获 $65M，订单 2026 年 8 月 31 日生效。","en":"On September 1, 2026 the U.S. Army announced a $192M production award for the TITAN next-generation mobile intelligence node, with Palantir as prime contractor at $127M and Anduril at $65M, effective August 31, 2026."},"evidenceIds":["EV-PLTR-005"]}],"evidence":[{"id":"EV-PLTR-001","title":"Palantir Q2 2026 Business Update","source":{"publisher":"Palantir Investor Relations","url":"https://investors.palantir.com/files/Palantir%20-%20Q2%202026%20Business%20Update.pdf","document":"Q2 2026 Business Update"},"sourceDate":"2026-08-04","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"TCV is not revenue and may include multi-year commitments or timing effects.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-PLTR-002","title":"Palantir Q1 2026 Form 10-Q","source":{"publisher":"Palantir Investor Relations","url":"https://investors.palantir.com/files/2026%20Q1%20PLTR%2010-Q.pdf","document":"2026 Q1 Form 10-Q"},"sourceDate":"2026-05-06","dataAsOf":"2026-03-31","calculation":null,"confidence":"high","counterevidence":"Many customer contracts contain termination-for-convenience provisions.","lastVerified":"2026-08-15","role":"supporting"},{"id":"EV-PLTR-003","title":"Point-in-time valuation input gap","source":{"publisher":"LunarTulip Research","document":"Market-cap input intentionally withheld pending licensed point-in-time tape verification"},"sourceDate":"2026-08-05","dataAsOf":"2026-08-05","calculation":"Enterprise value scenarios are sensitivity outputs, not observed market value or price targets.","confidence":"low","counterevidence":"Without a verified point-in-time enterprise value, no observed forward-sales multiple is asserted.","lastVerified":"2026-08-15","role":"supporting"},{"id":"EV-PLTR-004","title":"MarketWatch PLTR quote snapshot","source":{"publisher":"MarketWatch","url":"https://www.marketwatch.com/investing/stock/pltr","document":"PLTR quote page"},"sourceDate":"2026-09-11","dataAsOf":"2026-09-11","calculation":null,"confidence":"high","counterevidence":"Third-party quote pages can differ on market cap and 52-week high by a few percent.","lastVerified":"2026-09-14","role":"primary"},{"id":"EV-PLTR-005","title":"Army TITAN production award","source":{"publisher":"Breaking Defense","url":"https://breakingdefense.com/2026/09/army-awards-192m-to-palantir-and-anduril-to-produce-titan-system/","document":"Army TITAN award coverage"},"sourceDate":"2026-09-02","dataAsOf":"2026-09-01","calculation":null,"confidence":"high","counterevidence":"Contract awards are not revenue until recognized under ASC 606.","lastVerified":"2026-09-14","role":"primary"}],"financialBridge":[{"label":{"zh-CN":"Q2 2026 收入","en":"Q2 2026 revenue"},"period":"2026-Q2","value":1935.464,"unit":"USDm","evidenceIds":["EV-PLTR-001"]},{"label":{"zh-CN":"美国商业收入","en":"U.S. commercial revenue"},"period":"2026-Q2","value":764,"unit":"USDm","evidenceIds":["EV-PLTR-001"]},{"label":{"zh-CN":"美国政府收入","en":"U.S. government revenue"},"period":"2026-Q2","value":809,"unit":"USDm","evidenceIds":["EV-PLTR-001"]},{"label":{"zh-CN":"FY2026 收入指引中点","en":"FY2026 revenue guidance midpoint"},"period":"2026-FY","value":8154,"unit":"USDm","evidenceIds":["EV-PLTR-001"]},{"label":{"zh-CN":"Q3 2026 营收指引中点","en":"Q3 2026 revenue guidance midpoint"},"period":"2026-Q3","value":2162,"unit":"USDm","evidenceIds":["EV-PLTR-001"]}],"valuationScenarios":[{"id":"VAL-PLTR-BASE","label":{"zh-CN":"增长延续","en":"Growth continues"},"revenueBaseUsdM":8154,"revenueGrowthPct":45,"forwardRevenueUsdM":11823.3,"salesMultiple":24,"impliedEnterpriseValueUsdM":283759.2,"calculation":"8154 * 1.45 = 11823.3; 11823.3 * 24 = 283759.2","interpretation":{"zh-CN":"以指引中点为基数，假设 45% 的远期增长和 24 倍 sales multiple，输出企业价值。当日 observed multiple 未经核验，24 倍是设定的假设，不构成对现价的比较。","en":"Takes the guidance midpoint, assumes 45% forward growth and a 24× sales multiple, and outputs enterprise value. With no verified point-in-time multiple, 24× is an assumption rather than a comparison against the tape."}},{"id":"VAL-PLTR-BEAR","label":{"zh-CN":"增长降档","en":"Growth resets"},"revenueBaseUsdM":8154,"revenueGrowthPct":25,"forwardRevenueUsdM":10192.5,"salesMultiple":14,"impliedEnterpriseValueUsdM":142695,"calculation":"8154 * 1.25 = 10192.5; 10192.5 * 14 = 142695","interpretation":{"zh-CN":"增长退到 25%、倍数退到 14 倍时，企业价值约为基准情景的一半。它说明定价对增长和倍数同时敏感，不映射每股价值。","en":"At 25% growth and a 14× multiple, enterprise value lands near half the base case. The point is joint sensitivity to growth and multiple; nothing here maps to per-share value."}}],"counterevidenceClaimIds":["CL-PLTR-005"],"falsifiers":[{"id":"FAL-PLTR-001","metric":"U.S. commercial revenue YoY growth","operator":"<","threshold":100,"unit":"percent","horizon":"any of the next two reported quarters","claimIds":["CL-PLTR-004"],"rationale":{"zh-CN":"美国商业增速在 Q3 跌破 +100% 是本文框架的第一降档信号；仍属高增速，但不足以支撑当前倍数对久期的要求。","en":"U.S. commercial growth below +100% in Q3 is the first downgrade signal in this framework; still fast in absolute terms, but insufficient for the duration embedded in the current multiple."}},{"id":"FAL-PLTR-002","metric":"U.S. commercial TCV bookings YoY growth","operator":"<","threshold":50,"unit":"percent","horizon":"any of the next two reported quarters","claimIds":["CL-PLTR-004"],"rationale":{"zh-CN":"TCV 是领先序列。它先跌破 50%，合同与确认收入的同步关系就断了，而这条同步关系是预算迁移命题的全部观察基础。","en":"TCV is the leading series. If it breaks below 50% first, the alignment between contracts and recognized revenue is gone, and that alignment is the entire observable basis for the budget-migration thesis."}}],"versions":[{"version":"1.0.0","date":"2026-08-05","diff":{"zh-CN":"原始时点判断：把 PLTR 读作 AI 应用层商业化的强正向样本。","en":"Original point-in-time view: PLTR read as a strong positive sample for AI application-layer monetization."}},{"version":"2.0.0","date":"2026-08-15","diff":{"zh-CN":"删掉错误的 10–12 倍 P/S、技术价位、仓位与 HYP 编号，换成第一方证据、可复算的情景算术和量化证伪条件。","en":"Removed the erroneous 10–12× sales multiple, technical levels, position sizing and the HYP identifier, replacing them with primary evidence, reproducible scenario arithmetic and quantitative falsifiers."}},{"version":"3.0.0","date":"2026-09-14","diff":{"zh-CN":"2026-09-14 更新：纳入 9 月估值回撤、TITAN 量产合同与 Q3 对账窗口；保留 v2 证据链并追加公开市场来源。","en":"September 14 update: adds the September valuation pullback, the TITAN production award and the Q3 reconciliation window while keeping the v2 evidence chain and adding public-market sources."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"Palantir：商业化有硬数，定价零容错","question":"美国商业客户把 AIP 记在创新预算里，还是记在核心经营预算里？","standfirst":"Q2 的商业收入、TCV 与利润同时加速，这一点没有争议。还需要验证的是这笔支出的预算归属：创新预算会随周期收缩，核心经营预算会被保住。","whyItMatters":"上游算力支出早就能从几家公司的资本开支里读出来，应用层能不能反复拿到企业的经营预算，公开证据一直很薄。Palantir 是少数把这条链条两端都披露出来的样本：合同签了多少，收入确认了多少。它的定价方式又把增长久期推到研究中心。一家按 forward sales 定价的公司，容错来自增长还能持续多久；单季增速再高，也只解释其中一部分。","consensus":"美国商业收入增长 149%、TCV bookings 增长 153%、全年指引再次上调，这些已经在价格里。","differentiated":"增量信息来自三条序列能否继续同步：TCV bookings、确认收入、全年指引。TCV 记的是承诺额，多年期合同和签约时点都能让它单独走高，只有确认收入跟着走，采购和收入确认这两道关口才算真正穿过。第二个观察点是商业收入相对政府收入的位置，Q2 已经到约 94%，越过去意味着这家公司的收入结构被重写一次。","valuation":"估值只做企业价值敏感性。基数取公司自己的全年收入指引，增长和 sales multiple 都写进算式，读者可以重算。2026-08-05 当日经许可的企业价值和完全稀释股数不在手上，所以这里不陈述当日 forward P/S，也不给每股结论。","update":"v3.0 保留 2026-08-05 发布时点，asOf 更新至 2026-09-14，纳入 9 月回撤、TITAN 合同与 Q3 对账窗口。","evidenceClaimIds":["CL-PLTR-001","CL-PLTR-002","CL-PLTR-003","CL-PLTR-006","CL-PLTR-007"],"causalChain":[{"title":"AIP 接进业务流程","body":"Ontology 把数据、权限和动作放到同一层，模型输出落在具体决策上。","claimIds":["CL-PLTR-004"]},{"title":"预算类别改变","body":"TCV、bookings 与确认收入同步，才说明这笔钱通过了采购和收入确认。","claimIds":["CL-PLTR-002","CL-PLTR-004"]},{"title":"久期接受检验","body":"领先的合同指标一旦先减速，增长久期假设就要下修。","claimIds":["CL-PLTR-005"]}],"riskClaimIds":["CL-PLTR-005"],"forwardTestIds":["FAL-PLTR-001","FAL-PLTR-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Palantir: Hard commercial proof, zero valuation slack","question":"Are U.S. commercial customers funding AIP from innovation budgets or from the core operating plan?","standfirst":"Commercial revenue, TCV and margin accelerated together in Q2, and that much is settled. What still needs proving is where the money sits. Innovation budgets get cut in a downturn. Operating budgets get defended.","whyItMatters":"Capital-expenditure disclosure long ago made upstream AI spend easy to observe. Evidence that the application layer can repeatedly win enterprise operating budgets is far thinner. Palantir is unusual in disclosing both ends of that chain: what was contracted and what was recognized. It also prices off forward sales, which puts duration at the center of the work. For a company valued that way, the margin for error lives in how long growth lasts, and a single quarter's rate explains only part of it.","consensus":"149% U.S. commercial growth, 153% TCV-bookings growth and another guidance raise are already in the price.","differentiated":"The information gain comes from whether three series keep moving together: TCV bookings, recognized revenue and full-year guidance. TCV records committed value, and multi-year terms or signing timing can lift it on their own. Recognition is what confirms the spend cleared both procurement and the revenue standard. The second marker is commercial revenue against government revenue, already near 94% in Q2. Crossing it would rewrite the company's revenue mix.","valuation":"Valuation here stops at enterprise-value sensitivity. The base is the company's own full-year guidance, with growth and the sales multiple written into arithmetic a reader can rebuild. Licensed point-in-time enterprise value and fully diluted share count for August 5 were not available, so no observed forward-sales multiple and no per-share conclusion appear.","update":"Version 3.0 keeps the August 5 publication date, moves asOf to September 14, and adds the September pullback, the TITAN award and the Q3 reconciliation window.","evidenceClaimIds":["CL-PLTR-001","CL-PLTR-002","CL-PLTR-003","CL-PLTR-006","CL-PLTR-007"],"causalChain":[{"title":"AIP lands inside the workflow","body":"Ontology puts data, permissions and actions in one layer, so model output attaches to a specific decision.","claimIds":["CL-PLTR-004"]},{"title":"The budget line changes","body":"Only when TCV, bookings and recognized revenue move together has the spend cleared procurement and recognition.","claimIds":["CL-PLTR-002","CL-PLTR-004"]},{"title":"Duration gets tested","body":"If the leading contract indicators slow first, the growth-duration assumption has to come down.","claimIds":["CL-PLTR-005"]}],"riskClaimIds":["CL-PLTR-005"],"forwardTestIds":["FAL-PLTR-001","FAL-PLTR-002"]}},"dataGaps":[{"zh-CN":"2026-08-05 当日经许可的 point-in-time 企业价值和完全稀释股数不可得，因此本对象不陈述当日 forward P/S，也不给每股价值；情景里的 sales multiple 都是外生假设。","en":"Licensed point-in-time enterprise value and fully diluted share count for August 5 were unavailable, so the object states no observed forward-sales multiple and no per-share value. Every sales multiple in the scenarios is an exogenous assumption."},{"zh-CN":"Q3 2026 财报确切日期与财报前卖方一致预期分布不可得，增速对指引的比较以公司自身口径为准。","en":"The exact Q3 2026 earnings date and pre-earnings sell-side consensus distribution are unavailable; growth comparisons use company guidance only."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/palantir-ai-application-commercialization-2026q2","en":"https://lunartuliplab.com/en/deep-dive/palantir-ai-application-commercialization-2026q2"},"machineReadable":"https://lunartuliplab.com/research/palantir-ai-application-commercialization-2026q2","dateModified":"2026-09-14"},{"id":"RO-COMP-NET-001","slug":"cloudflare-monetization-density-2026q2","kind":"company-deep-dive","tickers":["NET"],"publishedAt":"2026-08-07","asOf":"2026-09-14","version":"2.0.0","sourceLabel":"Cloudflare agentic internet · 2026-09-14 edition","claims":[{"id":"CL-NET-001","type":"Fact","text":{"zh-CN":"Cloudflare 2026 年第二季度收入 6.961 亿美元，同比增长 36%；current RPO 同比增长 35%。","en":"Cloudflare reported Q2 2026 revenue of $696.1 million, up 36%, with current RPO up 35%."},"evidenceIds":["EV-NET-001"]},{"id":"CL-NET-002","type":"Fact","text":{"zh-CN":"截至 2026 年 6 月末，年化收入超过 10 万美元的大客户 4,698 家；当季自由现金流 5,640 万美元，对应 8% 的自由现金流率。","en":"At the end of June 2026 Cloudflare had 4,698 customers above $100,000 in annualized revenue, and generated $56.4 million of free cash flow in the quarter, an 8% margin."},"evidenceIds":["EV-NET-001","EV-NET-002"]},{"id":"CL-NET-003","type":"Fact","text":{"zh-CN":"公司在 2026 Investor Day 披露，2025-06-01 至 2026-05-31 期间，其网络上的日均 AI agent 请求量增长超过 1,700%。这是使用量指标，公司未同时给出对应的收入口径。","en":"At its 2026 Investor Day, Cloudflare disclosed growth of more than 1,700% in daily AI-agent requests across its network between June 1, 2025 and May 31, 2026. This is a usage metric, disclosed without any accompanying revenue bridge."},"evidenceIds":["EV-NET-003"]},{"id":"CL-NET-004","type":"Hypothesis","text":{"zh-CN":"决定 NET 收入的变量是 monetization density：一次 agent 动作穿过 Workers、网络、安全和身份等收费层的数量，比请求总量更接近收入。公司不披露 attach rate，这一条只能借财务序列间接检验。","en":"The variable that should drive NET's revenue is monetization density: the count of paid Workers, network, security and identity layers a single agent action touches sits closer to revenue than aggregate request volume. Cloudflare discloses no attach rate, so the hypothesis can only be tested indirectly through the financial series."},"evidenceIds":["EV-NET-001","EV-NET-003"]},{"id":"CL-NET-005","type":"Inference","text":{"zh-CN":"agent 请求继续高速增长、而收入、current RPO 与自由现金流率没有同步改善，说明大部分机器流量停在低价或未计费的层里。","en":"If agent requests keep compounding while revenue, current RPO and free-cash-flow margin fail to improve alongside them, most of the machine traffic is sitting in low-priced or unmetered layers."},"evidenceIds":["EV-NET-001","EV-NET-003"]},{"id":"CL-NET-006","type":"Fact","text":{"zh-CN":"Cloudflare Q2 FY26 DBNR（美元净留存率）120%，同比 +6 个百分点；CEO 在电话会表示历史上首次网络上超过 50% 的流量来自非人类 agentic 流量。","en":"Cloudflare Q2 FY26 dollar-based net retention was 120%, up 6 points year over year; the CEO said on the earnings call that for the first time more than 50% of network traffic came from non-human agentic traffic."},"evidenceIds":["EV-NET-004"]},{"id":"CL-NET-007","type":"Inference","text":{"zh-CN":"9 月 3 日 OpenAI 安全合作与 9 月 8 日 Meta Muse 联想推动 9 月 9 日单日 +10.51%，但两条催化均未披露财务条款；PS(TTM) 约 43x 已把 agentic 期权计入大半，强度兑现取决于 10 月 29 日 Q3 财报。","en":"The September 3 OpenAI security partnership and September 8 Meta Muse association helped drive a +10.51% single-day move on September 9, yet neither catalyst disclosed financial terms; at roughly 43× TTM sales much of the agentic option is already priced, and intensity reconciliation falls to the October 29 Q3 report."},"evidenceIds":["EV-NET-001","EV-NET-005"]}],"evidence":[{"id":"EV-NET-001","title":"Cloudflare Q2 2026 financial results","source":{"publisher":"Cloudflare Investor Relations","url":"https://cloudflare.net/news/news-details/2026/Cloudflare-Announces-Second-Quarter-2026-Financial-Results/default.aspx","document":"Q2 2026 earnings release"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"Revenue and RPO growth do not isolate AI-agent contribution.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-NET-002","title":"Cloudflare Q2 2026 investor presentation","source":{"publisher":"Cloudflare Investor Relations","url":"https://cloudflare.net/files/doc_financials/2026/q2/Q2-2026-Cloudflare-Investor-Presentation.pdf","document":"Q2 2026 Investor Presentation"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"Large-customer count is a scale measure, not direct evidence of agent monetization.","lastVerified":"2026-08-15","role":"supporting"},{"id":"EV-NET-003","title":"Cloudflare 2026 Investor Day","source":{"publisher":"Cloudflare Investor Relations","url":"https://cloudflare.net/files/doc_downloads/Presentations/2026/06/15/New/Master-Deck-2026-Investor-Day-FINAL.pdf","document":"2026 Investor Day master deck"},"sourceDate":"2026-06-09","dataAsOf":"2026-05-31","calculation":null,"confidence":"high","counterevidence":"Request growth is a usage metric and carries no disclosed revenue bridge.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-NET-004","title":"Cloudflare Q2 FY26 Exhibit 99.1","source":{"publisher":"Cloudflare Investor Relations","url":"https://cloudflare.net/files/doc_financials/2026/q2/Q2-26-Exhibit-99-1.pdf","document":"Q2 FY26 Exhibit 99.1"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"DBNR and traffic-mix commentary do not isolate AI revenue.","lastVerified":"2026-09-14","role":"primary"},{"id":"EV-NET-005","title":"September 2026 agentic rerating coverage","source":{"publisher":"The Motley Fool","url":"https://www.fool.com/investing/2026/09/09/why-cloudflare-rallied-today/","document":"September 9 rally coverage"},"sourceDate":"2026-09-09","dataAsOf":"2026-09-09","calculation":null,"confidence":"medium","counterevidence":"Market commentary does not establish contractual economics.","lastVerified":"2026-09-14","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"Q2 2026 收入","en":"Q2 2026 revenue"},"period":"2026-Q2","value":696.1,"unit":"USDm","evidenceIds":["EV-NET-001"]},{"label":{"zh-CN":"收入同比增速","en":"Revenue growth (YoY)"},"period":"2026-Q2","value":36,"unit":"percent","evidenceIds":["EV-NET-001"]},{"label":{"zh-CN":"自由现金流","en":"Free cash flow"},"period":"2026-Q2","value":56.4,"unit":"USDm","evidenceIds":["EV-NET-001"]},{"label":{"zh-CN":"大客户数（>$100k ARR）","en":"Large customers (>$100k ARR)"},"period":"2026-Q2","value":4698,"unit":"count","evidenceIds":["EV-NET-002"]},{"label":{"zh-CN":"DBNR","en":"DBNR"},"period":"2026-Q2","value":120,"unit":"percent","evidenceIds":["EV-NET-004"]},{"label":{"zh-CN":"Q3 FY26 营收指引下限","en":"Q3 FY26 revenue guidance low"},"period":"2026-Q3","value":736,"unit":"USDm","evidenceIds":["EV-NET-004"]}],"valuationScenarios":[{"id":"VAL-NET-BASE","label":{"zh-CN":"收费密度上升","en":"Density rises"},"revenueBaseUsdM":2784.4,"revenueGrowthPct":32,"forwardRevenueUsdM":3675.408,"salesMultiple":20,"impliedEnterpriseValueUsdM":73508.16,"calculation":"696.1 * 4 = 2784.4; 2784.4 * 1.32 = 3675.408; 3675.408 * 20 = 73508.16","interpretation":{"zh-CN":"以 Q2 年化后的 27.844 亿美元为基数，假设 32% 增长和 20 倍 sales multiple。它测的是收费密度改善这一种状态，与公司指引无关。","en":"Uses the annualized $2,784.4 million base with 32% growth and a 20× sales multiple. It tests one state of the world, density improving, and carries no relation to company guidance."}},{"id":"VAL-NET-BEAR","label":{"zh-CN":"流量没有变现","en":"Traffic does not monetize"},"revenueBaseUsdM":2784.4,"revenueGrowthPct":20,"forwardRevenueUsdM":3341.28,"salesMultiple":12,"impliedEnterpriseValueUsdM":40095.36,"calculation":"696.1 * 4 = 2784.4; 2784.4 * 1.20 = 3341.28; 3341.28 * 12 = 40095.36","interpretation":{"zh-CN":"增长退到 20%、倍数退到 12 倍，企业价值只剩基准情景的约 55%。请求量与财务序列一旦脱节，增长和倍数会同时降档。","en":"At 20% growth and a 12× multiple, enterprise value falls to roughly 55% of the base case. Once request volume and the financial series diverge, growth and the multiple reset together."}}],"counterevidenceClaimIds":["CL-NET-007"],"falsifiers":[{"id":"FAL-NET-001","metric":"revenue YoY growth","operator":"<","threshold":25,"unit":"percent","horizon":"any of the next two reported quarters","claimIds":["CL-NET-004"],"rationale":{"zh-CN":"请求量出现 1,700% 的增长之后，收入增速反而跌到 25% 以下，说明收费密度没有形成传导。25% 的选点低于当前的 36%，留出正常波动空间。","en":"Revenue growth below 25% after a 1,700% surge in requests would show that monetization density is not transmitting. The threshold sits under the current 36% to leave room for ordinary variation."}},{"id":"FAL-NET-002","metric":"current RPO YoY growth","operator":"<","threshold":25,"unit":"percent","horizon":"any of the next two reported quarters","claimIds":["CL-NET-004"],"rationale":{"zh-CN":"current RPO 记录已签约、预计在一年内确认的部分。它跌破 25%，说明新增的收费层没有进到合同里，密度改善也就缺少前瞻证据。","en":"Current RPO records contracted revenue expected within a year. Below 25%, additional paid layers are not entering contracts, and the density case loses its forward evidence."}},{"id":"FAL-NET-003","metric":"DBNR","operator":"<","threshold":118,"unit":"percent","horizon":"Q3 FY26 earnings","claimIds":["CL-NET-006"],"rationale":{"zh-CN":"DBNR 跌破 118 说明存量扩张与 agent 叙事脱节，9 月重估的留存基础被削弱。","en":"DBNR below 118 would show installed-base expansion diverging from the agent narrative, weakening the retention basis behind the September rerating."}}],"versions":[{"version":"1.0.0","date":"2026-08-15","diff":{"zh-CN":"从原 NET+TEAM 合并页拆出独立公司对象，删除无来源的 66% option premium、技术价位和仓位建议。","en":"Split out of the combined NET+TEAM page as a standalone company object, with the unsourced 66% option-premium estimate, technical levels and sizing guidance removed."}},{"version":"2.0.0","date":"2026-09-14","diff":{"zh-CN":"2026-09-14 更新：纳入 DBNR 120%、非人类流量过半、9 月 OpenAI/Meta 催化与 Q3 $736.0M 指引对账窗口。","en":"September 14 update: adds DBNR at 120%, majority non-human traffic, September OpenAI/Meta catalysts and the Q3 $736.0M reconciliation window."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"Cloudflare：流量过半已是机器","question":"机器请求的增长能不能穿过更多收费层，并在合同余额和现金流里留下痕迹？","standfirst":"1,700% 证明流量确实来了。它没有说明单位机器活动值多少钱：公司没有披露对应收入，没有给付费请求占比，也没有给跨产品 attach rate。需要跟踪的量是 monetization density（收费密度）。","whyItMatters":"一个 agent 完成任务会反复调用工具、抓取页面、重新验证身份，调用频次远高于人类会话。这对 Cloudflare 是双向的。请求量可以在几乎不收费的层里增长，也可以在同一个动作上叠加 Workers 执行、安全检查、身份校验和网络转发。前一种状态下，流量曲线好看而收入曲线不动；后一种状态下，收入可以慢于流量、快于客户数。区分这两种状态要靠财务序列，公开的使用量指标做不到。","consensus":"36% 的收入增长、35% 的 current RPO 增长和 1,700% 的 AI agent 请求增速，市场都已经看到。","differentiated":"把 1,700% 换成一个可以被证伪的问题：一次 agent 动作平均穿过几个收费产品，这个数字在变大还是变小。公开数据只能间接回答，可用的是收入增速、current RPO 增速、8% 的自由现金流率和 4,698 家大客户，四条序列同时改善才构成密度上升的证据。如果请求量继续翻倍而这四条不动，结论就是机器流量停在了低价或未计费的层里。","valuation":"情景用 Q2 收入年化后作基数（696.1 × 4），把增长和倍数都写成算式。年化单季会忽略季节性和公司自己的全年节奏，所以它是敏感性，不是指引，也不含每股结论。","update":"v2.0 保留 2026-08-07 发布时点，asOf 2026-09-14，纳入 DBNR、agentic 流量占比与 9 月催化。","evidenceClaimIds":["CL-NET-001","CL-NET-002","CL-NET-003","CL-NET-006"],"causalChain":[{"title":"一个任务放大成很多请求","body":"agent 反复调用工具和抓取，机器调用量与人类会话脱钩。","claimIds":["CL-NET-003"]},{"title":"调用穿过几个收费层","body":"Workers、安全、身份和网络产品决定单位活动的收入。","claimIds":["CL-NET-004"]},{"title":"合同和现金流验收","body":"current RPO 与自由现金流率必须跟上流量，否则密度只是叙事。","claimIds":["CL-NET-001","CL-NET-005"]}],"riskClaimIds":["CL-NET-007"],"forwardTestIds":["FAL-NET-001","FAL-NET-002","FAL-NET-003"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Cloudflare: Majority traffic is already machine","question":"Can machine-request growth cross more paid layers and register in contracted backlog and cash flow?","standfirst":"The 1,700% figure establishes that the traffic arrived. It says nothing about revenue per machine action: Cloudflare discloses no attributable revenue, no paid-request mix and no cross-product attach rate. The variable to underwrite is monetization density.","whyItMatters":"An agent finishing a task calls tools, fetches pages and re-authenticates far more often than a human session would. That cuts two ways for Cloudflare. Machine requests can compound inside layers that are barely priced, or the same action can pick up Workers compute, a security check, an identity verification and network egress. In the first state the traffic chart moves and the revenue line does not. In the second, revenue can trail raw traffic while running well ahead of customer count. Only the financial series separates the two; the published usage metric cannot.","consensus":"36% revenue growth, 35% current-RPO growth and 1,700% growth in AI-agent requests are all visible.","differentiated":"Restate the 1,700% as something falsifiable: how many paid products the average agent action touches, and whether that count is rising. The public data answers only indirectly. What is available is revenue growth, current-RPO growth, an 8% free-cash-flow margin and 4,698 large customers, and those four have to improve together before density counts as evidence. If requests keep doubling while the four sit still, the traffic is landing in layers that are cheap or unmetered.","valuation":"The scenarios annualize Q2 revenue as a base (696.1 × 4) and put growth and multiple into explicit arithmetic. Annualizing a single quarter ignores seasonality and the company's own full-year cadence, which is why the output is a sensitivity rather than guidance, and why no per-share conclusion follows.","update":"Version 2.0 keeps the August 7 publication date, asOf September 14, and adds DBNR, agentic traffic mix and September catalysts.","evidenceClaimIds":["CL-NET-001","CL-NET-002","CL-NET-003","CL-NET-006"],"causalChain":[{"title":"One task becomes many requests","body":"Agents call tools and fetch repeatedly, decoupling machine call volume from human sessions.","claimIds":["CL-NET-003"]},{"title":"Calls cross paid layers","body":"Workers, security, identity and network products set the revenue earned per action.","claimIds":["CL-NET-004"]},{"title":"Contracts and cash settle it","body":"Current RPO and free-cash-flow margin have to follow the traffic, or density stays a story.","claimIds":["CL-NET-001","CL-NET-005"]}],"riskClaimIds":["CL-NET-007"],"forwardTestIds":["FAL-NET-001","FAL-NET-002","FAL-NET-003"]}},"dataGaps":[{"zh-CN":"公司没有披露 AI agent 请求对应的收入和付费请求占比，也没有给跨产品 attach rate，因此 monetization density 只能通过收入、current RPO 和自由现金流率间接推断。","en":"Cloudflare discloses no revenue attributable to AI-agent requests, no paid-request mix and no cross-product attach rate, so monetization density can only be inferred indirectly from revenue, current RPO and free-cash-flow margin."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/cloudflare-monetization-density-2026q2","en":"https://lunartuliplab.com/en/deep-dive/cloudflare-monetization-density-2026q2"},"machineReadable":"https://lunartuliplab.com/research/cloudflare-monetization-density-2026q2","dateModified":"2026-09-14"},{"id":"RO-COMP-TEAM-001","slug":"atlassian-workflow-density-context-monetization-2026fy","kind":"company-deep-dive","tickers":["TEAM"],"publishedAt":"2026-08-07","asOf":"2026-09-14","version":"2.0.0","sourceLabel":"Atlassian Rovo monetization window · 2026-09-14 edition","claims":[{"id":"CL-TEAM-001","type":"Fact","text":{"zh-CN":"Atlassian FY2026 第四财季收入 17.66469 亿美元，同比增长约 28%；全年收入 65.72308 亿美元，同比增长 26%。","en":"Atlassian reported FY2026 fourth-quarter revenue of $1.766469 billion, up roughly 28%, and full-year revenue of $6.572308 billion, up 26%."},"evidenceIds":["EV-TEAM-001"]},{"id":"CL-TEAM-002","type":"Fact","text":{"zh-CN":"截至 2026 年 6 月 30 日，Subscription ARR 为 66.06 亿美元，同比增长 23%。","en":"Subscription ARR stood at $6.606 billion as of June 30, 2026, up 23%."},"evidenceIds":["EV-TEAM-001"]},{"id":"CL-TEAM-003","type":"Fact","text":{"zh-CN":"Atlassian 在 SEC 文件中把产品组合描述为互联应用、AI agents 与 Collections，并说明它们共享同一套 Cloud Platform 和数据模型。这是架构描述，公司没有给出对应的定价或收入拆分。","en":"In its SEC disclosure Atlassian describes the portfolio as connected apps, AI agents and Collections sharing one Cloud Platform and data model. That is an architecture description; no corresponding pricing or revenue split is provided."},"evidenceIds":["EV-TEAM-002"]},{"id":"CL-TEAM-004","type":"Hypothesis","text":{"zh-CN":"AI coding 可能减少一部分人类席位，同时制造更多任务、评审、依赖和治理对象，TEAM 的增长函数因此可能从 seat 数扩展到 workflow density 与组织上下文。公司未披露 agent 活动量或上下文收费口径，这一条无法从外部直接检验。","en":"AI coding may reduce some human seats while creating more tasks, reviews, dependencies and governance objects, so TEAM's growth function could broaden from seat count toward workflow density and organizational context. Atlassian discloses no agent-activity volume or context-pricing basis, which leaves the hypothesis untestable from outside."},"evidenceIds":["EV-TEAM-001","EV-TEAM-002"]},{"id":"CL-TEAM-005","type":"Inference","text":{"zh-CN":"管理层给出的 Rovo 分组对比没有落到收费口径上，agent 活动量、Rovo 的付费口径和上下文对应收入仍未披露。若 Subscription ARR 增速继续落在收入增速之下，而可核验的证据只有这类分组相关性，最合理的解读仍是席位压缩快于新计费单位的形成。","en":"The Rovo cohort comparison management supplied does not reach a billing basis: agent-activity volume, a paid-usage basis for Rovo and context-attributable revenue all remain undisclosed. If subscription ARR keeps running below revenue growth while cohort correlation stays the only verifiable evidence, the most reasonable reading is still that seat compression is arriving faster than a new billing unit."},"evidenceIds":["EV-TEAM-001","EV-TEAM-003"]},{"id":"CL-TEAM-006","type":"Fact","text":{"zh-CN":"Atlassian 在 FY2026 Q4 股东信中披露，采用 Rovo 的客户 ARR 增速持续超过未采用客户的 2 倍。公司没有给出分组定义、样本构建和调整方法，披露的是两组客户之间的同期相关性，不是 Rovo 自身的收入或因果增量。","en":"In its FY2026 Q4 shareholder letter, Atlassian disclosed that Rovo adopters continue to grow ARR more than twice as fast as non-adopters. No cohort definition, sample construction or adjustment was provided. What is disclosed is a contemporaneous correlation between two groups of customers, not revenue from Rovo or a causal increment."},"evidenceIds":["EV-TEAM-003"]},{"id":"CL-TEAM-007","type":"Fact","text":{"zh-CN":"Atlassian FY27 指引：总收入同比约 +13.0%，Cloud 收入约 +25.5%，订阅 ARR 约 +18.0%；FY26 FCF ,319.1M，FCF 利润率 20.1%，同比下滑 700 个基点。","en":"Atlassian FY2027 guidance calls for total revenue growth of about +13.0%, cloud revenue growth of about +25.5% and subscription ARR growth of about +18.0%; FY2026 FCF was ,319.1M at a 20.1% margin, down 700 basis points year over year."},"evidenceIds":["EV-TEAM-001"]},{"id":"CL-TEAM-008","type":"Inference","text":{"zh-CN":"2026 年 9 月 11 日收 $179.70、市值约 $45.60B，已进入 MarketBeat 共识目标价 $183.81 与 stockanalysis $196.23 区间的下沿；FY27 指引降速与 Rovo 无金额口径意味着二次点火需要 Q1 财报披露。","en":"On September 11, 2026 the stock closed at $179.70 with a market cap of about $45.60B, entering the lower band of the MarketBeat consensus target of $183.81 and the stockanalysis target of $196.23; FY2027 deceleration guidance and the absence of a Rovo revenue line mean the next catalyst needs Q1 disclosure."},"evidenceIds":["EV-TEAM-004","EV-TEAM-001"]}],"evidence":[{"id":"EV-TEAM-001","title":"Atlassian FY2026 Q4 financial results","source":{"publisher":"U.S. Securities and Exchange Commission","url":"https://www.sec.gov/Archives/edgar/data/1650372/000165037226000031/ex991q4fy26.htm","document":"Exhibit 99.1, Q4 and FY2026 results"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":"Q4 growth = 1766.469 / 1384.344 - 1 = 27.60%","confidence":"high","counterevidence":"Reported revenue includes migration and pricing effects; it does not isolate AI monetization.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-TEAM-002","title":"Atlassian FY2026 Q3 Form 10-Q description of business","source":{"publisher":"U.S. Securities and Exchange Commission","url":"https://www.sec.gov/Archives/edgar/data/1650372/000165037226000027/R9.htm","document":"FY2026 Q3 Form 10-Q, Description of Business"},"sourceDate":"2026-05-01","dataAsOf":"2026-03-31","calculation":null,"confidence":"high","counterevidence":"Product architecture does not prove incremental willingness to pay.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-TEAM-003","title":"Atlassian FY2026 Q4 shareholder letter — Rovo adopter cohort metric","source":{"publisher":"U.S. Securities and Exchange Commission","url":"https://www.sec.gov/Archives/edgar/data/1650372/000165037226000031/teamq42026shareholderlet.htm","document":"FY2026 Q4 shareholder letter, Exhibit 99.2"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"Atlassian did not disclose cohort definitions, sample construction or controls; faster ARR growth among adopters may reflect selection bias and does not isolate Rovo's causal contribution.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-TEAM-004","title":"MarketBeat TEAM consensus snapshot","source":{"publisher":"MarketBeat","url":"https://www.marketbeat.com/instant-alerts/analyst-atlassian-nasdaq-team-upgraded-by-wall-street-zen-to-buy-rating-2026-09-13/","document":"TEAM analyst snapshot"},"sourceDate":"2026-09-13","dataAsOf":"2026-09-11","calculation":null,"confidence":"medium","counterevidence":"Consensus target prices aggregate heterogeneous analyst models.","lastVerified":"2026-09-14","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"FY2026 Q4 收入","en":"FY2026 Q4 revenue"},"period":"2026-FY-Q4","value":1766.469,"unit":"USDm","evidenceIds":["EV-TEAM-001"]},{"label":{"zh-CN":"FY2026 全年收入","en":"FY2026 revenue"},"period":"2026-FY","value":6572.308,"unit":"USDm","evidenceIds":["EV-TEAM-001"]},{"label":{"zh-CN":"Subscription ARR","en":"Subscription ARR"},"period":"2026-06-30","value":6606,"unit":"USDm","evidenceIds":["EV-TEAM-001"]},{"label":{"zh-CN":"Subscription ARR 同比增速","en":"Subscription ARR growth (YoY)"},"period":"2026-06-30","value":23,"unit":"percent","evidenceIds":["EV-TEAM-001"]},{"label":{"zh-CN":"FY26 FCF 利润率","en":"FY2026 FCF margin"},"period":"2026-FY","value":20.1,"unit":"percent","evidenceIds":["EV-TEAM-001"]},{"label":{"zh-CN":"FY27 总收入增速指引","en":"FY2027 total revenue growth guide"},"period":"2027-FY","value":13,"unit":"percent","evidenceIds":["EV-TEAM-001"]}],"valuationScenarios":[{"id":"VAL-TEAM-BASE","label":{"zh-CN":"上下文形成计费单位","en":"Context becomes billable"},"revenueBaseUsdM":6572.308,"revenueGrowthPct":22,"forwardRevenueUsdM":8018.21576,"salesMultiple":8,"impliedEnterpriseValueUsdM":64145.72608,"calculation":"6572.308 * 1.22 = 8018.21576; 8018.21576 * 8 = 64145.72608","interpretation":{"zh-CN":"假设组织上下文变成新的收费对象，全年收入按 22% 增长、以 8 倍 sales multiple 计价。这是敏感性输出。公司披露的 Rovo 分组对比方向一致，但没有给出能支撑这条路径的收入分项。","en":"Assumes organizational context becomes a billable object, with full-year revenue growing 22% at an 8× sales multiple. The figure is a sensitivity output. Atlassian's Rovo cohort comparison points the same way, but supplies no revenue line item that would support the path."}},{"id":"VAL-TEAM-BEAR","label":{"zh-CN":"席位压缩占上风","en":"Seat compression wins"},"revenueBaseUsdM":6572.308,"revenueGrowthPct":12,"forwardRevenueUsdM":7360.98496,"salesMultiple":5,"impliedEnterpriseValueUsdM":36804.9248,"calculation":"6572.308 * 1.12 = 7360.98496; 7360.98496 * 5 = 36804.9248","interpretation":{"zh-CN":"人类席位减少快于 agent 与上下文收费形成时，增长退到 12%、倍数退到 5 倍，企业价值约为基准情景的 57%。","en":"If human seats contract faster than agent and context monetization forms, growth falls to 12% and the multiple to 5×, leaving enterprise value near 57% of the base case."}}],"counterevidenceClaimIds":["CL-TEAM-008"],"falsifiers":[{"id":"FAL-TEAM-001","metric":"subscription ARR YoY growth","operator":"<","threshold":15,"unit":"percent","horizon":"any of the next two reported quarters","claimIds":["CL-TEAM-004"],"rationale":{"zh-CN":"Subscription ARR 比确认收入更接近经常性口径，受迁移时点的影响更小。它跌破 15%，说明 workflow 扩张没有抵消席位压力和成熟 SaaS 的自然减速。","en":"Subscription ARR sits closer to a recurring measure than recognized revenue and is less exposed to migration timing. Below 15%, workflow expansion is not offsetting seat pressure and the natural deceleration of a mature SaaS base."}},{"id":"FAL-TEAM-002","metric":"cloud revenue YoY growth","operator":"<","threshold":18,"unit":"percent","horizon":"any of the next two reported quarters","claimIds":["CL-TEAM-004"],"rationale":{"zh-CN":"Cloud 是共享数据模型和 AI agents 实际运行的地方。Cloud 增速跌破 18%，共享上下文带来增量价值捕获这条判断就失去了承载它的收入线。","en":"Cloud is where the shared data model and the AI agents actually run. Cloud growth below 18% would leave the shared-context argument without a revenue line to carry it."}},{"id":"FAL-TEAM-003","metric":"total revenue YoY growth","operator":"<","threshold":13,"unit":"percent","horizon":"FY2027","claimIds":["CL-TEAM-007"],"rationale":{"zh-CN":"全年收入增速低于公司 +13% 指引，说明 RPO 转化与 Cloud 加速不足以抵消 Data Center 迁移拖累。","en":"Full-year revenue growth below the company's +13% guide would show RPO conversion and cloud acceleration failing to offset data-center migration drag."}}],"versions":[{"version":"1.0.0","date":"2026-08-15","diff":{"zh-CN":"从原合并页拆出独立对象，删除对 short squeeze 贡献的精确归因、技术价位与仓位内容。Rovo 采用组 ARR 增速 2 倍已核到 FY2026 Q4 股东信，改记为管理层披露的 cohort 相关性，方法未披露和选择偏差两项限制保留。","en":"Split out of the combined page as a standalone object, with the precise short-squeeze attribution, technical levels and sizing content removed. The Rovo 2× ARR-growth figure was traced to the FY2026 Q4 shareholder letter and is now recorded as a management-disclosed cohort correlation, with the undisclosed-method and selection-bias limitations retained."}},{"version":"2.0.0","date":"2026-09-14","diff":{"zh-CN":"2026-09-14 更新：纳入 FY27 降速指引、Rovo 无独立收入口径、FCF 利润率下滑与共识目标价区间。","en":"September 14 update: adds FY2027 deceleration guidance, no standalone Rovo revenue basis, lower FCF margin and the consensus target-price band."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"Atlassian：三重催化之后等口径","question":"人类 seat 承压之后，机器生成的任务、依赖和治理对象能不能变成新的计费单位？","standfirst":"FY2026 收入增长 26%，Subscription ARR 增长 23%。ARR 慢于收入这一点值得追问：协作软件的计费单位一直是人头，而 AI coding 同时压缩人头、制造工作状态。","whyItMatters":"协作软件二十年来的收入函数是席位数乘单价。AI coding 从两端同时动它：一端可能减少需要账号的工程师，另一端制造更多 pull request、评审、依赖关系和需要留痕的决策。这些机器生成的工作状态如果能变成收费对象，TEAM 的增长制度会换一次；如果不能，席位压缩会直接落在 ARR 上。管理层目前给出的分组对比停在客户层面，还没有落到能分开这两种结果的收费口径上。","consensus":"FY2026 收入增长 26%、Subscription ARR 增长 23%，股东信里还有一条 Rovo 采用客户 ARR 增速超过未采用客户 2 倍；产品组合正在围绕共享 Cloud Platform、AI agents 和 Collections 重组。","differentiated":"收入函数可能从人类 seat 扩到 workflow density（工作流密度）和组织上下文。管理层给出的唯一客户级证据，是 Rovo 采用组的 ARR 增速超过未采用组 2 倍，方向站在这条路径一边，但它承担不了结论：分组定义、样本构建和控制变量都没有披露，而采用顺序本身就有偏，企业软件里最先上新功能的，通常是预算和人员都在扩张的账户。口径上它衡量的是账户整体 ARR，账户扩张也可能来自席位、版本升级或 Cloud 迁移，与 Rovo 自身的收入并不是一回事。真正能推进判断的替代指标还是 Subscription ARR 和 Cloud 收入增速。算术上要提醒一句：26% 是全年收入增速，23% 是时点 ARR 增速，口径不同，单期差值说明不了什么，方向持续偏离才说明问题。","valuation":"两个情景都以 FY2026 的 65.72308 亿美元收入为基数，一个假设上下文变现成立，一个假设席位压缩占上风，只输出企业价值敏感性。倍数假设设在 8 倍和 5 倍，比另外两个对象低，因为基数是一家全年增速 26% 的成熟 SaaS；它们仍然是外生假设。","update":"v2.0 保留 2026-08-07 发布时点，asOf 2026-09-14，纳入 FY27 指引与卖方共识区间位置。","evidenceClaimIds":["CL-TEAM-001","CL-TEAM-002","CL-TEAM-006","CL-TEAM-007"],"causalChain":[{"title":"agent 制造工作状态","body":"更多代码和内容带来任务、评审和需要留痕的决策。","claimIds":["CL-TEAM-004"]},{"title":"共享上下文提高迁移成本","body":"应用、agents 和 Collections 跑在同一个平台与数据模型上。","claimIds":["CL-TEAM-003"]},{"title":"ARR 给出答案","body":"Subscription ARR 决定新增对象是否抵消了席位压力，分组对比替代不了这条总量序列。","claimIds":["CL-TEAM-002","CL-TEAM-005"]}],"riskClaimIds":["CL-TEAM-008"],"forwardTestIds":["FAL-TEAM-001","FAL-TEAM-002","FAL-TEAM-003"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Atlassian: After three catalysts, waiting for a revenue line","question":"As human seats come under pressure, can machine-created tasks, dependencies and governance objects become a billing unit?","standfirst":"FY2026 revenue grew 26%; subscription ARR grew 23%. That gap is worth interrogating. Collaboration software has always billed headcount, and AI coding compresses headcount while manufacturing work state.","whyItMatters":"For two decades the revenue function in collaboration software has been seats times price. AI coding pushes on both sides of it. Fewer engineers may need a licence, while pull requests, reviews, dependency chains and decisions that need an audit trail all multiply. If that machine-created work state becomes billable, TEAM's growth regime changes. If it does not, seat compression lands directly in ARR. The cohort comparison management has published stops at the account level and does not yet reach a billing basis that would separate the two outcomes.","consensus":"FY2026 revenue up 26%, subscription ARR up 23%, a shareholder-letter datapoint that Rovo adopters grow ARR more than twice as fast as non-adopters, and a portfolio being reorganized around a shared Cloud Platform, AI agents and Collections.","differentiated":"The revenue function may broaden from human seats toward workflow density and organizational context. The only customer-level evidence management has offered is that Rovo adopters grow ARR more than twice as fast as non-adopters. The direction favours the path, but the datapoint cannot carry the conclusion: no cohort definition, no sample construction, no controls, and adoption order is itself biased, because the accounts that take new functionality first are usually the ones already adding budget and headcount. The measure is account-level ARR, and an expanding account may be expanding on seats, edition upgrades or cloud migration rather than on Rovo. The proxies that actually move the judgment remain subscription ARR and cloud revenue growth. One caution on the arithmetic: 26% is a full-year revenue rate and 23% is a point-in-time ARR rate. A single-period gap between them proves little, while a persistent divergence would.","valuation":"Both scenarios start from FY2026 revenue of $6,572.308 million, one assuming context monetization works and one assuming seat compression wins, and output enterprise-value sensitivity only. The multiples are set at 8× and 5×, lower than in the other two objects because the base is a mature SaaS business growing 26%. They remain exogenous assumptions.","update":"Version 2.0 keeps the August 7 publication date, asOf September 14, and adds FY2027 guidance and consensus target positioning.","evidenceClaimIds":["CL-TEAM-001","CL-TEAM-002","CL-TEAM-006","CL-TEAM-007"],"causalChain":[{"title":"Agents manufacture work state","body":"More code and content produce tasks, reviews and decisions that need a record.","claimIds":["CL-TEAM-004"]},{"title":"Shared context raises switching cost","body":"Apps, agents and Collections run on one platform and one data model.","claimIds":["CL-TEAM-003"]},{"title":"ARR gives the answer","body":"Subscription ARR decides whether the new objects offset seat pressure; a cohort comparison cannot stand in for the aggregate series.","claimIds":["CL-TEAM-002","CL-TEAM-005"]}],"riskClaimIds":["CL-TEAM-008"],"forwardTestIds":["FAL-TEAM-001","FAL-TEAM-002","FAL-TEAM-003"]}},"dataGaps":[{"zh-CN":"FY2026 Q4 股东信披露 Rovo 采用客户的 ARR 增速超过未采用客户的 2 倍，但没有给出分组定义、样本构建和调整方法。采用 Rovo 的账户很可能本来就在扩张，这种同期分组对比分不开选择偏差；口径上它衡量的也是账户整体 ARR，不是 Rovo 自身的收入。要把它升级为因果证据，需要分组阈值与时点、可比对照组，以及 Rovo 的收入或付费口径。","en":"The FY2026 Q4 shareholder letter discloses that Rovo adopters grow ARR more than twice as fast as non-adopters, without cohort definitions, sample construction or adjustments. Accounts adopting Rovo are plausibly the ones already expanding, so a contemporaneous cohort comparison cannot separate selection bias, and the measure captures total account ARR rather than revenue from Rovo itself. Upgrading it to causal evidence would require the adoption threshold and timing, a matched control group, and either Rovo revenue or a disclosed paid-usage basis."},{"zh-CN":"公司尚未披露 agent 活动、workflow objects 或组织上下文对应的独立收入，也没有给 Rovo 的付费口径。","en":"Atlassian does not yet disclose standalone revenue tied to agent activity, workflow objects or organizational context, nor any paid-usage basis for Rovo."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/atlassian-workflow-density-context-monetization-2026fy","en":"https://lunartuliplab.com/en/deep-dive/atlassian-workflow-density-context-monetization-2026fy"},"machineReadable":"https://lunartuliplab.com/research/atlassian-workflow-density-context-monetization-2026fy","dateModified":"2026-09-14"},{"id":"RO-THEME-AI-VALUE-001","slug":"cloudflare-atlassian-ai-application-commercialization-2026q2","kind":"theme-study","tickers":["PLTR","NET","TEAM","DDOG","APP"],"publishedAt":"2026-08-07","asOf":"2026-08-07","version":"2.0.0","sourceLabel":"AI usage to value capture cross-sectional study · 2026-08-07 edition","claims":[{"id":"CL-THEME-001","type":"Derived","text":{"zh-CN":"PLTR、NET 与 TEAM 各自对应一条价值捕获路径：企业商业预算、机器流量穿过的收费层、组织协同上下文。三条路径的证据成熟度并不相同。","en":"PLTR, NET and TEAM each correspond to one capture path: commercial budgets, paid layers crossed by machine traffic, and organizational coordination context. The evidence behind the three is at different stages of maturity."},"evidenceIds":["EV-THEME-001","EV-THEME-002","EV-THEME-003"]},{"id":"CL-THEME-002","type":"Fact","text":{"zh-CN":"控制组中，DDOG 2026 年第二季度收入 11.21454 亿美元，同比增长 36%，自由现金流 2.79 亿美元。","en":"In the control group, DDOG reported Q2 2026 revenue of $1.121454 billion, up 36%, with free cash flow of $279 million."},"evidenceIds":["EV-THEME-004"]},{"id":"CL-THEME-003","type":"Fact","text":{"zh-CN":"另一控制组 APP 2026 年第二季度收入 19.24 亿美元，同比增长 53%，调整后 EBITDA 16.14 亿美元。调整后 EBITDA 是非 GAAP 指标。","en":"The second control, APP, reported Q2 2026 revenue of $1.924 billion, up 53%, and adjusted EBITDA of $1.614 billion. Adjusted EBITDA is a non-GAAP measure."},"evidenceIds":["EV-THEME-005"]},{"id":"CL-THEME-004","type":"Hypothesis","text":{"zh-CN":"AI usage 只有穿过一个客户认可的收费单位，并在收入、合同余额或现金流里留下痕迹，才构成可持续的 value capture。五家公司都没有按统一口径披露 AI 收入，这一条只能用间接指标检验。","en":"AI usage becomes durable value capture only when it crosses a billing unit the customer accepts and leaves a trace in revenue, contracted backlog or cash flow. None of the five companies discloses AI revenue on a common basis, so the proposition can only be tested through indirect measures."},"evidenceIds":["EV-THEME-001","EV-THEME-002","EV-THEME-003","EV-THEME-004","EV-THEME-005"]},{"id":"CL-THEME-005","type":"Inference","text":{"zh-CN":"DDOG 与 APP 说明与 AI 相关的增长可以同时完成较强的现金转化。核心组若长期只报告 adoption 或流量、财务传导达不到控制组的水平，主题判断应当降级为个股逻辑。","en":"DDOG and APP show that AI-linked growth can coexist with strong cash conversion. If the core cohort keeps reporting adoption or traffic while its financial transmission stays short of the controls, the theme should be downgraded to company-specific logic."},"evidenceIds":["EV-THEME-004","EV-THEME-005"]}],"evidence":[{"id":"EV-THEME-001","title":"PLTR canonical company object","source":{"publisher":"LunarTulip Research","document":"RO-COMP-PLTR-001"},"sourceDate":"2026-08-05","dataAsOf":"2026-08-05","calculation":null,"confidence":"high","counterevidence":"PLTR may be company-specific rather than representative of the application layer.","lastVerified":"2026-08-15","role":"supporting"},{"id":"EV-THEME-002","title":"NET canonical company object","source":{"publisher":"LunarTulip Research","document":"RO-COMP-NET-001"},"sourceDate":"2026-08-07","dataAsOf":"2026-08-07","calculation":null,"confidence":"high","counterevidence":"Agent requests have no disclosed revenue attribution.","lastVerified":"2026-08-15","role":"supporting"},{"id":"EV-THEME-003","title":"TEAM canonical company object","source":{"publisher":"LunarTulip Research","document":"RO-COMP-TEAM-001"},"sourceDate":"2026-08-07","dataAsOf":"2026-08-07","calculation":null,"confidence":"high","counterevidence":"Workflow-density revenue is not separately disclosed.","lastVerified":"2026-08-15","role":"supporting"},{"id":"EV-THEME-004","title":"Datadog Q2 2026 financial results","source":{"publisher":"Datadog Investor Relations","url":"https://investors.datadoghq.com/news-releases/news-release-details/datadog-announces-second-quarter-2026-financial-results/","document":"Q2 2026 earnings release"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"The release does not quantify revenue directly attributable to AI workloads.","lastVerified":"2026-08-15","role":"primary"},{"id":"EV-THEME-005","title":"AppLovin Q2 2026 financial results","source":{"publisher":"AppLovin Investor Relations","url":"https://investors.applovin.com/news/news-details/2026/AppLovin-Announces-Second-Quarter-2026-Financial-Results/default.aspx","document":"Q2 2026 earnings release"},"sourceDate":"2026-08-05","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"APP's advertising model is not directly comparable with enterprise software.","lastVerified":"2026-08-15","role":"primary"}],"financialBridge":[{"label":{"zh-CN":"PLTR Q2 收入增速","en":"PLTR Q2 revenue growth"},"period":"2026-Q2","value":93,"unit":"percent","evidenceIds":["EV-THEME-001"]},{"label":{"zh-CN":"NET Q2 收入增速","en":"NET Q2 revenue growth"},"period":"2026-Q2","value":36,"unit":"percent","evidenceIds":["EV-THEME-002"]},{"label":{"zh-CN":"TEAM FY2026 收入增速","en":"TEAM FY2026 revenue growth"},"period":"2026-FY","value":26,"unit":"percent","evidenceIds":["EV-THEME-003"]},{"label":{"zh-CN":"DDOG Q2 自由现金流（控制组）","en":"DDOG Q2 free cash flow (control)"},"period":"2026-Q2","value":279,"unit":"USDm","evidenceIds":["EV-THEME-004"]},{"label":{"zh-CN":"APP Q2 调整后 EBITDA（控制组）","en":"APP Q2 adjusted EBITDA (control)"},"period":"2026-Q2","value":1614,"unit":"USDm","evidenceIds":["EV-THEME-005"]}],"valuationScenarios":[{"id":"VAL-THEME-CAPTURE","label":{"zh-CN":"价值捕获扩散","en":"Capture broadens"},"revenueBaseUsdM":100,"revenueGrowthPct":30,"forwardRevenueUsdM":130,"salesMultiple":12,"impliedEnterpriseValueUsdM":1560,"calculation":"Normalized cohort revenue index: 100 * 1.30 = 130; 130 * 12 = 1560","interpretation":{"zh-CN":"标准化指数：基数 100，增长 30%，12 倍 sales multiple。它只用于比较两种制度，既不代表五家公司的市值之和，也不对应其中任何一家。","en":"A normalized index: base 100, 30% growth, a 12× sales multiple. It exists to compare two regimes, and represents neither the combined value of the five companies nor any one of them."}},{"id":"VAL-THEME-USAGE","label":{"zh-CN":"使用量停在 adoption","en":"Usage stays adoption"},"revenueBaseUsdM":100,"revenueGrowthPct":15,"forwardRevenueUsdM":115,"salesMultiple":7,"impliedEnterpriseValueUsdM":805,"calculation":"Normalized cohort revenue index: 100 * 1.15 = 115; 115 * 7 = 805","interpretation":{"zh-CN":"增长 15%、7 倍倍数。使用量没有穿过收费单位时，增长和倍数一起降档，指数落到另一种情景的约 52%。","en":"15% growth at a 7× multiple. When usage does not cross a billing unit, growth and the multiple reset together, leaving the index near 52% of the other case."}}],"counterevidenceClaimIds":["CL-THEME-005"],"falsifiers":[{"id":"FAL-THEME-001","metric":"core cohort median reported revenue growth","operator":"<","threshold":20,"unit":"percent","horizon":"next two common reporting cycles","claimIds":["CL-THEME-004"],"rationale":{"zh-CN":"口径取中位数，这样 PLTR 的高增速无法单独抬高整组读数。PLTR、NET、TEAM 的收入增速中位数跌到 20% 以下，说明价值捕获没有横向扩散。","en":"The median is used instead of the mean so that PLTR's rate cannot lift the cohort on its own. A median below 20% across PLTR, NET and TEAM would show that capture is not broadening."}},{"id":"FAL-THEME-002","metric":"companies with improving growth or cash-conversion evidence","operator":"<","threshold":3,"unit":"count","horizon":"next two common reporting cycles","claimIds":["CL-THEME-004"],"rationale":{"zh-CN":"五家里少于三家在收入增速、合同指标或现金转化上出现改善，横截面就失去了统计含量，主题退回个股逻辑。","en":"If fewer than three of the five improve on revenue growth, contract indicators or cash conversion, the cross-section loses whatever statistical content it had and the theme reverts to company-specific stories."}}],"versions":[{"version":"1.0.0","date":"2026-08-07","diff":{"zh-CN":"原页面把 NET 与 TEAM 合并，讨论 AI 应用商业化的扩散。","en":"The original page combined NET and TEAM to argue that AI application monetization was broadening."}},{"version":"2.0.0","date":"2026-08-15","diff":{"zh-CN":"升级为 Theme Study：核心组 PLTR、NET、TEAM，控制组 DDOG、APP，并移除技术面与仓位内容。","en":"Upgraded to a Theme Study with PLTR, NET and TEAM as the core cohort and DDOG and APP as controls, and with technical and positioning content removed."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"从 AI usage 到 value capture：五家公司的横截面检验","question":"使用量增长在什么条件下会变成可归因的收入、合同余额和现金流？","standfirst":"PLTR、NET、TEAM 停在同一条路径的三个位置：预算已经确认、收费单位待验证、计费口径本身受质疑。DDOG 与 APP 作为控制组，给出传导完成之后财务上该长什么样。","whyItMatters":"adoption 指标几乎可以自动上涨：请求数、活跃用户、模型调用量都随部署扩大而增加，报出来不需要客户多付一分钱。收费单位是另一回事，它要求客户接受一个新的计量对象，并在合同或账单上认下来。把五家公司放在同一个问题下比较，能把公司特有的爆发和产业层面的扩散分开。单看 PLTR，很难判断 149% 是产业信号还是一家公司的合同周期；把 NET 和 TEAM 摆在旁边，路径差异就出来了。","consensus":"AI 使用量在快速增长，多家公司收入提速，这些都已经被接受。","differentiated":"把分析单位从 adoption 换成收费单位和财务收货点：预算（PLTR）、单次动作穿过的收费层（NET）、协同与治理对象（TEAM）。三者的证据走到的距离不同，PLTR 的合同与确认收入已经同步，NET 手上只有使用量，TEAM 连计费口径都还在变化。控制组给出对照的量级：DDOG 当季 11.21454 亿美元收入对应 2.79 亿美元自由现金流，APP 19.24 亿美元收入对应 16.14 亿美元调整后 EBITDA。两家都在增长的同时完成了现金转化，也都没有披露可归因于 AI 的收入。核心组要么走到类似的财务质量，要么这个主题退回三个独立的公司故事。","valuation":"主题层用标准化收入指数（基数 100）比较增长和倍数两种制度，不合成五家公司的市值，也不对任何一家给估值结论。指数的用途只有一个：30% 增长配 12 倍，与 15% 增长配 7 倍之间差了接近一倍，说明制度切换的代价来自增长和倍数同时移动。","update":"v2.0 保留 2026-08-07 的时点，把原来的 NET+TEAM 合并页升级为含两个控制组的 Theme Study，并移除技术面和仓位内容。","evidenceClaimIds":["CL-THEME-001","CL-THEME-002","CL-THEME-003"],"causalChain":[{"title":"使用量进入生产环境","body":"预算审批、机器调用和工作对象是三个可观察的入口。","claimIds":["CL-THEME-001"]},{"title":"收费单位被客户接受","body":"客户要为决策、单次动作或组织上下文签下一个计量口径。","claimIds":["CL-THEME-004"]},{"title":"财务报表给出裁决","body":"收入、合同余额和现金转化决定价值是否真的被捕获。","claimIds":["CL-THEME-002","CL-THEME-003","CL-THEME-005"]}],"riskClaimIds":["CL-THEME-005"],"forwardTestIds":["FAL-THEME-001","FAL-THEME-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"From AI usage to value capture: a five-company cross-section","question":"Under what conditions does usage growth become attributable revenue, contracted backlog and cash flow?","standfirst":"PLTR, NET and TEAM sit at three points on the same path: budget confirmed, billing unit unproven, billing basis itself in question. DDOG and APP serve as controls, showing what completed transmission looks like in the financials.","whyItMatters":"Adoption metrics can rise almost mechanically. Requests, active users and model calls all grow as deployment widens, and reporting them costs the customer nothing. A billing unit is a different matter: it requires the customer to accept a new unit of measurement and sign for it. Holding five companies against one question separates company-specific breakouts from industry diffusion. On its own, 149% at PLTR could be an industry signal or one company's contract cycle. Placed next to NET and TEAM, the difference in path becomes legible.","consensus":"Rapid AI-usage growth and revenue acceleration at several companies are already accepted.","differentiated":"Move the unit of analysis from adoption to the billing unit and the point of financial receipt: budgets at PLTR, paid layers crossed per action at NET, coordination and governance objects at TEAM. The evidence has travelled different distances. PLTR's contracts and recognized revenue already move together, NET has usage only, and at TEAM the billing basis itself is still shifting. The controls set the comparison: DDOG turned $1,121.454 million of quarterly revenue into $279 million of free cash flow, and APP turned $1,924 million into $1,614 million of adjusted EBITDA. Both grew while converting to cash, and neither disclosed AI-attributable revenue. The core cohort either reaches comparable financial quality or the theme reduces to three separate company stories.","valuation":"The theme layer uses a normalized revenue index based at 100 to compare two regimes of growth and multiple, without aggregating market values or reaching a valuation conclusion on any single name. The index does one job: 30% growth at 12× sits close to twice 15% growth at 7×, which shows the cost of a regime change comes from growth and the multiple moving at once.","update":"Version 2.0 preserves the August 7 information set, upgrades the original NET+TEAM page into a Theme Study with two controls, and removes technical and positioning content.","evidenceClaimIds":["CL-THEME-001","CL-THEME-002","CL-THEME-003"],"causalChain":[{"title":"Usage reaches production","body":"Budget approval, machine calls and work objects are three observable entry points.","claimIds":["CL-THEME-001"]},{"title":"A billing unit gets accepted","body":"The customer signs for a unit of measurement: a decision, an action, or a piece of organizational context.","claimIds":["CL-THEME-004"]},{"title":"The financials rule on it","body":"Revenue, backlog and cash conversion decide whether the value was actually captured.","claimIds":["CL-THEME-002","CL-THEME-003","CL-THEME-005"]}],"riskClaimIds":["CL-THEME-005"],"forwardTestIds":["FAL-THEME-001","FAL-THEME-002"]}},"dataGaps":[{"zh-CN":"五家公司没有按统一口径披露 AI 相关收入，因此本研究对每家使用最接近价值捕获的自有财务指标，横向可比性也因此受限。","en":"The five companies do not disclose AI-related revenue on a common basis, so the study uses each company's closest own-financial proxy for capture, and cross-sectional comparability is limited accordingly."},{"zh-CN":"DDOG 与 APP 只作为传导机制的控制组，不构成业务模式或估值可比公司；APP 的广告模式与企业软件的收费逻辑差别很大。","en":"DDOG and APP are controls for the transmission mechanism, not business-model or valuation comparables. APP's advertising model prices very differently from enterprise software."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/cloudflare-atlassian-ai-application-commercialization-2026q2","en":"https://lunartuliplab.com/en/deep-dive/cloudflare-atlassian-ai-application-commercialization-2026q2"},"machineReadable":"https://lunartuliplab.com/research/cloudflare-atlassian-ai-application-commercialization-2026q2","dateModified":"2026-08-15"},{"id":"RO-THEME-AI-HARDWARE-001","slug":"global-ai-hardware-profit-pools-2026-09","kind":"theme-study","tickers":["000660.KS","MU","005930.KS","285A.T","SNDK","001309.SZ","301308.SZ","300308.SZ","300502.SZ","300394.SZ","LITE","COHR","AAOI"],"publishedAt":"2026-09-02","asOf":"2026-09-01","version":"1.0.0","sourceLabel":"全球 AI 硬件利润池横截面研究（存储 · 光模块 · 人形机器人）· 2026-09-02 版｜Global AI hardware profit-pool cross-section (memory · optics · humanoid robotics) · 2026-09-02 edition","claims":[{"id":"CL-HWX-001","type":"Fact","text":{"zh-CN":"2026-09-01 TrendForce 的 DDR5 16Gb 现货均价为 54.083 美元，当日上涨 0.31%；2026-08-24 更新的 DDR5 32GB RDIMM 模组均价周涨 4.06%；2026-08-17 当周中国市场 DDR5 16Gb 与 24Gb 报价周涨 14.29%。以上均为报价序列，同期成交量与需求确认仍然偏弱。","en":"On 2026-09-01 the TrendForce DDR5 16Gb spot average was $54.083, up 0.31% on the day; the DDR5 32GB RDIMM module average updated 2026-08-24 rose 4.06% week on week; and in the week of 2026-08-17 China-market DDR5 16Gb and 24Gb quotes rose 14.29% week on week. All of these are quoted-price series, and transaction volume and demand confirmation over the same period remained weak."},"evidenceIds":["EV-HWX-TRENDFORCE"]},{"id":"CL-HWX-002","type":"Fact","text":{"zh-CN":"韩国产业通商资源部披露的 DDR5 16Gb 月度固定价从 2026 年 4 月的 35.0 美元升至 8 月的 46.5 美元，其中 7 月环比 +12.5%、8 月环比 +3.3%；NAND 128Gb 固定价月涨幅从 5 月的 9.5% 降至 8 月的 1.3%。同一份数据显示韩国 2026 年 8 月半导体出口 466.5 亿美元，同比增长 209%。","en":"Korea's Ministry of Trade, Industry and Energy reports the DDR5 16Gb monthly fixed price rising from $35.0 in April 2026 to $46.5 in August, with month-on-month increases of 12.5% in July and 3.3% in August. The monthly increase in NAND 128Gb fixed prices fell from 9.5% in May to 1.3% in August. The same release shows Korean semiconductor exports of $46.65 billion in August 2026, up 209% year over year."},"evidenceIds":["EV-HWX-MOTIE"]},{"id":"CL-HWX-003","type":"Derived","text":{"zh-CN":"现货报价与月度固定价方向相同、加速度相反：现货二阶导向上，合约价二阶导向下（7 月 +12.5% → 8 月 +3.3%）。把两条线合并成单一“存储涨价”指标，会同时高估模组利润的持续时间并低估上游原厂的定价能力。","en":"Spot quotes and monthly fixed prices move in the same direction with opposite acceleration: the second derivative of spot is rising while the second derivative of contract prices is falling (+12.5% in July to +3.3% in August). Collapsing the two into a single 'memory prices are up' indicator overstates the duration of module profit and understates upstream pricing power at the same time."},"evidenceIds":["EV-HWX-TRENDFORCE","EV-HWX-MOTIE"]},{"id":"CL-HWX-004","type":"Fact","text":{"zh-CN":"NVIDIA FY27Q2 收入 962 亿美元，其中数据中心收入 890 亿美元；公司披露的未来供应与产能承诺（supply and capacity commitments）从上一季的 1190 亿美元提高到 2790 亿美元，并说明主要涉及存储采购。该金额是未来承诺义务，不是已付款采购，条款上部分可调整或可延期，也不等于任一供应商的确定性收入。","en":"NVIDIA reported FY27Q2 revenue of $96.2 billion, including $89.0 billion of data-centre revenue, and disclosed future supply and capacity commitments rising from $119.0 billion in the prior quarter to $279.0 billion, noting they relate principally to memory procurement. The figure is a forward commitment obligation rather than paid procurement, parts of it are adjustable or deferrable under their terms, and it does not equate to contracted revenue for any single supplier."},"evidenceIds":["EV-HWX-NVDA"]},{"id":"CL-HWX-005","type":"Fact","text":{"zh-CN":"SK 海力士 2026 年第二季度收入 79.3 万亿韩元，环比增长 51%；营业利润 60.5 万亿韩元，营业利润率 76%，营业利润同比增长 557%。当季 93.9 万亿韩元净利润中约 63.3 万亿来自投资资产收益。公司同时披露 HBM4 已量产出货、321 层 NAND 年底计划升至韩国本土产能约一半、净现金 69.4 万亿韩元。","en":"SK hynix reported Q2 2026 revenue of KRW 79.3 trillion, up 51% sequentially, with operating profit of KRW 60.5 trillion at a 76% margin and operating profit up 557% year over year. Roughly KRW 63.3 trillion of the KRW 93.9 trillion in net income came from gains on investment assets. The company also disclosed volume shipments of HBM4, a plan to lift 321-layer NAND to about half of Korean domestic capacity by year-end, and net cash of KRW 69.4 trillion."},"evidenceIds":["EV-HWX-SKHYNIX","EV-HWX-REUTERS-SKH"]},{"id":"CL-HWX-006","type":"Fact","text":{"zh-CN":"美光 FY2026 第三财季收入 414.6 亿美元，环比增长 73.7%；GAAP 营业利润 333.2 亿美元，营业利润率 80.4%；经营现金流 253.9 亿美元。Cloud Memory 与 Core Data Center 两个业务单元当季营业利润率分别为 78% 和 83%；公司 FQ4 指引为收入 500 亿美元、毛利率约 86%。","en":"Micron reported fiscal 2026 third-quarter revenue of $41.46 billion, up 73.7% sequentially, with GAAP operating profit of $33.32 billion at an 80.4% margin and operating cash flow of $25.39 billion. The Cloud Memory and Core Data Center units posted quarterly operating margins of 78% and 83%, and FQ4 guidance calls for $50.0 billion of revenue at roughly 86% gross margin."},"evidenceIds":["EV-HWX-MU"]},{"id":"CL-HWX-007","type":"Fact","text":{"zh-CN":"三星电子 2026 年第二季度总收入 171.5 万亿韩元；半导体（DS）部门收入 127.5 万亿韩元、营业利润 89.2 万亿韩元。集团口径包含手机、显示与代工等非存储业务。","en":"Samsung Electronics reported Q2 2026 group revenue of KRW 171.5 trillion, with the Device Solutions segment contributing KRW 127.5 trillion of revenue and KRW 89.2 trillion of operating profit. The group basis includes handsets, displays and foundry alongside memory."},"evidenceIds":["EV-HWX-SAMSUNG"]},{"id":"CL-HWX-008","type":"Fact","text":{"zh-CN":"Kioxia FY2026 第一财季收入 1.767 万亿日元，环比增长 76.2%；非 GAAP 营业利润 1.326 万亿日元，非 GAAP 营业利润率 75%。非 GAAP 口径由公司定义，与 GAAP 结果不同。","en":"Kioxia reported fiscal 2026 first-quarter revenue of JPY 1.767 trillion, up 76.2% sequentially, with non-GAAP operating profit of JPY 1.326 trillion at a 75% non-GAAP operating margin. The non-GAAP basis is company-defined and differs from the GAAP result."},"evidenceIds":["EV-HWX-KIOXIA"]},{"id":"CL-HWX-009","type":"Fact","text":{"zh-CN":"SanDisk FY2026 第四财季收入 89.7 亿美元，环比增长 51%，其中约三分之二来自价格；数据中心收入环比增长 103%；GAAP 营业利润 70.4 亿美元。","en":"SanDisk reported fiscal 2026 fourth-quarter revenue of $8.97 billion, up 51% sequentially with roughly two-thirds of the increase from price, data-centre revenue up 103% sequentially, and GAAP operating profit of $7.04 billion."},"evidenceIds":["EV-HWX-SNDK"]},{"id":"CL-HWX-010","type":"Fact","text":{"zh-CN":"德明利 2026 年上半年收入 169.1 亿元，同比增长 311.6%；归母净利润 60.2 亿元，综合毛利率 47.1%；同期经营活动现金流 -67.4 亿元，存货 214.1 亿元，占总资产 65.0%。","en":"Demingli reported first-half 2026 revenue of RMB 16.91 billion, up 311.6% year over year, with attributable net profit of RMB 6.02 billion at a 47.1% blended gross margin. Operating cash flow was RMB -6.74 billion and inventory stood at RMB 21.41 billion, or 65.0% of total assets."},"evidenceIds":["EV-HWX-DEMINGLI"]},{"id":"CL-HWX-011","type":"Fact","text":{"zh-CN":"江波龙 2026 年上半年收入 240.9 亿元，同比增长 136.3%；归母净利润 105.8 亿元；同期经营活动现金流 -31.5 亿元，存货 257.8 亿元，占总资产 60.1%。","en":"Longsys reported first-half 2026 revenue of RMB 24.09 billion, up 136.3% year over year, with attributable net profit of RMB 10.58 billion. Operating cash flow was RMB -3.15 billion and inventory stood at RMB 25.78 billion, or 60.1% of total assets."},"evidenceIds":["EV-HWX-LONGSYS"]},{"id":"CL-HWX-012","type":"Fact","text":{"zh-CN":"中际旭创 2026 年上半年收入 417.8 亿元，同比增长 182.5%；归母净利润 136.5 亿元，同比增长 241.7%；上半年经营活动现金流与归母净利润之比为 0.13。","en":"Innolight reported first-half 2026 revenue of RMB 41.78 billion, up 182.5% year over year, and attributable net profit of RMB 13.65 billion, up 241.7%. First-half operating cash flow equalled 0.13 times attributable net profit."},"evidenceIds":["EV-HWX-INNOLIGHT"]},{"id":"CL-HWX-013","type":"Fact","text":{"zh-CN":"新易盛 2026 年上半年收入 209.1 亿元，同比增长 100.3%；归母净利润 75.3 亿元，同比增长 91.0%；上半年经营活动现金流与归母净利润之比为 0.21。","en":"Eoptolink reported first-half 2026 revenue of RMB 20.91 billion, up 100.3% year over year, and attributable net profit of RMB 7.53 billion, up 91.0%. First-half operating cash flow equalled 0.21 times attributable net profit."},"evidenceIds":["EV-HWX-EOPTOLINK"]},{"id":"CL-HWX-014","type":"Fact","text":{"zh-CN":"天孚通信 2026 年上半年收入 28.3 亿元，同比增长 15.2%；归母净利润 12.0 亿元，同比增长 33.9%；上半年经营活动现金流与归母净利润之比为 0.83。","en":"T&S Communications reported first-half 2026 revenue of RMB 2.83 billion, up 15.2% year over year, and attributable net profit of RMB 1.20 billion, up 33.9%. First-half operating cash flow equalled 0.83 times attributable net profit."},"evidenceIds":["EV-HWX-TFC"]},{"id":"CL-HWX-015","type":"Fact","text":{"zh-CN":"Lumentum FY2026 第四财季收入 10.06 亿美元，同比增长 109.3%；非 GAAP 营业利润率 36.6%，下一季非 GAAP 营业利润率指引 39.5%—40.5%，下一季收入指引中值约 12.5 亿美元。公司 GAAP 净利润受一次性、非现金债务清偿损失影响，因此经营比较采用其披露的非 GAAP 口径。","en":"Lumentum reported fiscal 2026 fourth-quarter revenue of $1.006 billion, up 109.3% year over year, with a 36.6% non-GAAP operating margin, next-quarter non-GAAP operating margin guidance of 39.5%–40.5% and a next-quarter revenue guidance midpoint near $1.25 billion. GAAP net income is affected by a one-time, non-cash loss on debt extinguishment, so operating comparisons use the company's disclosed non-GAAP basis."},"evidenceIds":["EV-HWX-LUMENTUM"]},{"id":"CL-HWX-016","type":"Fact","text":{"zh-CN":"Coherent FY2026 第四财季收入 20.46 亿美元，同比增长 33.8%，数据中心与通信业务同比增长 58.6%；非 GAAP 营业利润率 21.8%。全财年经营活动现金流 7,950 万美元，资本开支 11.03 亿美元，存货从 14.38 亿美元升至 25.81 亿美元；下一季收入指引中值约 23 亿美元。","en":"Coherent reported fiscal 2026 fourth-quarter revenue of $2.046 billion, up 33.8% year over year, with datacom and communications up 58.6% and a 21.8% non-GAAP operating margin. Full-year operating cash flow was $79.5 million against capex of $1.103 billion, inventory rose from $1.438 billion to $2.581 billion, and the next-quarter revenue guidance midpoint is near $2.3 billion."},"evidenceIds":["EV-HWX-COHERENT"]},{"id":"CL-HWX-017","type":"Fact","text":{"zh-CN":"AAOI 2026 年第二季度收入 1.92 亿美元，同比增长约 86.4%；第三季度收入指引中值约 2.73 亿美元。当季非 GAAP 转为盈利 550 万美元，GAAP 口径仍为亏损。","en":"AAOI reported second-quarter 2026 revenue of $192 million, up roughly 86.4% year over year, and guided to a third-quarter revenue midpoint near $273 million. The quarter produced $5.5 million of non-GAAP profit while remaining loss-making on a GAAP basis."},"evidenceIds":["EV-HWX-AAOI"]},{"id":"CL-HWX-018","type":"Derived","text":{"zh-CN":"按各自指引中值计算，Lumentum、Coherent 与 AAOI 下一财季收入分别对应约 24%、12% 和 42% 的环比增长（12.5/10.06、23/20.46、2.73/1.92）。三家同时给出环比增长指引，说明光模块的需求斜率在 2026-09-01 时点尚未转向。","en":"On their own guidance midpoints, Lumentum, Coherent and AAOI imply roughly 24%, 12% and 42% sequential revenue growth for the coming fiscal quarter (12.5/10.06, 23/20.46, 2.73/1.92). All three guiding to sequential growth indicates the optics demand slope had not turned as of 2026-09-01."},"evidenceIds":["EV-HWX-LUMENTUM","EV-HWX-COHERENT","EV-HWX-AAOI"]},{"id":"CL-HWX-019","type":"Derived","text":{"zh-CN":"本对象使用的“峰值盈利年化倍数”为市值除以最近季度（或半年）利润的简单年化，属于 2026-09-01 收盘时点的快照筛查，不是 forward PE，也不是价格结论。美国公司的年化倍数分母是非 GAAP 净利润，与并列的非 GAAP 营业利润率是两条不同口径，不可互换；中国公司使用归母净利润。市值与 52 周回撤未与完全稀释股数逐一核对，因此不进入事实层。","en":"The peak-earnings annualised multiples used in this object divide market capitalisation by a simple annualisation of the latest quarterly (or half-year) profit. They are point-in-time screens at the 2026-09-01 close, not forward P/E and not price conclusions. For the U.S. names the denominator is non-GAAP net income, which is a different basis from the non-GAAP operating margin quoted alongside it and cannot be substituted for it; the Chinese names use attributable net profit. Market capitalisations and 52-week drawdowns were not reconciled to fully diluted share counts and therefore stay out of the fact layer."},"evidenceIds":["EV-HWX-MARKET","EV-HWX-MU","EV-HWX-KIOXIA","EV-HWX-SNDK"]},{"id":"CL-HWX-020","type":"Derived","text":{"zh-CN":"全球存储链可以拆成三层性质不同的利润：约束所有者（SK 海力士、美光、三星，拥有 HBM、服务器 DRAM 与企业级 NAND 的产能、工艺与客户认证）、NAND 周期弹性（Kioxia、SanDisk，直接暴露于 NAND 价格与数据中心 SSD）、库存价差（德明利、江波龙，利润大量来自低成本库存与成品售价之间的价差）。三层可以同时受益，证伪条件完全不同。","en":"The global memory chain separates into three profit layers of different character: constraint owners (SK hynix, Micron, Samsung, holding the capacity, process and customer qualification behind HBM, server DRAM and enterprise NAND); NAND cycle elasticity (Kioxia and SanDisk, directly exposed to NAND pricing and data-centre SSD); and inventory spread (Demingli and Longsys, where much of the profit sits between low-cost inventory and realised selling prices). All three can benefit at once, and their falsification conditions are entirely different."},"evidenceIds":["EV-HWX-SKHYNIX","EV-HWX-MU","EV-HWX-SAMSUNG","EV-HWX-KIOXIA","EV-HWX-SNDK","EV-HWX-DEMINGLI","EV-HWX-LONGSYS"]},{"id":"CL-HWX-021","type":"Hypothesis","text":{"zh-CN":"同一轮 AI 资本开支内部，结构性利润会优先并持续留在约束所有者手里，因为供给紧张最先进入它们的 ASP、毛利与现金流。9—11 月的核心变量因此是利润能否继续留在资产负债表更干净、客户锁定更强的公司。各公司未按统一口径披露 HBM 或企业级产品的分部利润，这一条只能用营业利润率、现金转化与合约价加速度间接检验。","en":"Within one AI capex cycle, structural profit should accrue first and persist with the constraint owners, because tight supply reaches their ASPs, gross margin and cash flow before anyone else's. The core variable from September to November is therefore whether profit keeps landing with the companies carrying cleaner balance sheets and tighter customer lock-in. None of these companies discloses segment profit for HBM or enterprise product on a common basis, so the proposition can only be tested indirectly through operating margin, cash conversion and the acceleration of contract prices."},"evidenceIds":["EV-HWX-SKHYNIX","EV-HWX-MU","EV-HWX-MOTIE","EV-HWX-NVDA"]},{"id":"CL-HWX-022","type":"Inference","text":{"zh-CN":"德明利与江波龙 7—8 倍的静态年化倍数，分母含有低成本库存、急涨 ASP 与较低减值计提共同形成的峰值利润。若 Q3 毛利率继续下滑、经营现金流继续净流出、存货占资产与周转继续恶化，表面低 PE 只是周期分母过高，而不是估值折价。","en":"The 7–8x static annualised multiples at Demingli and Longsys rest on a denominator containing peak profit formed jointly by low-cost inventory, sharply rising ASPs and light impairment. If Q3 gross margin keeps falling, operating cash flow keeps running negative and inventory-to-assets and turnover keep deteriorating, the apparent low P/E is an inflated cyclical denominator rather than a valuation discount."},"evidenceIds":["EV-HWX-DEMINGLI","EV-HWX-LONGSYS","EV-HWX-MOTIE"]},{"id":"CL-HWX-023","type":"Inference","text":{"zh-CN":"光模块的估值分水岭已经从增长转向现金转化。中际旭创与新易盛上半年 CFO ÷ 归母净利润只有 0.13 与 0.21，天孚通信为 0.83；Coherent 全年经营现金流 7,950 万美元对应 11.03 亿美元资本开支。在 30—100 倍以上的 run-rate 要求下，现金回收延迟会迅速放大估值波动。","en":"The valuation dividing line in optics has moved from growth to cash conversion. First-half operating cash flow covered only 0.13 and 0.21 of attributable net profit at Innolight and Eoptolink against 0.83 at T&S Communications, and Coherent generated $79.5 million of full-year operating cash flow against $1.103 billion of capex. Against run-rate multiples of 30x to 100x and above, delayed cash recovery amplifies valuation volatility quickly."},"evidenceIds":["EV-HWX-INNOLIGHT","EV-HWX-EOPTOLINK","EV-HWX-TFC","EV-HWX-COHERENT"]},{"id":"CL-HWX-024","type":"Inference","text":{"zh-CN":"人形机器人在全球公共市场缺少可与 HBM、1.6T 光模块并列的盈利载体：领先整机公司多数未上市或被大型集团包裹，上市零部件公司很少单列人形机器人收入。在公告级订单、单列收入与毛利率同时出现之前，它只能作为长期技术期权，不能进入同口径的利润池比较。","en":"Humanoid robotics lacks a profit vehicle in global public markets comparable to HBM or 1.6T optics: most leading whole-machine developers are private or held inside large groups, and listed component suppliers rarely disclose humanoid revenue as a separate line. Until announcement-grade orders, separately disclosed revenue and gross margin appear together, it remains a long-dated technology option rather than a member of a like-for-like profit-pool comparison."},"evidenceIds":["EV-HWX-HUMANOID-GAP"]},{"id":"CL-HWX-025","type":"Inference","text":{"zh-CN":"反方情景是超额利润率被供给侧正常化拉低：长期协议提高需求可见度的同时限制现货上涨对原厂 ASP 的额外弹性，2027 年新增产能进一步压缩价差。此外有媒体报道称中国厂商 CXMT 已开始小批量生产 HBM3E，该报道未获公司确认，本对象不据此推导任何价格或份额影响，仅记为待验证的供给侧风险。","en":"The bear case is supply-side normalisation compressing excess margin: long-term agreements raise demand visibility while capping the incremental ASP elasticity manufacturers capture from rising spot, and 2027 capacity additions narrow the spread further. Separately, press reporting states that Chinese manufacturer CXMT has begun small-volume production of HBM3E. That report is not company-confirmed, no pricing or share effect is inferred from it here, and it is recorded only as an unverified supply-side risk."},"evidenceIds":["EV-HWX-SKHYNIX","EV-HWX-SAMSUNG","EV-HWX-MOTIE"]}],"evidence":[{"id":"EV-HWX-TRENDFORCE","title":"TrendForce DRAM spot and module quotes","source":{"publisher":"TrendForce","url":"https://www.trendforce.com/price/dram/dram_spot","document":"DRAM spot price board and 2026-08-17 channel price update"},"sourceDate":"2026-09-01","dataAsOf":"2026-09-01","calculation":null,"confidence":"medium","counterevidence":"These are quoted prices rather than confirmed transactions. Volume and demand confirmation stayed weak through this leg, so rising quotes cannot be read as verified demand.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-MOTIE","title":"Korea August 2026 export release with DRAM and NAND fixed prices","source":{"publisher":"Ministry of Trade, Industry and Energy, Republic of Korea","url":"https://www.motir.go.kr/kor/article/ATCL3f49a5a8c/172145/view","document":"2026 年 8 월 수출입 동향 / August 2026 trade release"},"sourceDate":"2026-09-01","dataAsOf":"2026-08-31","calculation":"Fixed-price deceleration: July +12.5% MoM, August +3.3% MoM on DDR5 16Gb; NAND 128Gb +9.5% MoM in May to +1.3% MoM in August.","confidence":"high","counterevidence":"Aggregate export value mixes logic, memory and packaging, and monthly fixed prices are a reference series rather than every manufacturer's realised ASP.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-NVDA","title":"NVIDIA FY2027 Q2 CFO commentary","source":{"publisher":"U.S. Securities and Exchange Commission","url":"https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm","document":"FY27Q2 CFO Commentary"},"sourceDate":"2026-08-26","dataAsOf":"2026-07-26","calculation":null,"confidence":"high","counterevidence":"Supply and capacity commitments are forward obligations that are partly adjustable or deferrable. They are neither paid procurement nor a locked demand schedule for any individual memory supplier.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-SKHYNIX","title":"SK hynix Q2 2026 business results","source":{"publisher":"SK hynix Investor Relations","url":"https://news.skhynix.com/en/q2-2026-business-results/","document":"Q2 2026 business results"},"sourceDate":"2026-07-28","dataAsOf":"2026-06-30","calculation":"Operating margin = KRW 60.5tn / KRW 79.3tn = 76.3%","confidence":"high","counterevidence":"Net income includes roughly KRW 63.3tn of gains on investment assets, so any net-income-based multiple is distorted and only operating profit describes the operating business.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-MU","title":"Micron fiscal 2026 third-quarter results","source":{"publisher":"Micron Technology Investor Relations","url":"https://www.globenewswire.com/news-release/2026/06/24/3317151/14450/en/micron-technology-inc-reports-record-results-for-the-third-quarter-of-fiscal-2026.html","document":"FQ3 FY2026 earnings release"},"sourceDate":"2026-06-24","dataAsOf":"2026-05-28","calculation":"GAAP operating margin = $33.32bn / $41.46bn = 80.4%","confidence":"high","counterevidence":"Margins of this level reflect an unusually tight pricing environment; guidance of roughly 86% gross margin has not yet been tested by a reported quarter.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-SAMSUNG","title":"Samsung Electronics Q2 2026 results","source":{"publisher":"Samsung Electronics Newsroom","url":"https://news.samsung.com/global/samsung-electronics-announces-second-quarter-2026-results","document":"Q2 2026 results announcement"},"sourceDate":"2026-07-31","dataAsOf":"2026-06-30","calculation":"DS operating margin = KRW 89.2tn / KRW 127.5tn = 70.0%","confidence":"high","counterevidence":"Group results blend handsets, displays and foundry with memory, so the consolidated multiple is not comparable with a pure-play memory manufacturer.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-REUTERS-SKH","title":"Reuters report on SK hynix Q2 profit and the market reaction","source":{"publisher":"Reuters","url":"https://www.reuters.com/world/asia-pacific/sk-hynix-q2-profit-jumps-557-ai-chip-demand-misses-forecasts-2026-07-28/","document":"SK hynix Q2 profit jumps 557% on AI chip demand, misses forecasts"},"sourceDate":"2026-07-28","dataAsOf":"2026-06-30","calculation":null,"confidence":"medium","counterevidence":"Press reporting of a single-session market reaction is not a first-party disclosure and does not establish a durable change in fundamentals.","lastVerified":"2026-09-02","role":"supporting"},{"id":"EV-HWX-KIOXIA","title":"Kioxia Holdings FY2026 Q1 results presentation","source":{"publisher":"Kioxia Holdings Investor Relations","url":"https://www.kioxia-holdings.com/en-jp/ir/library/presentation.html","document":"FY2026 Q1 results presentation"},"sourceDate":"2026-08-07","dataAsOf":"2026-06-30","calculation":"Non-GAAP operating margin = JPY 1.326tn / JPY 1.767tn = 75.0%","confidence":"medium","counterevidence":"The operating profit cited is a company-defined non-GAAP measure and is not directly comparable with SanDisk's GAAP operating profit.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-SNDK","title":"SanDisk fiscal 2026 fourth-quarter results","source":{"publisher":"U.S. Securities and Exchange Commission","url":"https://www.sec.gov/Archives/edgar/data/2023554/000162828026053346/sndkq4-26ex991xpressrelease.htm","document":"Exhibit 99.1, FQ4 FY2026 press release"},"sourceDate":"2026-08-05","dataAsOf":"2026-07-03","calculation":null,"confidence":"high","counterevidence":"About two-thirds of the sequential revenue increase came from price, so the same operating leverage works in reverse on any NAND ASP inflection.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-DEMINGLI","title":"德明利 2026 年半年度报告 / Demingli 2026 interim report","source":{"publisher":"巨潮资讯网 CNINFO（中国证监会指定信息披露平台）","url":"https://www.cninfo.com.cn/new/fulltextSearch?keyWord=%E5%BE%B7%E6%98%8E%E5%88%A9","document":"深圳市德明利技术股份有限公司 2026 年半年度报告（001309.SZ）"},"sourceDate":"2026-08-28","dataAsOf":"2026-06-30","calculation":"Operating cash flow / attributable net profit = -6.74 / 6.02 = -1.12; inventory share of total assets = 65.0%","confidence":"high","counterevidence":"Reported profit embeds inventory revaluation gains; the interim report does not separate spread-driven profit from operating profit on a sustainable product basis.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-LONGSYS","title":"江波龙 2026 年半年度报告 / Longsys 2026 interim report","source":{"publisher":"巨潮资讯网 CNINFO（中国证监会指定信息披露平台）","url":"https://www.cninfo.com.cn/new/fulltextSearch?keyWord=%E6%B1%9F%E6%B3%A2%E9%BE%99","document":"深圳市江波龙电子股份有限公司 2026 年半年度报告（301308.SZ）"},"sourceDate":"2026-08-28","dataAsOf":"2026-06-30","calculation":"Operating cash flow / attributable net profit = -3.15 / 10.58 = -0.30; inventory share of total assets = 60.1%","confidence":"high","counterevidence":"The same inventory-spread caveat applies, and enterprise-grade product revenue is not disclosed separately from consumer inventory gains.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-INNOLIGHT","title":"中际旭创 2026 年半年度报告 / Innolight 2026 interim report","source":{"publisher":"巨潮资讯网 CNINFO（中国证监会指定信息披露平台）","url":"https://static.cninfo.com.cn/finalpage/2026-08-22/1225491752.PDF","document":"中际旭创股份有限公司 2026 年半年度报告（300308.SZ）"},"sourceDate":"2026-08-22","dataAsOf":"2026-06-30","calculation":"Operating cash flow / attributable net profit = 0.13","confidence":"high","counterevidence":"Revenue and profit growth do not settle profit quality while cash conversion stays near 0.13; prepayments, inventory build and customer concentration are not resolved by the interim report.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-EOPTOLINK","title":"新易盛 2026 年半年度报告 / Eoptolink 2026 interim report","source":{"publisher":"巨潮资讯网 CNINFO（中国证监会指定信息披露平台）","url":"https://www.cninfo.com.cn/new/fulltextSearch?keyWord=%E6%96%B0%E6%98%93%E7%9B%9B","document":"新易盛股份有限公司 2026 年半年度报告（300502.SZ）"},"sourceDate":"2026-08-28","dataAsOf":"2026-06-30","calculation":"Operating cash flow / attributable net profit = 0.21","confidence":"high","counterevidence":"Attributable net profit grew more slowly than revenue, and the cash-conversion ratio remains well below the peer high in the same cohort.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-TFC","title":"天孚通信 2026 年半年度报告 / T&S Communications 2026 interim report","source":{"publisher":"巨潮资讯网 CNINFO（中国证监会指定信息披露平台）","url":"https://www.cninfo.com.cn/new/fulltextSearch?keyWord=%E5%A4%A9%E5%AD%9A%E9%80%9A%E4%BF%A1","document":"苏州天孚光通信股份有限公司 2026 年半年度报告（300394.SZ）"},"sourceDate":"2026-08-28","dataAsOf":"2026-06-30","calculation":"Operating cash flow / attributable net profit = 0.83","confidence":"high","counterevidence":"The strongest cash conversion in the cohort comes with the slowest revenue growth, so cash quality alone does not establish an attractive entry basis.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-LUMENTUM","title":"Lumentum fiscal 2026 fourth-quarter and full-year results","source":{"publisher":"Lumentum Investor Relations","url":"https://investor.lumentum.com/financial-news-releases/news-details/2026/Lumentum-Announces-Fourth-Quarter-and-Full-Fiscal-Year-2026-Results/default.aspx","document":"FQ4 and FY2026 results release"},"sourceDate":"2026-08-11","dataAsOf":"2026-06-27","calculation":"Guidance midpoint sequential growth = 12.5 / 10.06 - 1 = 24.3%","confidence":"high","counterevidence":"GAAP net income is distorted by a one-time, non-cash loss on debt extinguishment, so GAAP and non-GAAP results diverge sharply and cannot be mixed in one multiple.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-COHERENT","title":"Coherent fiscal 2026 fourth-quarter results","source":{"publisher":"Coherent Corp. Investor Relations","url":"https://www.coherent.com/content/dam/coherent/site/en/documents/investors/financial-releases/2026/august-12/earnings-release-fy26-q4.pdf","document":"FY26 Q4 earnings release"},"sourceDate":"2026-08-12","dataAsOf":"2026-06-30","calculation":"Guidance midpoint sequential growth = 23 / 20.46 - 1 = 12.4%; full-year operating cash flow $79.5m against $1,103m of capex","confidence":"high","counterevidence":"Free cash flow was deeply negative for the year, and the inventory build from $1.438bn to $2.581bn precedes any confirmation of customer acceptance.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-AAOI","title":"Applied Optoelectronics second-quarter 2026 results","source":{"publisher":"Applied Optoelectronics Investor Relations","url":"https://investors.ao-inc.com/news-releases/news-release-details/applied-optoelectronics-reports-second-quarter-2026-results","document":"Q2 2026 earnings release"},"sourceDate":"2026-08-06","dataAsOf":"2026-06-30","calculation":"Guidance midpoint sequential growth = 2.73 / 1.92 - 1 = 42.2%","confidence":"high","counterevidence":"The company remains loss-making on GAAP and the non-GAAP profit base is small, so earnings multiples carry no explanatory value and execution and financing risk stay elevated.","lastVerified":"2026-09-02","role":"primary"},{"id":"EV-HWX-MARKET","title":"Point-in-time market snapshot, 2026-09-01 close","source":{"publisher":"Google Finance, MarketWatch and exchange closing data","url":"https://www.google.com/finance/quote/MU%3ANASDAQ","document":"Screen quotes and market capitalisations captured at the 2026-09-01 close"},"sourceDate":"2026-09-01","dataAsOf":"2026-09-01","calculation":"Peak-earnings screen = market capitalisation / (latest-quarter net income x 4); H1 screen = market capitalisation / (H1 attributable net profit x 2)","confidence":"low","counterevidence":"These are screen quotes rather than a licensed point-in-time tape, and were not reconciled to fully diluted share counts, so every market capitalisation, screen P/E and 52-week drawdown here is approximate and stays out of the fact layer.","lastVerified":"2026-09-02","role":"supporting"},{"id":"EV-HWX-CONSENSUS","title":"Third-party sell-side consensus snapshot for MU","source":{"publisher":"Seeking Alpha","url":"https://seekingalpha.com/symbol/MU/earnings/revisions","document":"MU earnings estimates and revisions, captured 2026-09-01"},"sourceDate":"2026-09-01","dataAsOf":"2026-09-01","calculation":"Forward screen = 2026-09-01 price / vendor-aggregated FY2027 consensus EPS. The specific consensus figures are presented in the canonical editorial page only.","confidence":"low","counterevidence":"A vendor-aggregated sell-side consensus is not company guidance, moves with every print and revision, and differs across vendors on fiscal-year alignment and GAAP versus non-GAAP treatment. It is therefore excluded from the fact layer of this object.","lastVerified":"2026-09-02","role":"supporting"},{"id":"EV-HWX-HUMANOID-GAP","title":"Humanoid robotics segment-disclosure gap in global public markets","source":{"publisher":"Lunartulip Research","document":"Disclosure survey: no listed component supplier in the reviewed sample reports separately disclosed humanoid robotics revenue or gross margin"},"sourceDate":"2026-09-01","dataAsOf":"2026-09-01","calculation":null,"confidence":"low","counterevidence":"Absence of segment disclosure is not evidence that the business is absent; a company could be shipping meaningful volume without breaking it out, and private whole-machine developers are invisible to this survey entirely.","lastVerified":"2026-09-02","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"韩国半导体出口（2026 年 8 月）","en":"Korean semiconductor exports (August 2026)"},"period":"2026-08","value":46650,"unit":"USDm","displayValue":{"zh-CN":"466.5 亿美元，同比 +209%","en":"$46.65bn, +209% YoY"},"evidenceIds":["EV-HWX-MOTIE"]},{"label":{"zh-CN":"DDR5 16Gb 月度固定价环比（7 月）","en":"DDR5 16Gb fixed price, MoM (July)"},"period":"2026-07","value":12.5,"unit":"percent","evidenceIds":["EV-HWX-MOTIE"]},{"label":{"zh-CN":"DDR5 16Gb 月度固定价环比（8 月）","en":"DDR5 16Gb fixed price, MoM (August)"},"period":"2026-08","value":3.3,"unit":"percent","displayValue":{"zh-CN":"+3.3%（固定价 46.5 美元，4 月为 35.0 美元）","en":"+3.3% (fixed price $46.5, from $35.0 in April)"},"evidenceIds":["EV-HWX-MOTIE"]},{"label":{"zh-CN":"NAND 128Gb 月度固定价环比（8 月）","en":"NAND 128Gb fixed price, MoM (August)"},"period":"2026-08","value":1.3,"unit":"percent","displayValue":{"zh-CN":"+1.3%（5 月为 +9.5%）","en":"+1.3% (from +9.5% in May)"},"evidenceIds":["EV-HWX-MOTIE"]},{"label":{"zh-CN":"DDR5 16Gb 现货均价当日涨幅","en":"DDR5 16Gb spot average, daily change"},"period":"2026-09-01","value":0.31,"unit":"percent","displayValue":{"zh-CN":"+0.31%（均价 54.083 美元，报价口径）","en":"+0.31% (average $54.083, quoted basis)"},"evidenceIds":["EV-HWX-TRENDFORCE"]},{"label":{"zh-CN":"NVIDIA 未来供应与产能承诺","en":"NVIDIA forward supply and capacity commitments"},"period":"2026-FY27Q2","value":279000,"unit":"USDm","displayValue":{"zh-CN":"2,790 亿美元（上一季 1,190 亿美元）· 承诺义务，非已付款采购","en":"$279bn (from $119bn) · commitment obligation, not paid procurement"},"evidenceIds":["EV-HWX-NVDA"]},{"label":{"zh-CN":"SK 海力士 Q2 营业利润率","en":"SK hynix Q2 operating margin"},"period":"2026-Q2","value":76,"unit":"percent","displayValue":{"zh-CN":"76%（营业利润 60.5 万亿韩元 ÷ 收入 79.3 万亿韩元）","en":"76% (KRW 60.5tn operating profit / KRW 79.3tn revenue)"},"evidenceIds":["EV-HWX-SKHYNIX"]},{"label":{"zh-CN":"美光 FQ3 GAAP 营业利润率","en":"Micron FQ3 GAAP operating margin"},"period":"2026-FQ3","value":80.4,"unit":"percent","displayValue":{"zh-CN":"80.4%（营业利润 333.2 亿美元 ÷ 收入 414.6 亿美元）","en":"80.4% ($33.32bn operating profit / $41.46bn revenue)"},"evidenceIds":["EV-HWX-MU"]},{"label":{"zh-CN":"美光 FQ3 经营现金流","en":"Micron FQ3 operating cash flow"},"period":"2026-FQ3","value":25390,"unit":"USDm","displayValue":{"zh-CN":"253.9 亿美元","en":"$25.39bn"},"evidenceIds":["EV-HWX-MU"]},{"label":{"zh-CN":"三星 DS 部门 Q2 营业利润率","en":"Samsung DS Q2 operating margin"},"period":"2026-Q2","value":70,"unit":"percent","displayValue":{"zh-CN":"约 70%（营业利润 89.2 万亿韩元 ÷ 收入 127.5 万亿韩元）","en":"About 70% (KRW 89.2tn / KRW 127.5tn)"},"evidenceIds":["EV-HWX-SAMSUNG"]},{"label":{"zh-CN":"Kioxia FY26Q1 非 GAAP 营业利润率","en":"Kioxia FY26Q1 non-GAAP operating margin"},"period":"2026-FY26Q1","value":75,"unit":"percent","displayValue":{"zh-CN":"75%（非 GAAP 营业利润 1.326 万亿日元 ÷ 收入 1.767 万亿日元）","en":"75% (JPY 1.326tn non-GAAP operating profit / JPY 1.767tn revenue)"},"evidenceIds":["EV-HWX-KIOXIA"]},{"label":{"zh-CN":"SanDisk FQ4 收入","en":"SanDisk FQ4 revenue"},"period":"2026-FQ4","value":8970,"unit":"USDm","displayValue":{"zh-CN":"89.7 亿美元（环比 +51%，约三分之二来自价格）","en":"$8.97bn (+51% QoQ, about two-thirds from price)"},"evidenceIds":["EV-HWX-SNDK"]},{"label":{"zh-CN":"SanDisk FQ4 数据中心收入环比","en":"SanDisk FQ4 data-centre revenue, QoQ"},"period":"2026-FQ4","value":103,"unit":"percent","evidenceIds":["EV-HWX-SNDK"]},{"label":{"zh-CN":"德明利 H1 经营现金流 ÷ 归母净利润","en":"Demingli H1 operating cash flow ÷ net profit"},"period":"2026-H1","value":-112,"unit":"percent","displayValue":{"zh-CN":"-1.12×（经营现金流 -67.4 亿元 ÷ 归母净利润 60.2 亿元）","en":"-1.12x (RMB -6.74bn / RMB 6.02bn)"},"evidenceIds":["EV-HWX-DEMINGLI"]},{"label":{"zh-CN":"德明利 H1 存货占总资产","en":"Demingli H1 inventory as share of total assets"},"period":"2026-H1","value":65,"unit":"percent","displayValue":{"zh-CN":"65.0%（存货 214.1 亿元）","en":"65.0% (inventory RMB 21.41bn)"},"evidenceIds":["EV-HWX-DEMINGLI"]},{"label":{"zh-CN":"江波龙 H1 经营现金流 ÷ 归母净利润","en":"Longsys H1 operating cash flow ÷ net profit"},"period":"2026-H1","value":-29.8,"unit":"percent","displayValue":{"zh-CN":"-0.30×（经营现金流 -31.5 亿元 ÷ 归母净利润 105.8 亿元）","en":"-0.30x (RMB -3.15bn / RMB 10.58bn)"},"evidenceIds":["EV-HWX-LONGSYS"]},{"label":{"zh-CN":"江波龙 H1 存货占总资产","en":"Longsys H1 inventory as share of total assets"},"period":"2026-H1","value":60.1,"unit":"percent","displayValue":{"zh-CN":"60.1%（存货 257.8 亿元）","en":"60.1% (inventory RMB 25.78bn)"},"evidenceIds":["EV-HWX-LONGSYS"]},{"label":{"zh-CN":"中际旭创 H1 CFO ÷ 归母净利润","en":"Innolight H1 CFO ÷ attributable net profit"},"period":"2026-H1","value":13,"unit":"percent","displayValue":{"zh-CN":"0.13×","en":"0.13x"},"evidenceIds":["EV-HWX-INNOLIGHT"]},{"label":{"zh-CN":"新易盛 H1 CFO ÷ 归母净利润","en":"Eoptolink H1 CFO ÷ attributable net profit"},"period":"2026-H1","value":21,"unit":"percent","displayValue":{"zh-CN":"0.21×","en":"0.21x"},"evidenceIds":["EV-HWX-EOPTOLINK"]},{"label":{"zh-CN":"天孚通信 H1 CFO ÷ 归母净利润","en":"T&S Communications H1 CFO ÷ attributable net profit"},"period":"2026-H1","value":83,"unit":"percent","displayValue":{"zh-CN":"0.83×","en":"0.83x"},"evidenceIds":["EV-HWX-TFC"]},{"label":{"zh-CN":"Lumentum FQ4 非 GAAP 营业利润率","en":"Lumentum FQ4 non-GAAP operating margin"},"period":"2026-FQ4","value":36.6,"unit":"percent","displayValue":{"zh-CN":"36.6%（下一季指引 39.5%—40.5%）","en":"36.6% (guided to 39.5%–40.5%)"},"evidenceIds":["EV-HWX-LUMENTUM"]},{"label":{"zh-CN":"Coherent FY2026 经营现金流","en":"Coherent FY2026 operating cash flow"},"period":"2026-FY","value":79.5,"unit":"USDm","displayValue":{"zh-CN":"7,950 万美元（同期资本开支 11.03 亿美元）","en":"$79.5m (against $1,103m of capex)"},"evidenceIds":["EV-HWX-COHERENT"]},{"label":{"zh-CN":"AAOI Q2 收入","en":"AAOI Q2 revenue"},"period":"2026-Q2","value":192,"unit":"USDm","displayValue":{"zh-CN":"1.92 亿美元（同比约 +86.4%）","en":"$192m (about +86.4% YoY)"},"evidenceIds":["EV-HWX-AAOI"]}],"valuationScenarios":[{"id":"VAL-HWX-CONSTRAINT","label":{"zh-CN":"约束所有者：峰值盈利快筛","en":"Constraint owners: peak-earnings screen"},"metrics":[{"label":{"zh-CN":"峰值盈利年化倍数（美光）","en":"Peak-earnings annualised multiple (Micron)"},"value":{"zh-CN":"约 9.5×（市值 ÷ 最近季度 GAAP 净利润 × 4）","en":"About 9.5x (market cap ÷ latest-quarter GAAP net income x 4)"}},{"label":{"zh-CN":"行情端 TTM 市盈率","en":"Screen TTM P/E"},"value":{"zh-CN":"SK 海力士约 7.2×（分母含约 63.3 万亿韩元投资收益）；美光 21.1×；三星集团约 11.2×","en":"SK hynix about 7.2x (denominator includes roughly KRW 63.3tn of investment gains); Micron 21.1x; Samsung group about 11.2x"}},{"label":{"zh-CN":"较 52 周高点回撤","en":"Drawdown from 52-week high"},"value":{"zh-CN":"SK 海力士约 -43%；美光约 -26%；三星约 -33%","en":"SK hynix about -43%; Micron about -26%; Samsung about -33%"}},{"label":{"zh-CN":"口径限制","en":"Basis limitation"},"value":{"zh-CN":"2026-09-01 收盘快照；不是 forward PE，也不是价格结论","en":"2026-09-01 close snapshot; not a forward P/E and not a price conclusion"}}],"calculation":"Micron peak-earnings screen = market capitalisation / (latest-quarter GAAP net income x 4). Screen P/Es use trailing twelve-month reported net income without normalisation. All inputs captured at the 2026-09-01 close and not reconciled to fully diluted share counts.","interpretation":{"zh-CN":"个位数的峰值盈利快筛与十倍出头的行情市盈率，说明市场已经在计入利润显著正常化。这一组的研究问题因此是正常化之后的利润底部有多高，而不是当前倍数看起来是否便宜。三家公司会计口径、业务纯度与货币都不同，四个指标并列只用于观察市场已经计入多少，不构成跨市场可比估值。","en":"Single-digit peak-earnings screens alongside low-twenties trailing P/Es show the market already embedding a substantial normalisation of profit. The research question for this group is therefore how high the floor sits after normalisation, not whether today's multiple looks cheap. Accounting basis, business purity and currency all differ across the three names, so these metrics sit side by side to show what is priced in, not to supply a comparable cross-market valuation."}},{"id":"VAL-HWX-NAND","label":{"zh-CN":"NAND 周期弹性：Kioxia / SanDisk","en":"NAND cycle elasticity: Kioxia / SanDisk"},"metrics":[{"label":{"zh-CN":"峰值盈利年化倍数","en":"Peak-earnings annualised multiple"},"value":{"zh-CN":"Kioxia 约 7.9×（非 GAAP 口径）；SanDisk 约 8.7×（GAAP 口径）","en":"Kioxia about 7.9x (non-GAAP basis); SanDisk about 8.7x (GAAP basis)"}},{"label":{"zh-CN":"较 52 周高点回撤","en":"Drawdown from 52-week high"},"value":{"zh-CN":"Kioxia 约 -55%；SanDisk 约 -35%","en":"Kioxia about -55%; SanDisk about -35%"}},{"label":{"zh-CN":"利润来源","en":"Source of profit"},"value":{"zh-CN":"SanDisk 环比增长中约三分之二来自价格；数据中心收入环比 +103%","en":"About two-thirds of SanDisk's sequential growth came from price; data-centre revenue +103% QoQ"}},{"label":{"zh-CN":"口径限制","en":"Basis limitation"},"value":{"zh-CN":"两个分母的会计口径不同，不能直接相互比较","en":"The two denominators sit on different accounting bases and are not directly comparable"}}],"calculation":"Screen = market capitalisation / (latest-quarter net income x 4), using Kioxia's company-defined non-GAAP net income and SanDisk's GAAP net income. Market data captured at the 2026-09-01 close.","interpretation":{"zh-CN":"这两家的低倍数建立在极高的 NAND 价格与利润率之上，分母的周期位置比倍数本身更重要。NAND 每一次 ASP 反转，利润弹性都会双向工作；SanDisk 财报后的下跌说明高增长已经被计入。它们衡量的是周期弹性，与约束所有权是两类不同的研究对象。","en":"The low multiples here rest on extreme NAND pricing and margins, so where the denominator sits in the cycle matters more than the multiple itself. Every NAND ASP inflection works in both directions on profit, and SanDisk's post-print decline shows the growth is already in the price. These names measure cycle elasticity, which is a different research object from constraint ownership."}},{"id":"VAL-HWX-MODULE-CN","label":{"zh-CN":"中国存储模组：静态倍数与情景反向检验","en":"Chinese memory modules: static multiple and scenario back-test"},"metrics":[{"label":{"zh-CN":"静态年化倍数","en":"Static annualised multiple"},"value":{"zh-CN":"德明利约 7.9×；江波龙约 7.1×（市值 ÷ H1 归母净利润 × 2）","en":"Demingli about 7.9x; Longsys about 7.1x (market cap ÷ H1 attributable net profit x 2)"}},{"label":{"zh-CN":"现金与库存质量","en":"Cash and inventory quality"},"value":{"zh-CN":"CFO ÷ 净利润 -1.12× / -0.30×；存货占总资产 65.0% / 60.1%","en":"CFO ÷ net profit of -1.12x / -0.30x; inventory at 65.0% / 60.1% of total assets"}},{"label":{"zh-CN":"作者情景假设：市值升至 2026-09-01 快照的十倍","en":"Author scenario: market cap rises to ten times the 2026-09-01 snapshot"},"value":{"zh-CN":"德明利在 20× 下需约 474 亿元、25× 下需约 379 亿元可持续年利润；江波龙对应约 750 亿元与约 600 亿元","en":"Demingli would need about RMB 47.4bn of sustainable annual profit at 20x and about RMB 37.9bn at 25x; Longsys about RMB 75bn and about RMB 60bn"}},{"label":{"zh-CN":"情景性质","en":"Nature of the scenario"},"value":{"zh-CN":"作者设定的反向检验，不是预测、不是价格结论，也不含任何交易含义","en":"An author-defined back-test: not a forecast, not a price conclusion, and carrying no trading implication"}}],"calculation":"Static multiple: 94.8 / (6.02 x 2) = 7.9x; 150 / (10.58 x 2) = 7.1x (RMB bn, 2026-09-01 close). Scenario back-test: 94.8 x 10 = 948; 948 / 20 = 47.4; 948 / 25 = 37.9; 150 x 10 = 1500; 1500 / 20 = 75; 1500 / 25 = 60. Required profit of RMB 47.4bn is about four times the annualised H1 base of RMB 12.04bn.","interpretation":{"zh-CN":"这组算术只回答一个问题：当前的静态低倍数需要多大幅度的盈利结构变化才能被延续，而不是判断它会不会发生。所需利润量级需要企业级产品份额、主控能力、客户粘性与现金回收同时扩张，现货报价脉冲只能提供其中一段。分母中已包含低成本库存、急涨 ASP 与较低减值计提共同形成的峰值利润，因此表面便宜与安全边际不是一回事。","en":"This arithmetic answers one question: how much would the earnings structure have to change for today's low static multiple to be sustained. It is not a judgment on whether that happens. Profit of the required magnitude needs enterprise product share, controller capability, customer stickiness and cash recovery to expand together, and a spot-quote impulse supplies only part of the distance. The denominator already contains peak profit formed by cheap inventory, sharply rising ASPs and light impairment, so apparent cheapness and margin of safety are not the same thing."}},{"id":"VAL-HWX-OPTICS","label":{"zh-CN":"光模块：增长要求与现金转化","en":"Optics: growth demanded versus cash conversion"},"metrics":[{"label":{"zh-CN":"中国三家 H1 净利润年化倍数","en":"Annualised H1 net profit multiple, Chinese names"},"value":{"zh-CN":"中际旭创约 37×；新易盛约 37×；天孚通信约 119×","en":"Innolight about 37x; Eoptolink about 37x; T&S Communications about 119x"}},{"label":{"zh-CN":"美国两家最近季度非 GAAP 净利润年化倍数","en":"Annualised latest-quarter non-GAAP net income multiple, U.S. names"},"value":{"zh-CN":"Lumentum 约 64×；Coherent 约 38×（分母为非 GAAP 净利润，不是营业利润率）","en":"Lumentum about 64x; Coherent about 38x (denominator is non-GAAP net income, not operating margin)"}},{"label":{"zh-CN":"现金转化","en":"Cash conversion"},"value":{"zh-CN":"H1 CFO ÷ 净利润 0.13 / 0.21 / 0.83；Coherent 全年 CFO 7,950 万美元对应资本开支 11.03 亿美元","en":"H1 CFO ÷ net profit of 0.13 / 0.21 / 0.83; Coherent's $79.5m of full-year CFO against $1,103m of capex"}},{"label":{"zh-CN":"AAOI","en":"AAOI"},"value":{"zh-CN":"非 GAAP 盈利 550 万美元、GAAP 仍亏损，盈利基数过小，PE 不具解释力","en":"$5.5m of non-GAAP profit and still loss-making on GAAP; the base is too small for P/E to carry meaning"}}],"calculation":"Innolight: 10100 / (136.5 x 2) = 37.0x; Eoptolink: 5600 / (75.3 x 2) = 37.2x; T&S: 2870 / (12.0 x 2) = 119.6x (RMB 100m units, 2026-09-01 close). U.S. multiples divide market capitalisation by four times latest-quarter non-GAAP net income as disclosed by each company.","interpretation":{"zh-CN":"六家公司的产品结构、会计口径与资本强度都不同，把它们压成一个市盈率排名会抹掉全部有效信息。有意义的读法是把估值要求与现金转化放在一起：30—100 倍以上的 run-rate 要求，配 0.13—0.21 的现金转化，意味着利润质量的验证窗口很窄。天孚通信提供了同业里的现金上限参照，代价是最慢的增长与最高的估值基数。","en":"Product mix, accounting basis and capital intensity differ across all six names, and compressing them into one P/E ranking erases every piece of useful information. The productive reading pairs the valuation demanded with the cash actually converted: run-rate multiples of 30x to 100x and above against 0.13 to 0.21 cash conversion leaves a narrow window to validate profit quality. T&S Communications sets the cash-conversion ceiling within the cohort, at the cost of the slowest growth and the highest valuation base."}}],"counterevidenceClaimIds":["CL-HWX-022","CL-HWX-023","CL-HWX-024","CL-HWX-025"],"falsifiers":[{"id":"FAL-HWX-001","metric":"DDR5 16Gb monthly fixed price, month-on-month change","operator":"<","threshold":0,"unit":"percent","horizon":"any monthly fixed-price release from September to November 2026","claimIds":["CL-HWX-003","CL-HWX-021"],"rationale":{"zh-CN":"月度固定价的环比涨幅已经从 7 月的 12.5% 降到 8 月的 3.3%。一旦转负，上游 ASP 的加速度就不只是见顶，而是进入回落，约束所有权优先获得结构性利润这条前提会被直接削弱。阈值取 0 而不是某个正数，是为了让破位不依赖对减速幅度的主观判断。","en":"The month-on-month increase in fixed prices has already fallen from 12.5% in July to 3.3% in August. Turning negative would mean upstream ASP acceleration has not merely peaked but reversed, which directly weakens the premise that constraint ownership captures structural profit first. The threshold sits at zero rather than at some positive rate so that a break does not depend on a subjective view of how much deceleration counts."}},{"id":"FAL-HWX-002","metric":"Micron reported GAAP gross margin, FQ4 FY2026","operator":"<","threshold":80,"unit":"percent","horizon":"the FQ4 FY2026 result reported in September 2026","claimIds":["CL-HWX-006","CL-HWX-021"],"rationale":{"zh-CN":"公司自己的 FQ4 指引是约 86% 毛利率。阈值设在 80% 而不是 86%，留出约 6 个百分点的缓冲，避免把正常的指引偏差当成破位；真的跌到 80% 以下，意味着 ASP 或产品组合的兑现明显低于公司口径，约束所有者的定价能力需要重估。","en":"The company's own FQ4 guidance is roughly 86% gross margin. The threshold sits at 80% rather than 86% to leave about six points of buffer, so ordinary variance against guidance is not mistaken for a break. An actual print below 80% would mean ASP or mix is converting materially below the company's own basis, and the pricing power of the constraint owners would need to be reassessed."}},{"id":"FAL-HWX-003","metric":"Number of Demingli and Longsys simultaneously improving on Q3 gross margin, operating cash flow and inventory-to-total-assets","operator":"<","threshold":2,"unit":"count","horizon":"the Q3 2026 reports filed in October 2026","claimIds":["CL-HWX-010","CL-HWX-011","CL-HWX-022"],"rationale":{"zh-CN":"这两家公司的研究框架要求四项信号同时出现，其中三项可以在三季报里直接读出：毛利率不再下滑、经营现金流净流出明显收窄、存货占资产与周转不再恶化。两家都做不到，说明利润仍然由库存价差而不是经营质量驱动，7—8 倍的静态倍数只是周期分母过高。","en":"The research framework for these two requires four signals to appear together, and three of them read directly off the Q3 report: gross margin stops falling, negative operating cash flow narrows materially, and inventory-to-assets and turnover stop deteriorating. If neither company clears all three, profit is still driven by the inventory spread rather than operating quality, and the 7–8x static multiple is an inflated cyclical denominator."}},{"id":"FAL-HWX-004","metric":"Innolight and Eoptolink full-year 2026 operating cash flow ÷ attributable net profit","operator":"<","threshold":0.4,"unit":"multiple","horizon":"the FY2026 annual reports","claimIds":["CL-HWX-012","CL-HWX-013","CL-HWX-023"],"rationale":{"zh-CN":"上半年读数是 0.13 与 0.21，同业的天孚通信为 0.83，两端给出了区间。全年仍低于 0.4，说明利润没有转成现金，而不是半年度的时点扰动；在 30—100 倍以上的 run-rate 要求下，这足以推翻“盈利兑现最靠前”的定位。阈值取两端之间而非贴近上半年读数，避免把季节性回款当成结构性改善。","en":"The first-half readings were 0.13 and 0.21 against 0.83 at peer T&S Communications, which brackets the range. Staying below 0.4 for the full year would show profit is not converting to cash rather than reflecting half-year timing, and against run-rate multiples of 30x to 100x and above that is enough to overturn the ranking as furthest along on earnings conversion. The threshold sits between the two ends rather than close to the half-year reading, so that seasonal collections are not mistaken for structural improvement."}},{"id":"FAL-HWX-005","metric":"Number of Lumentum, Coherent and AAOI reporting next-quarter revenue at or above their own guidance midpoint","operator":"<","threshold":2,"unit":"count","horizon":"the fiscal quarters reported from October to November 2026","claimIds":["CL-HWX-018"],"rationale":{"zh-CN":"三家给出的指引中值分别对应约 24%、12% 和 42% 的环比增长，这是需求斜率仍在的全部公开依据。三家里少于两家达到自己的中值，说明斜率证据退回单公司层面，不能再支撑一个横截面判断。用公司自己的中值而不是外部一致预期作阈值，是为了让检验只依赖一手披露。","en":"The three guidance midpoints imply roughly 24%, 12% and 42% sequential growth, and that is the entire public basis for saying the demand slope is intact. Fewer than two of three clearing their own midpoint would push the evidence back to single-company stories and out of cross-sectional reach. The test uses each company's own midpoint rather than external consensus so that it depends only on first-party disclosure."}},{"id":"FAL-HWX-006","metric":"Number of listed humanoid robotics component suppliers in the reviewed sample disclosing separate humanoid revenue and gross margin","operator":"<","threshold":1,"unit":"count","horizon":"the Q3 2026 reports filed in October 2026","claimIds":["CL-HWX-024"],"rationale":{"zh-CN":"单列收入与毛利率是把人形机器人从叙事升级为利润池的最低门槛，低于这个门槛就没有可比的财务对象。阈值取 1 而不是更高，是因为当前样本中的家数为零，任何一家开始单列都足以让判断进入修订。","en":"Separately disclosed revenue and gross margin are the minimum threshold for moving humanoid robotics from narrative to profit pool; below it there is no comparable financial object at all. The threshold is one rather than something higher because the current count in the sample is zero, so a single company beginning to break the line out is enough to put the judgment into revision."}}],"versions":[{"version":"1.0.0","date":"2026-09-02","diff":{"zh-CN":"首版伴生研究对象，与人工审阅的中英文 canonical 展示稿一同发布。相对原稿的编辑处理：TrendForce 现货改记为报价序列并加注成交与需求确认偏弱；NVIDIA 2,790 亿美元统一表述为未来供应与产能承诺，不再暗示已付款采购或绝对需求锁定；中际旭创财报后的市场反应改为下跌 7.72%；AAOI 同比修正为约 86.4%；美国公司的年化倍数分母明确为非 GAAP 净利润，与非 GAAP 营业利润率分列；市值、市盈率与 52 周回撤统一标注 2026-09-01 快照且不进入事实层；第三方一致预期与 CXMT 相关报道移出 claim 层，分别记入证据台账与数据缺口；十倍路径改写为标注日期与 20×/25× 假设的作者情景反向检验；赔率表整体标注为研究判断。","en":"First companion research object, published alongside the human-reviewed Chinese and English canonical pages. Editorial handling relative to the source drafts: TrendForce spot is recorded as a quoted-price series with an explicit note that volume and demand confirmation stayed weak; NVIDIA's $279bn is stated consistently as forward supply and capacity commitments rather than paid procurement or a locked demand schedule; the Innolight post-print market reaction is recorded as a 7.72% decline; AAOI's year-over-year growth is corrected to about 86.4%; the denominator of the U.S. annualised multiples is stated explicitly as non-GAAP net income and kept separate from non-GAAP operating margin; market capitalisations, screen P/Es and 52-week drawdowns are labelled as a 2026-09-01 snapshot and excluded from the fact layer; the third-party consensus snapshot and the CXMT report are moved out of the claim layer into the evidence ledger and data gaps respectively; the ten-times path is rewritten as a dated author scenario back-test with explicit 20x and 25x assumptions; and the odds table is labelled as research judgment throughout."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"主线在换手？全球财报季后，存储与光模块的赔率重排","question":"同一轮 AI 资本开支里，利润正在留在哪一层，什么条件下会转移？","standfirst":"现货二阶导向上，合约价二阶导向下。存储的结构性利润先进入约束所有者，NAND 放大周期，中国模组放大库存时点；光模块的报表仍强，容错率已经变低。“存储接棒、光模块交棒”过于整齐，全球财报没有给出这样的画面。","whyItMatters":"把六个环节放进同一个截面，是因为它们共用一轮资本开支，却不共用同一段利润。HBM、服务器 DRAM 与企业级 NAND 的产能、工艺和客户认证集中在少数几家原厂，供给紧张最先进入它们的 ASP、毛利与现金流；NAND 的价格弹性更大，利润在 ASP 反转时双向工作；中国模组把低成本库存与成品售价之间的价差完整写进利润表，也把备货压力写进现金流量表。光模块是另一条线：收入与利润仍在兑现，现金转化却出现 0.13 到 0.83 的巨大分散。分层之后，同一个“AI 硬件”标签下会出现完全不同的证伪条件，这正是横截面研究要解决的问题。","consensus":"存储涨价、AI 资本开支强劲、光模块高增长，这些已经被普遍接受；SK 海力士与美光较 52 周高点的回撤、中际旭创财报后的下跌，也已经反映在价格里。","differentiated":"增量来自三次拆分。第一，把“存储涨价”拆成现货报价与月度固定价两条二阶导：现货加速而合约涨幅从 12.5% 收窄到 3.3%，同比利润仍高，环比加速度却会先见顶。第二，把“存储链”拆成约束所有者、NAND 周期弹性与库存价差三层，它们可以同时上涨，证伪条件完全不同。第三，把“光模块板块”拆成盈利兑现与现金转化两个维度：0.13、0.21 与 0.83 的现金转化差异，配上 30—100 倍以上的 run-rate 要求，比板块标签更能解释后续的估值波动。需求侧的口径同样要拆：NVIDIA 的 2,790 亿美元是未来供应与产能承诺，不是已付款采购。","valuation":"估值只做跨市场、跨币种、跨会计口径的周期筛查，不合成统一的美元收入或企业价值情景。四组情景分别覆盖约束所有者、NAND 弹性、中国模组与光模块，全部以 2026-09-01 收盘快照为时点。文中的峰值盈利年化倍数是市值除以最近季度或半年利润的简单年化，不是 forward PE，也不构成价格结论；美国公司的年化倍数分母是非 GAAP 净利润，与并列的非 GAAP 营业利润率不可互换。模组的十倍情景是作者设定的反向检验，用来回答“需要多大的盈利结构变化”，不是预测。","update":"v1.0 与人工审阅的中英文 canonical 展示稿同时发布，本对象只承担索引、证据台账、财务桥、估值情景、量化证伪与版本记录，不作为页面正文来源。相对原稿的主要修正见版本说明：报价与成交的区分、NVIDIA 承诺口径、中际旭创市场反应、AAOI 同比、非 GAAP 分母、快照标注，以及一致预期与 CXMT 报道的降级处理。","evidenceClaimIds":["CL-HWX-001","CL-HWX-002","CL-HWX-003","CL-HWX-004","CL-HWX-005","CL-HWX-006","CL-HWX-007","CL-HWX-008","CL-HWX-009","CL-HWX-010","CL-HWX-011","CL-HWX-012","CL-HWX-013","CL-HWX-014","CL-HWX-015","CL-HWX-016","CL-HWX-017","CL-HWX-018","CL-HWX-019","CL-HWX-020"],"causalChain":[{"title":"两条价格线的加速度分叉","body":"现货报价重新加速，月度固定价的环比涨幅从 12.5% 收窄到 3.3%，同一轮涨价里出现两种斜率。","claimIds":["CL-HWX-001","CL-HWX-002","CL-HWX-003"]},{"title":"结构性利润先落在约束所有者","body":"产能、工艺与客户认证集中在上游，供给紧张最先进入 SK 海力士与美光的 ASP、毛利与现金流。","claimIds":["CL-HWX-005","CL-HWX-006","CL-HWX-021"]},{"title":"弹性层放大周期与库存时点","body":"Kioxia、SanDisk 承接 NAND 价格弹性；德明利、江波龙的利润来自库存价差，现金流与存货决定它能留多久。","claimIds":["CL-HWX-008","CL-HWX-009","CL-HWX-010","CL-HWX-011","CL-HWX-022"]},{"title":"光模块的门槛从增长换成现金","body":"六家公司同时给出强劲收入或指引，分化落在现金转化、资本开支与产能兑现上。","claimIds":["CL-HWX-012","CL-HWX-016","CL-HWX-018","CL-HWX-023"]},{"title":"证据边界决定利润池排序","body":"人形机器人尚无单列收入可比，供给正常化与未确认的第三方报道构成主要反向情景。","claimIds":["CL-HWX-024","CL-HWX-025"]}],"riskClaimIds":["CL-HWX-022","CL-HWX-023","CL-HWX-024","CL-HWX-025"],"forwardTestIds":["FAL-HWX-001","FAL-HWX-002","FAL-HWX-003","FAL-HWX-004","FAL-HWX-005","FAL-HWX-006"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Is the baton really changing hands? Repricing memory and optics after a global reporting season","question":"Within one AI capex cycle, which layer is keeping the profit, and what would move it?","standfirst":"The second derivative of spot is rising while the second derivative of contract prices is falling. Structural memory profit lands first with the constraint owners, NAND amplifies the cycle, and the Chinese module makers amplify inventory timing. In optics the reported numbers are still strong and the margin for error has thinned. The tidy version — memory takes the baton, optics hands it over — is not what the global reporting season produced.","whyItMatters":"These six nodes belong in one cross-section because they share a capex cycle without sharing the same dollar of profit. Capacity, process and customer qualification for HBM, server DRAM and enterprise NAND sit with a handful of manufacturers, so tight supply reaches their ASPs, gross margin and cash flow first. NAND carries more price elasticity, and that elasticity works in both directions on any ASP inflection. The Chinese module makers write the spread between low-cost inventory and realised prices straight into the income statement, and the cost of building that inventory into the cash-flow statement. Optics is a separate line entirely: revenue and profit keep converting while cash conversion disperses from 0.13 to 0.83. Once the layers are separated, one 'AI hardware' label turns out to contain completely different falsification conditions, which is the problem a cross-section exists to solve.","consensus":"Rising memory prices, strong AI capex and high optics growth are all widely accepted, and the drawdowns at SK hynix and Micron from their 52-week highs, along with Innolight's post-print decline, are already in the price.","differentiated":"The information gain comes from three separations. First, split 'memory prices are up' into spot quotes and monthly fixed prices: spot is re-accelerating while contract increases narrowed from 12.5% to 3.3%, so year-over-year profit can stay high while sequential acceleration peaks first. Second, split the memory chain into constraint owners, NAND cycle elasticity and inventory spread; all three can rise together and none shares a falsification condition with the others. Third, split the optics sector into earnings conversion and cash conversion: the gap between 0.13, 0.21 and 0.83, set against run-rate multiples of 30x to 100x and above, explains subsequent valuation volatility far better than the sector label. The demand side needs the same discipline: NVIDIA's $279 billion is a forward supply and capacity commitment, not paid procurement.","valuation":"Valuation here is a cycle screen across markets, currencies and accounting bases, with no synthetic common-currency revenue or enterprise-value scenario. Four scenarios cover the constraint owners, NAND elasticity, the Chinese module makers and optics, all struck at the 2026-09-01 close. The peak-earnings annualised multiples divide market capitalisation by a simple annualisation of the latest quarterly or half-year profit; they are not forward P/E and not price conclusions. For the U.S. names the denominator is non-GAAP net income and cannot be exchanged with the non-GAAP operating margin quoted alongside it. The ten-times module scenario is an author-defined back-test answering how much the earnings structure would have to change, not a forecast.","update":"Version 1.0 ships alongside the human-reviewed Chinese and English canonical pages. This object carries the index entry, evidence ledger, financial bridge, valuation scenarios, quantitative falsifiers and version history; it is not the source of the page body. The principal corrections against the source drafts are listed in the version note: separating quotes from transactions, restating the NVIDIA commitment basis, recording the Innolight market reaction, correcting AAOI's year-over-year rate, naming the non-GAAP denominator, labelling the market snapshot, and demoting the consensus snapshot and the CXMT report out of the claim layer.","evidenceClaimIds":["CL-HWX-001","CL-HWX-002","CL-HWX-003","CL-HWX-004","CL-HWX-005","CL-HWX-006","CL-HWX-007","CL-HWX-008","CL-HWX-009","CL-HWX-010","CL-HWX-011","CL-HWX-012","CL-HWX-013","CL-HWX-014","CL-HWX-015","CL-HWX-016","CL-HWX-017","CL-HWX-018","CL-HWX-019","CL-HWX-020"],"causalChain":[{"title":"Two price lines, opposite acceleration","body":"Spot quotes re-accelerate while month-on-month fixed-price increases narrow from 12.5% to 3.3%, producing two slopes inside one upcycle.","claimIds":["CL-HWX-001","CL-HWX-002","CL-HWX-003"]},{"title":"Structural profit lands with the constraint owners","body":"Capacity, process and qualification sit upstream, so tight supply reaches SK hynix and Micron's ASPs, margin and cash flow first.","claimIds":["CL-HWX-005","CL-HWX-006","CL-HWX-021"]},{"title":"The elastic layer amplifies cycle and timing","body":"Kioxia and SanDisk carry NAND price elasticity; Demingli and Longsys earn on the inventory spread, and cash flow and inventory decide how long it stays.","claimIds":["CL-HWX-008","CL-HWX-009","CL-HWX-010","CL-HWX-011","CL-HWX-022"]},{"title":"In optics the gate moves from growth to cash","body":"All six names posted strong revenue or guidance; the dispersion sits in cash conversion, capex and capacity execution.","claimIds":["CL-HWX-012","CL-HWX-016","CL-HWX-018","CL-HWX-023"]},{"title":"Evidence boundaries rank the profit pools","body":"Humanoid robotics has no separately disclosed revenue to compare, while supply normalisation and an unconfirmed third-party report frame the bear case.","claimIds":["CL-HWX-024","CL-HWX-025"]}],"riskClaimIds":["CL-HWX-022","CL-HWX-023","CL-HWX-024","CL-HWX-025"],"forwardTestIds":["FAL-HWX-001","FAL-HWX-002","FAL-HWX-003","FAL-HWX-004","FAL-HWX-005","FAL-HWX-006"]}},"dataGaps":[{"zh-CN":"跨市场口径不统一：会计准则、财季与货币各不相同，本对象因此不构造统一美元收入或企业价值情景，估值只以各自口径的峰值盈利快筛呈现。非日历财季公司（美光、SanDisk、Lumentum、Kioxia）的 dataAsOf 按各自披露的财季惯例记录，A 股半年报的披露日未逐一核对。","en":"Cross-market bases do not align: accounting standards, fiscal calendars and currencies all differ, so this object builds no common-currency revenue or enterprise-value scenario and presents valuation only as peak-earnings screens on each company's own basis. For the non-calendar reporters (Micron, SanDisk, Lumentum, Kioxia) dataAsOf follows each company's disclosed fiscal calendar, and the exact filing dates of the A-share interim reports were not individually reconciled."},{"zh-CN":"行情与市值来自 Google Finance、MarketWatch 与 2026-09-01 交易所收盘数据，未与完全稀释股数逐一核对。因此市值、行情端市盈率与 52 周回撤都是近似快照，不进入事实层，也不用于正式估值。","en":"Prices and market capitalisations come from Google Finance, MarketWatch and exchange closing data for 2026-09-01, and were not reconciled to fully diluted share counts. Market capitalisations, screen P/Es and 52-week drawdowns are therefore approximate snapshots, excluded from the fact layer and not used for formal valuation."},{"zh-CN":"峰值盈利年化倍数的分母含峰值 ASP、低成本库存与较低减值计提；SK 海力士的 TTM 分母包含约 63.3 万亿韩元投资资产收益；三星是多业务集团。这些倍数都不是 forward PE，也不是价格结论。","en":"The denominators of the peak-earnings annualised multiples embed peak ASPs, low-cost inventory and light impairment. SK hynix's trailing denominator includes roughly KRW 63.3tn of gains on investment assets, and Samsung remains a multi-business group. None of these multiples is a forward P/E or a price conclusion."},{"zh-CN":"第三方一致预期：美光的前瞻市盈率与 FY2027 一致预期每股收益取自 Seeking Alpha，抓取日期 2026-09-01，具体数字仅在 canonical 展示稿中呈现。它是数据商汇总的卖方一致预期快照，不是公司指引，会随每次财报与修正移动，不同数据商在财年口径与 GAAP/非 GAAP 处理上的差异也会改变结果，因此不进入本对象的 claim 层。","en":"Third-party consensus: the forward earnings screen and FY2027 consensus EPS for Micron come from Seeking Alpha, captured on 2026-09-01, and the specific figures appear only in the canonical editorial page. It is a vendor-aggregated sell-side consensus snapshot rather than company guidance, moves with every print and revision, and changes across vendors with fiscal-year alignment and GAAP versus non-GAAP treatment, so it stays out of this object's claim layer."},{"zh-CN":"CXMT 小批量生产 HBM3E 的说法来自媒体援引知情人士的报道，未获公司确认。本对象不据此推导价格、份额或供需影响，只作为 2027 年供给侧的待验证风险记录，等待公司或监管层面的一手披露。","en":"The report that CXMT has begun small-volume production of HBM3E comes from press reporting citing people familiar with the matter and is not company-confirmed. No pricing, share or supply-demand effect is inferred from it here; it is recorded only as an unverified 2027 supply-side risk pending first-party company or regulatory disclosure."},{"zh-CN":"人形机器人：原稿提到的中国零部件与整机样本公司，本对象未逐一绑定一手披露链接，因此不进入 claim 层，只保留“样本内无公司单列人形机器人收入与毛利率”这一结构性判断，并作为待补证据处理。","en":"Humanoid robotics: the Chinese component and whole-machine sample companies referenced in the source drafts are not individually bound to first-party disclosure links here, so they stay out of the claim layer. Only the structural finding — that no company in the reviewed sample discloses humanoid revenue and gross margin separately — is retained, with the underlying datapoints treated as evidence still to be sourced."},{"zh-CN":"口径分离尚未闭合：中国公司使用归母净利润，美国公司的年化倍数使用非 GAAP 净利润，而非 GAAP 营业利润率是第三条口径。要把六家光模块公司放进严格可比的估值框架，需要统一正常化利润、净现金、税率与股份数，本对象没有做这一步。","en":"The basis separation is not closed: the Chinese names use attributable net profit, the U.S. annualised multiples use non-GAAP net income, and non-GAAP operating margin is a third basis. Placing all six optics names in a strictly comparable valuation framework would require normalising profit, net cash, tax rate and share count, and this object does not take that step."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/global-ai-hardware-profit-pools-2026-09","en":"https://lunartuliplab.com/en/deep-dive/global-ai-hardware-profit-pools-2026-09"},"machineReadable":"https://lunartuliplab.com/research/global-ai-hardware-profit-pools-2026-09","dateModified":"2026-09-02"},{"id":"RO-THEME-MEMORY-CASH-CAPTURE-001","slug":"hbm-memory-cash-capture-alpha-2026-09","kind":"theme-study","tickers":["MU","000660.KS","005930.KS","SNDK","285A.T"],"publishedAt":"2026-09-10","asOf":"2026-09-10","version":"1.0.0","sourceLabel":"Memory capital-cycle series I · HBM revenue-to-cash-capture gap · 2026-09-10 edition","claims":[{"id":"CL-MEMTH-001","type":"Fact","text":{"zh-CN":"SanDisk FY2026 经营现金流为 116.71 亿美元，购置固定资产为 1.77 亿美元，公司列示自由现金流为 114.94 亿美元；同期收入为 202.48 亿美元。","en":"SanDisk reported FY2026 operating cash flow of $11.671 billion, purchases of property, plant and equipment of $177 million, company-presented free cash flow of $11.494 billion, and revenue of $20.248 billion."},"evidenceIds":["EV-MEMTH-SNDK"]},{"id":"CL-MEMTH-002","type":"Derived","text":{"zh-CN":"报告自由现金流率为 56.77%；再扣除 2.75 亿美元合资企业净出资及 24.76 亿美元新商业模式相关现金影响后，调整后自由现金流为 87.43 亿美元，对应收入的 43.18%，两种口径相差 13.59 个百分点。","en":"Reported free-cash-flow margin was 56.77%. After a further $275 million of net joint-venture contributions and $2.476 billion of cash effects related to the new business model, adjusted free cash flow was $8.743 billion, or 43.18% of revenue, a gap of 13.59 percentage points."},"evidenceIds":["EV-MEMTH-SNDK"]},{"id":"CL-MEMTH-003","type":"Fact","text":{"zh-CN":"美光披露 HBM4 12-high 量产爬坡速度约为 HBM3E 12-high 的两倍，累计 HBM4 出货收入超过 10 亿美元；SK 海力士披露 HBM4 已开始批量出货，三星披露 HBM4 销售扩大并送出 HBM4E 样品。","en":"Micron disclosed that its HBM4 12-high production ramp was running at roughly twice the pace of HBM3E 12-high and that cumulative HBM4 shipment revenue had exceeded $1 billion. SK hynix reported volume HBM4 shipments, while Samsung reported expanding HBM4 sales and HBM4E samples delivered to major customers."},"evidenceIds":["EV-MEMTH-MU","EV-MEMTH-SKH","EV-MEMTH-SAMSUNG"]},{"id":"CL-MEMTH-004","type":"Hypothesis","text":{"zh-CN":"下一段可重复验证的 HBM Alpha 更可能来自“产品收入—调整后现金捕获差”：产品收入增长、现金转换稳定且合格供给到达偏慢，支持更高周期底部；收入与利润增长但调整后现金转换走弱，则提示盈利质量先于标题利润见顶。","en":"The next repeatably testable HBM alpha may sit in the product-revenue-to-adjusted-cash-capture gap. Rising product revenue, stable cash conversion and slow qualified supply would support a higher cycle floor; rising revenue and profit alongside weakening adjusted cash conversion would signal earnings quality peaking ahead of headline profit."},"evidenceIds":["EV-MEMTH-SNDK","EV-MEMTH-MU","EV-MEMTH-SKH","EV-MEMTH-SAMSUNG"]}],"evidence":[{"id":"EV-MEMTH-SNDK","title":"SanDisk FY2026 Results and Form 10-K","source":{"publisher":"SanDisk Investor Relations","url":"https://investor.sandisk.com/node/8136/pdf","document":"FY2026 Results"},"sourceDate":"2026-08-05","dataAsOf":"2026-07-03","calculation":"11494 / 20248 = 56.77%; (11494 - 275 - 2476) / 20248 = 43.18%.","confidence":"high","counterevidence":"SanDisk is a NAND-cycle control, not an HBM producer; the bridge demonstrates method rather than HBM product economics.","lastVerified":"2026-09-10","role":"primary"},{"id":"EV-MEMTH-MU","title":"Micron FY2026 Q3 prepared remarks and Form 10-Q","source":{"publisher":"Micron Investor Relations","url":"https://s25.q4cdn.com/621799436/files/doc_financials/2026/q3/Q3-FY26-Prepared-Remarks.pdf","document":"FY2026 Q3 prepared remarks"},"sourceDate":"2026-06-24","dataAsOf":"2026-05-28","calculation":null,"confidence":"high","counterevidence":"Micron does not disclose a complete standalone HBM profit and cash-flow bridge.","lastVerified":"2026-09-10","role":"primary"},{"id":"EV-MEMTH-SKH","title":"SK hynix Q2 2026 business results","source":{"publisher":"SK hynix Newsroom","url":"https://news.skhynix.com/en/q2-2026-business-results/","document":"Q2 2026 business results"},"sourceDate":"2026-07-30","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"Company-level profit includes memory products beyond HBM.","lastVerified":"2026-09-10","role":"primary"},{"id":"EV-MEMTH-SAMSUNG","title":"Samsung Electronics Q2 2026 results","source":{"publisher":"Samsung Newsroom","url":"https://news.samsung.com/global/samsung-electronics-announces-second-quarter-2026-results","document":"Q2 2026 results"},"sourceDate":"2026-07-30","dataAsOf":"2026-06-30","calculation":null,"confidence":"high","counterevidence":"Device Solutions combines memory, foundry and System LSI; HBM profit is not separately disclosed.","lastVerified":"2026-09-10","role":"primary"}],"financialBridge":[{"label":{"zh-CN":"SanDisk FY2026 收入","en":"SanDisk FY2026 revenue"},"period":"2026-FY","value":20248,"unit":"USDm","evidenceIds":["EV-MEMTH-SNDK"]},{"label":{"zh-CN":"报告自由现金流","en":"Reported free cash flow"},"period":"2026-FY","value":11494,"unit":"USDm","evidenceIds":["EV-MEMTH-SNDK"]},{"label":{"zh-CN":"调整后自由现金流","en":"Adjusted free cash flow"},"period":"2026-FY","value":8743,"unit":"USDm","evidenceIds":["EV-MEMTH-SNDK"]},{"label":{"zh-CN":"报告与调整后 FCF 率差","en":"Reported-to-adjusted FCF-margin gap"},"period":"2026-FY","value":13.59,"unit":"percent","evidenceIds":["EV-MEMTH-SNDK"]}],"valuationScenarios":[{"id":"VAL-MEMTH-CASH-CAPTURE","label":{"zh-CN":"现金捕获差压力测试","en":"Cash-capture-gap stress test"},"metrics":[{"label":{"zh-CN":"报告 FCF 率","en":"Reported FCF margin"},"value":{"zh-CN":"56.77%","en":"56.77%"}},{"label":{"zh-CN":"调整后 FCF 率","en":"Adjusted FCF margin"},"value":{"zh-CN":"43.18%","en":"43.18%"}}],"calculation":"(11671 - 177) / 20248; (11671 - 177 - 275 - 2476) / 20248.","interpretation":{"zh-CN":"两种口径都可复算，但回答不同问题；差距的连续变化比任一单期绝对值更适合作为因子。","en":"Both bases are reproducible but answer different questions. The sequential change in the gap is a better factor candidate than either one-period level."}},{"id":"VAL-MEMTH-DURATION","label":{"zh-CN":"竞争优势期分叉","en":"Competitive-duration fork"},"metrics":[{"label":{"zh-CN":"正向条件","en":"Positive condition"},"value":{"zh-CN":"现金差收窄","en":"Cash gap narrows"}},{"label":{"zh-CN":"反向条件","en":"Negative condition"},"value":{"zh-CN":"有效供给加速","en":"Qualified supply accelerates"}}],"calculation":"Scenario classification; no per-share valuation is produced.","interpretation":{"zh-CN":"产品收入与调整后现金转换同向、而合格供给偏慢时，竞争优势期可延长；反之应缩短。","en":"Competitive duration can extend when product revenue and adjusted cash conversion rise together while qualified supply stays slow; the reverse combination shortens it."}}],"counterevidenceClaimIds":["CL-MEMTH-004"],"falsifiers":[{"id":"FAL-MEMTH-001","metric":"Adjusted cash-conversion direction while HBM product revenue grows","operator":"<","threshold":0,"unit":"percent","horizon":"each of the next two reporting cycles","claimIds":["CL-MEMTH-004"],"rationale":{"zh-CN":"若 HBM 产品收入增长时调整后现金转换连续环比下降，收入—现金捕获差扩大，结构性重估假设应降级。","en":"If adjusted cash conversion falls sequentially while HBM product revenue rises, the gap is widening and the structural rerating hypothesis should be downgraded."}},{"id":"FAL-MEMTH-002","metric":"Supplier programs advancing from plan to qualified sellable output","operator":">=","threshold":2,"unit":"count","horizon":"each of the next two reporting cycles","claimIds":["CL-MEMTH-004"],"rationale":{"zh-CN":"若同一观察窗内至少两个供应项目跨入认证后的可售产出，供给斜率快于当前假设，竞争优势期应缩短。","en":"If at least two supplier programmes reach post-qualification sellable output in the same observation window, supply is arriving faster than assumed and competitive duration should shorten."}}],"versions":[{"version":"1.0.0","date":"2026-09-10","diff":{"zh-CN":"首版双语 canonical 研究对象；保留现金桥图与四层因果图，并与 MU、SNDK 资本周期姊妹篇组成三篇系列。","en":"First bilingual canonical object, retaining the cash bridge and four-layer causal diagram and forming a three-part series with the MU and SNDK capital-cycle studies."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"HBM 传统周期估值是否会被打破：下一段 Alpha 可能藏在“收入—现金捕获差”里","question":"HBM 的合同化究竟抬高了周期底部，还是高利润正在通过资本开支制造下一轮供给？","standfirst":"需求、价格、良率与产能已经进入共识区。下一段更可验证的预期差，是产品收入增长能否在剔除客户资金、合资出资与新增产能投资后留下可重复现金。","whyItMatters":"收入、利润与现金捕获并不天然同步。把有效供给、产品商业化、财务归属与现金桥串成同一时点序列，才能区分结构性利润底盘与周期尖峰。","consensus":"AI 存储需求、HBM4 放量和长期协议提高收入可见度已经被市场广泛接受。","differentiated":"研究重点从需求总量转向收入—现金捕获差，以及高利润诱发资本开支、制程迁移与新进入者的二阶供给反馈。","valuation":"估值不依赖单点收入增速或毛利率，而拆成归属清晰的产品收入、调整后现金转换与竞争优势期。SanDisk 现金桥是方法对照，不是 HBM 产品利润替代值。","update":"v1.0 保留中文原稿的两个图表和全部研究判断，新增 claim-parity 英文版、伴生证据台账与系列导航。","evidenceClaimIds":["CL-MEMTH-001","CL-MEMTH-002","CL-MEMTH-003"],"causalChain":[{"title":"有效供给","body":"计划先经过设备、投产、认证和良率，才成为可售产出。","claimIds":["CL-MEMTH-003"]},{"title":"财务归属","body":"产品收入必须与分部及公司现金流桥接，不能把公司利润直接等同 HBM 单品利润。","claimIds":["CL-MEMTH-001","CL-MEMTH-004"]},{"title":"连续验证","body":"收入增长与调整后现金转换同向才提高盈利上修质量。","claimIds":["CL-MEMTH-002","CL-MEMTH-004"]}],"riskClaimIds":["CL-MEMTH-002"],"forwardTestIds":["FAL-MEMTH-001","FAL-MEMTH-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Can HBM break the traditional cycle valuation? The next alpha may sit in the revenue-to-cash capture gap","question":"Has HBM contracting raised the cycle floor, or are peak profits financing the next supply wave?","standfirst":"Demand, pricing, yield and capacity are already crowded debates. The more testable expectations gap is whether product-revenue growth still leaves repeatable cash after customer funding, joint-venture contributions and new capacity investment.","whyItMatters":"Revenue, profit and cash capture do not move together by construction. Effective supply, product commercialisation, financial attribution and the cash bridge have to sit in one point-in-time series to separate a structural earnings floor from a cyclical peak.","consensus":"AI-memory demand, the HBM4 ramp and long-term agreements improving revenue visibility are widely accepted.","differentiated":"The work moves from aggregate demand to the revenue-to-cash capture gap and to the second-order supply response induced by high profit: capex, process migration and new entrants.","valuation":"Valuation is decomposed into attributable product revenue, adjusted cash conversion and competitive-advantage period rather than a single revenue-growth or gross-margin point. The SanDisk bridge is a methodological control, not a proxy for HBM product profit.","update":"Version 1.0 preserves both figures and every research judgment in the Chinese source, adding a claim-parity English edition, companion evidence ledger and series navigation.","evidenceClaimIds":["CL-MEMTH-001","CL-MEMTH-002","CL-MEMTH-003"],"causalChain":[{"title":"Effective supply","body":"Plans must pass through equipment, production, qualification and yield before becoming sellable output.","claimIds":["CL-MEMTH-003"]},{"title":"Financial attribution","body":"Product revenue must bridge into segment and company cash flow; company profit cannot stand in for standalone HBM economics.","claimIds":["CL-MEMTH-001","CL-MEMTH-004"]},{"title":"Continuous validation","body":"Earnings revisions have higher quality only when revenue growth and adjusted cash conversion move together.","claimIds":["CL-MEMTH-002","CL-MEMTH-004"]}],"riskClaimIds":["CL-MEMTH-002"],"forwardTestIds":["FAL-MEMTH-001","FAL-MEMTH-002"]}},"dataGaps":[{"zh-CN":"主要 HBM 供应商尚未披露可与公司现金流完整桥接的 HBM 单品利润、资本开支与现金转换；因此本对象提出因子假设，不声称已经完成收益预测验证。","en":"Major HBM suppliers do not yet disclose standalone HBM profit, capex and cash conversion that reconcile fully to company cash flow. The object therefore states a factor hypothesis rather than claiming completed return-prediction validation."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/hbm-memory-cash-capture-alpha-2026-09","en":"https://lunartuliplab.com/en/deep-dive/hbm-memory-cash-capture-alpha-2026-09"},"machineReadable":"https://lunartuliplab.com/research/hbm-memory-cash-capture-alpha-2026-09","dateModified":"2026-09-10"},{"id":"RO-COMP-MU-CAPITAL-CYCLE-001","slug":"micron-hbm-capital-cycle-2026-09","kind":"company-deep-dive","tickers":["MU"],"publishedAt":"2026-09-10","asOf":"2026-09-07","version":"1.0.0","sourceLabel":"Memory capital-cycle series II · Micron HBM capital cycle · 2026-09-07 information set","claims":[{"id":"CL-MUCC-001","type":"Fact","text":{"zh-CN":"美光 FQ3-26 营收 414.6 亿美元，同比增长约 346%；非 GAAP 毛利率 84.9%，FQ4 指引为营收 500 亿美元±10 亿美元、非 GAAP 毛利率约 86%。","en":"Micron reported FQ3-26 revenue of $41.46 billion, up roughly 346% year over year, and a non-GAAP gross margin of 84.9%; FQ4 guidance called for revenue of $50.0 billion ± $1.0 billion and a non-GAAP gross margin of roughly 86%."},"evidenceIds":["EV-MUCC-Q3"]},{"id":"CL-MUCC-002","type":"Fact","text":{"zh-CN":"美光披露 FY27 资本开支每季将高于 FQ4 约 100 亿美元的水平，过半增量用于洁净室建设。","en":"Micron disclosed that FY27 capex would run above the roughly $10 billion FQ4 level in every quarter, with more than half of the increment allocated to cleanroom construction."},"evidenceIds":["EV-MUCC-Q3"]},{"id":"CL-MUCC-003","type":"Derived","text":{"zh-CN":"在 WACC 9.5%、永续增速 3%、十年窗口及 EV 近似市值的简化反向 DCF 下，2026-09-04 快照隐含未来五年营收复合 21.57% 与稳态 FCF 利润率 28.99%；结果对取价与模型假设敏感。","en":"Under a simplified reverse DCF using a 9.5% WACC, 3% terminal growth, a ten-year window and enterprise value approximated by market capitalisation, the 2026-09-04 snapshot implied 21.57% five-year revenue CAGR and a 28.99% steady FCF margin. The output is sensitive to the price date and model assumptions."},"evidenceIds":["EV-MUCC-MODEL"]},{"id":"CL-MUCC-004","type":"Hypothesis","text":{"zh-CN":"HBM 合同化让存储“脱离周期”的命题，需要 2027 年 HBM4/4E 合约价、三家供应商 HBM 毛利率及 HBM 与商品 DRAM 联动性三项同时通过；研究稿对三项存活至 2028 的主观概率为约 25%。","en":"The thesis that HBM contracting lets memory escape the cycle requires three tests to pass together: 2027 HBM4/4E contract prices, HBM gross margins across all three suppliers, and decoupling between HBM and commodity DRAM. The research note assigns a subjective probability of roughly 25% that all three survive through 2028."},"evidenceIds":["EV-MUCC-Q3","EV-MUCC-MODEL"]}],"evidence":[{"id":"EV-MUCC-Q3","title":"Micron FY2026 Q3 prepared remarks","source":{"publisher":"Micron Investor Relations","url":"https://s25.q4cdn.com/621799436/files/doc_financials/2026/q3/Q3-FY26-Prepared-Remarks.pdf","document":"FY2026 Q3 prepared remarks"},"sourceDate":"2026-06-24","dataAsOf":"2026-05-28","calculation":null,"confidence":"high","counterevidence":"Guidance and capex plans are forward-looking and may be revised.","lastVerified":"2026-09-10","role":"primary"},{"id":"EV-MUCC-MODEL","title":"Micron point-in-time research uniform and reverse-DCF-lite output","source":{"publisher":"LunarTulip Research","document":"MU capital-cycle uniform · 2026-09-07"},"sourceDate":"2026-09-07","dataAsOf":"2026-09-04","calculation":"WACC 9.5%; terminal growth 3%; ten-year window; EV approximated by market capitalisation before net cash/debt.","confidence":"medium","counterevidence":"The inverse problem is non-unique and point-in-time market inputs came from a third-party quote service.","lastVerified":"2026-09-10","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"FQ3-26 营收","en":"FQ3-26 revenue"},"period":"2026-FQ3","value":41456,"unit":"USDm","evidenceIds":["EV-MUCC-Q3"]},{"label":{"zh-CN":"FQ3-26 非 GAAP 毛利率","en":"FQ3-26 non-GAAP gross margin"},"period":"2026-FQ3","value":84.9,"unit":"percent","evidenceIds":["EV-MUCC-Q3"]},{"label":{"zh-CN":"FQ4-26 营收指引中点","en":"FQ4-26 revenue guidance midpoint"},"period":"2026-FQ4","value":50000,"unit":"USDm","evidenceIds":["EV-MUCC-Q3"]}],"valuationScenarios":[{"id":"VAL-MUCC-REVERSE-DCF","label":{"zh-CN":"市场隐含高原","en":"Market-implied plateau"},"metrics":[{"label":{"zh-CN":"五年营收复合","en":"Five-year revenue CAGR"},"value":{"zh-CN":"21.57%","en":"21.57%"}},{"label":{"zh-CN":"稳态 FCF 利润率","en":"Steady FCF margin"},"value":{"zh-CN":"28.99%","en":"28.99%"}}],"calculation":"Reverse-DCF-lite v1 at the 2026-09-04 point-in-time quote; WACC 9.5%, terminal growth 3%, ten-year window.","interpretation":{"zh-CN":"这是用于读出市场要求的敏感性结果，不是内在价值或目标价。","en":"This is a sensitivity reading of what the market requires, not intrinsic value or a price target."}},{"id":"VAL-MUCC-PLATEAU-FORK","label":{"zh-CN":"高原化与均值回归","en":"Plateau versus mean reversion"},"metrics":[{"label":{"zh-CN":"研究稿高原化概率","en":"Research-note plateau probability"},"value":{"zh-CN":"约 45%","en":"about 45%"}},{"label":{"zh-CN":"脱离周期概率","en":"Escape-from-cycle probability"},"value":{"zh-CN":"约 25%","en":"about 25%"}}],"calculation":"Author-assigned scenario weights in the 2026-09-07 information set; not statistical forecasts.","interpretation":{"zh-CN":"高原化允许需求不崩但利润率从尖峰收敛；概率是研究判断，不是历史频率。","en":"The plateau case allows demand to hold while margin compresses from the peak. The probabilities are research judgments, not historical frequencies."}}],"counterevidenceClaimIds":["CL-MUCC-004"],"falsifiers":[{"id":"FAL-MUCC-001","metric":"2027 HBM4/4E contract-price YoY decline","operator":">","threshold":15,"unit":"percent","horizon":"calendar 2027","claimIds":["CL-MUCC-004"],"rationale":{"zh-CN":"同比跌幅超过 15% 说明定制品价格竞争已经足以破坏“脱离周期”的第一项条件。","en":"A year-over-year decline greater than 15% would mean custom-product price competition has broken the first condition of the escape-from-cycle thesis."}},{"id":"FAL-MUCC-002","metric":"HBM gross margin across all three suppliers","operator":"<","threshold":60,"unit":"percent","horizon":"full-year 2027","claimIds":["CL-MUCC-004"],"rationale":{"zh-CN":"任一供应商全年 HBM 毛利率低于 60%，三项同时成立的命题即失效。","en":"If any supplier's full-year HBM gross margin falls below 60%, the joint three-condition thesis fails."}}],"versions":[{"version":"1.0.0","date":"2026-09-10","diff":{"zh-CN":"将 2026-09-07 Desk 原稿按原文上架为独立双语 canonical 页面，并保留时点制服、反向 DCF 边界、验证日历与姊妹篇关系。","en":"Published the 2026-09-07 Desk source as an independent bilingual canonical page, retaining its point-in-time uniform, reverse-DCF boundaries, validation calendar and sibling relationship."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"报表在顶点，资本在涌入：美光的“脱离周期”命题进入可证伪窗口","question":"HBM 合同化能否抵消 2027—2028 年集中到达的供给响应？","standfirst":"短期涨价与分配制仍可能兑现，但决定周期是否被打破的变量不在下一份财报，而在资本开支转成合格供给后的价格、利润率与商品联动性。","whyItMatters":"记录利润正在诱发洁净室、制程迁移和新进入者投资。资本周期的物理时滞让短期紧缺与中期回报率回落可以同时成立。","consensus":"FQ3-26 的记录营收、毛利率与 FQ4 指引，以及 HBM 长期协议提高收入可见度，已经进入市场定价。","differentiated":"增量信息是四路供给响应及 2H27—2028 的集中到达窗；命题必须通过价格、毛利率和商品联动性三项测试。","valuation":"反向 DCF 只读出市场要求的约 20% 五年营收复合与约 30% 稳态 FCF 利润率，不把低前瞻市盈率直接等同低估。","update":"v1.0 基本原样保留 Desk 中文稿，新增 claim-parity 英文稿、公开官网 canonical、系列导航与伴生证据对象；不依赖或回链 Desk 私有 URL。","evidenceClaimIds":["CL-MUCC-001","CL-MUCC-002","CL-MUCC-003"],"causalChain":[{"title":"涨价脉冲","body":"收入增长主要由价格贡献，报表仍可能继续强。","claimIds":["CL-MUCC-001"]},{"title":"资本响应","body":"记录利润推动洁净室与行业产能投资。","claimIds":["CL-MUCC-002"]},{"title":"2027—2028 裁决","body":"价格、利润率和商品联动性共同决定命题是否存活。","claimIds":["CL-MUCC-004"]}],"riskClaimIds":["CL-MUCC-002"],"forwardTestIds":["FAL-MUCC-001","FAL-MUCC-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Financials at the peak, capital flooding in: Micron's escape-from-cycle thesis enters a falsifiable window","question":"Can HBM contracting offset the supply response arriving in 2027–2028?","standfirst":"Near-term price and allocation dynamics can still deliver, but the variable that decides whether the cycle is broken is not the next print. It is what happens to price, margin and commodity linkage once capex becomes qualified supply.","whyItMatters":"Record profit is inducing cleanroom construction, process migration and entrant investment. Capital-cycle lead times allow near-term scarcity and medium-term return compression to coexist.","consensus":"Record FQ3-26 revenue and margin, FQ4 guidance and long-term HBM agreements improving visibility are already in the price.","differentiated":"The information gain is the four-way supply response and its concentrated 2H27–2028 arrival window. The thesis must pass contract-price, gross-margin and commodity-linkage tests together.","valuation":"The reverse DCF reads market requirements of roughly 20% five-year revenue CAGR and roughly 30% steady FCF margin; it does not equate a low forward P/E with undervaluation.","update":"Version 1.0 retains the Desk Chinese source substantially as written, adds a claim-parity English edition, public-site canonical, series navigation and companion evidence object, and carries no dependency on or link back to the private Desk URL.","evidenceClaimIds":["CL-MUCC-001","CL-MUCC-002","CL-MUCC-003"],"causalChain":[{"title":"Price pulse","body":"Price contributed most of the revenue increase, and reported results can remain strong.","claimIds":["CL-MUCC-001"]},{"title":"Capital response","body":"Record profit drives cleanroom and industry-capacity investment.","claimIds":["CL-MUCC-002"]},{"title":"2027–2028 adjudication","body":"Price, margin and commodity linkage jointly decide whether the thesis survives.","claimIds":["CL-MUCC-004"]}],"riskClaimIds":["CL-MUCC-002"],"forwardTestIds":["FAL-MUCC-001","FAL-MUCC-002"]}},"dataGaps":[{"zh-CN":"美光尚未披露完整的 HBM 单品利润与现金流桥；2026-09-04 行情、市值和一致预期来自第三方快照，未与完全稀释股数及净现金逐项闭合。","en":"Micron does not disclose a complete standalone HBM profit and cash-flow bridge. The 2026-09-04 quote, market capitalisation and consensus inputs are third-party snapshots not fully reconciled to diluted shares and net cash."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/micron-hbm-capital-cycle-2026-09","en":"https://lunartuliplab.com/en/deep-dive/micron-hbm-capital-cycle-2026-09"},"machineReadable":"https://lunartuliplab.com/research/micron-hbm-capital-cycle-2026-09","dateModified":"2026-09-10"},{"id":"RO-COMP-SNDK-CAPITAL-CYCLE-001","slug":"sandisk-nand-capital-cycle-2026-09","kind":"company-deep-dive","tickers":["SNDK"],"publishedAt":"2026-09-10","asOf":"2026-09-07","version":"1.0.0","sourceLabel":"Memory capital-cycle series III · SanDisk NAND capital cycle · 2026-09-07 information set","claims":[{"id":"CL-SNDKCC-001","type":"Fact","text":{"zh-CN":"SanDisk FY2026 营收 202.48 亿美元，同比增长 175%；经营现金流 116.71 亿美元，购置固定资产 1.77 亿美元。","en":"SanDisk reported FY2026 revenue of $20.248 billion, up 175% year over year, operating cash flow of $11.671 billion and purchases of property, plant and equipment of $177 million."},"evidenceIds":["EV-SNDKCC-10K"]},{"id":"CL-SNDKCC-002","type":"Fact","text":{"zh-CN":"公司披露 10 份长期协议、8 家数据中心与 Edge 客户、最低定价下 939 亿美元预期收入、598 亿美元 RPO、12.42 亿美元合同负债及 15 亿美元客户保证金。","en":"The company disclosed ten long-term agreements with eight datacentre and edge customers, $93.9 billion of expected revenue at floor pricing, $59.8 billion of RPO, $1.242 billion of contract liabilities and $1.5 billion of customer deposits."},"evidenceIds":["EV-SNDKCC-10K"]},{"id":"CL-SNDKCC-003","type":"Derived","text":{"zh-CN":"报告 FCF 利润率约 56.77%；按研究稿口径剔除合资企业净出资与新商业模式现金影响后，调整后 FCF 利润率约 43.18%。两者均为周期时点口径，不应直接外推为永续利润率。","en":"Reported FCF margin was roughly 56.77%; after removing net joint-venture contributions and cash effects related to the new business model on the research note's basis, adjusted FCF margin was roughly 43.18%. Both are cycle-point readings and should not be extrapolated directly as perpetuities."},"evidenceIds":["EV-SNDKCC-10K"]},{"id":"CL-SNDKCC-004","type":"Hypothesis","text":{"zh-CN":"合同化真实抬高 NAND 周期下限，但能否成为年金取决于供给转松后客户是否继续按 floor 履约，以及 floor 对应毛利率是否足以支撑中期模型。","en":"Contracting has genuinely raised the NAND-cycle floor, but whether it becomes an annuity depends on customers continuing to perform at the floor after supply loosens and on the floor gross margin being sufficient to support the medium-term model."},"evidenceIds":["EV-SNDKCC-10K","EV-SNDKCC-MODEL"]}],"evidence":[{"id":"EV-SNDKCC-10K","title":"SanDisk FY2026 results and Form 10-K","source":{"publisher":"SanDisk Investor Relations","url":"https://investor.sandisk.com/static-files/47c2209d-7ef3-488d-b9f4-62d371ad8740","document":"FY2026 Form 10-K"},"sourceDate":"2026-08-17","dataAsOf":"2026-07-03","calculation":"(11671 - 177) / 20248 = 56.77%; (11494 - 275 - 2476) / 20248 = 43.18%.","confidence":"high","counterevidence":"Contract floors have not yet been tested through a NAND oversupply cycle, and floor gross margin is undisclosed.","lastVerified":"2026-09-10","role":"primary"},{"id":"EV-SNDKCC-MODEL","title":"SanDisk point-in-time research uniform and reverse-DCF-lite output","source":{"publisher":"LunarTulip Research","document":"SNDK capital-cycle uniform · 2026-09-07"},"sourceDate":"2026-09-07","dataAsOf":"2026-09-04","calculation":"WACC 9.5%; terminal growth 3%; ten-year window; EV approximated by market capitalisation before net cash/debt.","confidence":"medium","counterevidence":"The inverse problem is non-unique and market inputs came from a third-party point-in-time quote service.","lastVerified":"2026-09-10","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"FY2026 营收","en":"FY2026 revenue"},"period":"2026-FY","value":20248,"unit":"USDm","evidenceIds":["EV-SNDKCC-10K"]},{"label":{"zh-CN":"报告自由现金流","en":"Reported free cash flow"},"period":"2026-FY","value":11494,"unit":"USDm","evidenceIds":["EV-SNDKCC-10K"]},{"label":{"zh-CN":"调整后自由现金流","en":"Adjusted free cash flow"},"period":"2026-FY","value":8743,"unit":"USDm","evidenceIds":["EV-SNDKCC-10K"]}],"valuationScenarios":[{"id":"VAL-SNDKCC-REVERSE-DCF","label":{"zh-CN":"利润率永久化压力测试","en":"Margin-perpetuity stress test"},"metrics":[{"label":{"zh-CN":"隐含五年营收复合","en":"Implied five-year revenue CAGR"},"value":{"zh-CN":"8.98%","en":"8.98%"}},{"label":{"zh-CN":"隐含稳态 FCF 利润率","en":"Implied steady FCF margin"},"value":{"zh-CN":"56.77%","en":"56.77%"}}],"calculation":"Reverse-DCF-lite v1 at the 2026-09-04 point-in-time quote; WACC 9.5%, terminal growth 3%, ten-year window.","interpretation":{"zh-CN":"结果把现价的主要重量定位在利润率持续期，而非高营收复合；这是敏感性读数，不是目标价。","en":"The output locates most of the market value in margin duration rather than high revenue compounding. It is a sensitivity reading, not a price target."}},{"id":"VAL-SNDKCC-CASH-BASIS","label":{"zh-CN":"报告与调整后现金率","en":"Reported versus adjusted cash margin"},"metrics":[{"label":{"zh-CN":"报告 FCF 率","en":"Reported FCF margin"},"value":{"zh-CN":"56.77%","en":"56.77%"}},{"label":{"zh-CN":"调整后 FCF 率","en":"Adjusted FCF margin"},"value":{"zh-CN":"43.18%","en":"43.18%"}}],"calculation":"(11671 - 177) / 20248; (11671 - 177 - 275 - 2476) / 20248.","interpretation":{"zh-CN":"现金率差显示合同资金与合资投入会改变周期尖峰的可重复性判断。","en":"The cash-margin gap shows how contract funding and JV investment change the repeatability assessment at a cycle peak."}}],"counterevidenceClaimIds":["CL-SNDKCC-004"],"falsifiers":[{"id":"FAL-SNDKCC-001","metric":"4Q26 NAND contract-price change","operator":"<=","threshold":0,"unit":"percent","horizon":"October-November 2026","claimIds":["CL-SNDKCC-004"],"rationale":{"zh-CN":"合约价不再上涨，是供给纪律叙事第一根支柱断裂的高频信号。","en":"A non-positive contract-price change is the high-frequency signal that the first pillar of the supply-discipline narrative has broken."}},{"id":"FAL-SNDKCC-002","metric":"Disclosed floor gross margin","operator":"<","threshold":40,"unit":"percent","horizon":"upon first company disclosure","claimIds":["CL-SNDKCC-004"],"rationale":{"zh-CN":"floor 毛利率低于 40% 将直接削弱合同足以支撑 80% 中期毛利率目标的年金叙事。","en":"A floor gross margin below 40% would directly weaken the claim that contracting can support the 80% medium-term gross-margin model as an annuity."}}],"versions":[{"version":"1.0.0","date":"2026-09-10","diff":{"zh-CN":"将 2026-09-07 Desk 原稿按原文上架为独立双语 canonical 页面，并保留研究制服、现金口径桥、验证日历与 MU 姊妹篇关系。","en":"Published the 2026-09-07 Desk source as an independent bilingual canonical page, retaining its research uniform, cash-basis bridge, validation calendar and MU sibling relationship."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"合同抬高的是下限，市场定价的是尖峰：SanDisk 与 NAND 资本周期的压力测试","question":"NAND 的 take-or-pay 合同抬高了周期底部，还是把尖峰现金流提前资本化？","standfirst":"合同化已经写进资产负债表，但 floor 尚未经历供给过剩压力测试。报告与调整后自由现金流之间的差距，是判断年金叙事能否成立的第一道桥。","whyItMatters":"NAND 玩家更多、制程迁移本身能形成隐性扩产，供给纪律比 DRAM 更脆弱。SanDisk 自身不扩晶圆的纪律无法约束整个行业。","consensus":"长期协议、RPO、客户保证金与 FY28—30 中期模型共同支持周期下限抬高，市场已经采信这部分方向。","differentiated":"现价主要要求尖峰 FCF 利润率长期化；2H27 供给转松、floor 毛利率及合同负债变化是更关键的裁决变量。","valuation":"反向 DCF 输出约 9% 五年营收复合与约 57% 稳态 FCF 利润率，说明定价重量集中在利润率永久化；报告口径与调整后约 43% 的现金率必须分开。","update":"v1.0 基本原样保留 Desk 中文稿，新增 claim-parity 英文稿、公开官网 canonical、系列导航与伴生证据对象；不依赖或回链 Desk 私有 URL。","evidenceClaimIds":["CL-SNDKCC-001","CL-SNDKCC-002","CL-SNDKCC-003"],"causalChain":[{"title":"合同下限","body":"RPO、最低收入与保证金提高收入可见度。","claimIds":["CL-SNDKCC-002"]},{"title":"现金口径","body":"预收、押金与合资现金使报告 FCF 高于调整后捕获。","claimIds":["CL-SNDKCC-003"]},{"title":"过剩测试","body":"供给转松后 floor 履约与毛利率才决定合同是否接近年金。","claimIds":["CL-SNDKCC-004"]}],"riskClaimIds":["CL-SNDKCC-003"],"forwardTestIds":["FAL-SNDKCC-001","FAL-SNDKCC-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Contracts raise the floor; the market prices the peak: Stress-testing SanDisk and the NAND capital cycle","question":"Do NAND take-or-pay contracts raise the cycle floor, or capitalise peak cash flow in advance?","standfirst":"Contracting is already visible on the balance sheet, but the floor has not been stress-tested through oversupply. The gap between reported and adjusted free cash flow is the first bridge in judging whether the annuity narrative can hold.","whyItMatters":"NAND has more players, and process migration itself creates hidden supply. Industry discipline is more fragile than in DRAM, and SanDisk cannot impose its own wafer discipline on every supplier.","consensus":"Long-term agreements, RPO, customer deposits and the FY28–30 medium-term model support a higher cycle floor, and the market already accepts that direction.","differentiated":"The market primarily requires peak FCF margin to persist. The 2H27 supply window, floor gross margin and changes in contract liabilities are the more consequential adjudication variables.","valuation":"The reverse DCF produces roughly 9% five-year revenue CAGR and roughly 57% steady FCF margin, concentrating the valuation burden on margin perpetuity. Reported cash margin and the roughly 43% adjusted reading must remain separate.","update":"Version 1.0 retains the Desk Chinese source substantially as written, adds a claim-parity English edition, public-site canonical, series navigation and companion evidence object, and carries no dependency on or link back to the private Desk URL.","evidenceClaimIds":["CL-SNDKCC-001","CL-SNDKCC-002","CL-SNDKCC-003"],"causalChain":[{"title":"Contract floor","body":"RPO, minimum revenue and deposits improve revenue visibility.","claimIds":["CL-SNDKCC-002"]},{"title":"Cash basis","body":"Prepayments, deposits and JV cash make reported FCF exceed adjusted capture.","claimIds":["CL-SNDKCC-003"]},{"title":"Oversupply test","body":"Performance at the floor and floor gross margin after supply loosens decide whether the contracts resemble an annuity.","claimIds":["CL-SNDKCC-004"]}],"riskClaimIds":["CL-SNDKCC-003"],"forwardTestIds":["FAL-SNDKCC-001","FAL-SNDKCC-002"]}},"dataGaps":[{"zh-CN":"公司尚未披露 floor 对应毛利率，take-or-pay 条款也未经历 NAND 供给过剩期；HBF 尚无客户认证、收入或定价，因此不进入本对象事实层。","en":"The company has not disclosed gross margin at the floor, and the take-or-pay terms have not been tested through NAND oversupply. HBF has no disclosed customer qualification, revenue or pricing and therefore remains outside this object's fact layer."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/sandisk-nand-capital-cycle-2026-09","en":"https://lunartuliplab.com/en/deep-dive/sandisk-nand-capital-cycle-2026-09"},"machineReadable":"https://lunartuliplab.com/research/sandisk-nand-capital-cycle-2026-09","dateModified":"2026-09-10"},{"id":"RO-COMP-NVDA-FY27Q2-001","slug":"nvidia-fy27q2-supercycle-2026-09","kind":"company-deep-dive","tickers":["NVDA"],"publishedAt":"2026-09-14","asOf":"2026-09-14","version":"1.0.0","sourceLabel":"NVIDIA FY27Q2 supercycle · 2026-09-14 edition","claims":[{"id":"CL-NVDA-001","type":"Fact","text":{"zh-CN":"NVIDIA FY27 Q2 营收 $96,221M，同比 +106%；数据中心分部 $89.0B，同比 +117%。","en":"NVIDIA FY27 Q2 revenue was $96,221M, up 106% year over year; Data Center revenue was $89.0B, up 117%."},"evidenceIds":["EV-NVDA-001"]},{"id":"CL-NVDA-002","type":"Fact","text":{"zh-CN":"Q3 FY27 指引营收 $108.0B ± 2%，不假设任何来自中国的数据中心 compute 收入；Q3 毛利率(GAAP 与非 GAAP)74.0% ± 50bps。","en":"Q3 FY27 guidance is revenue of $108.0B ± 2% with no assumed China data-center compute revenue; Q3 gross margin (GAAP and non-GAAP) is 74.0% ± 50bps."},"evidenceIds":["EV-NVDA-001"]},{"id":"CL-NVDA-003","type":"Fact","text":{"zh-CN":"10-Q 披露供应采购承诺 $279B、担保义务上限 $108.5B（其中 $105B 对 SB Energy/OpenAI 关联园区），应收账款周转天数由 45 天升至 60 天。","en":"The 10-Q discloses supply purchase commitments of $279B, guarantee obligations capped at $108.5B (including $105B for the SB Energy/OpenAI-related campus), and receivable days sales outstanding rising from 45 to 60 days."},"evidenceIds":["EV-NVDA-002"]},{"id":"CL-NVDA-004","type":"Derived","text":{"zh-CN":"CFO 在 FY27 Q2 电话会给出 FY2028 公司总营收约 70% 增长展望（供给受限前提）；以 fwd PE 约 26x 定价隐含的是周期化形态而非纯软件久期。","en":"On the FY27 Q2 call the CFO guided to approximately 70% total company revenue growth in fiscal 2028, subject to supply constraints; pricing at roughly 26× forward earnings embeds a cyclical form rather than pure software duration."},"evidenceIds":["EV-NVDA-001","EV-NVDA-003"]},{"id":"CL-NVDA-005","type":"Inference","text":{"zh-CN":"需求侧方向验证已完成，分歧转移到报表形态：毛利率下行路径、担保集中与 DSO 拉长改变需求一旦降速时的受力方式。","en":"Directional demand validation is complete; the debate shifts to statement shape: the gross-margin path, guarantee concentration and longer DSO change how the balance sheet absorbs any demand slowdown."},"evidenceIds":["EV-NVDA-002","EV-NVDA-003"]},{"id":"CL-NVDA-006","type":"Hypothesis","text":{"zh-CN":"若 Q3 毛利率守住 74.0% ± 50bps 且 DSO 不再同向走高，FY2028 约 70% 总营收展望仍可作为倍数支撑；反之报表代价会先于需求证伪。","en":"If Q3 gross margin holds 74.0% ± 50bps and DSO stops rising in tandem, the roughly 70% FY2028 total revenue outlook can still support the multiple; otherwise balance-sheet costs reprice ahead of demand falsification."},"evidenceIds":["EV-NVDA-001","EV-NVDA-002"]}],"evidence":[{"id":"EV-NVDA-001","title":"NVIDIA Q2 FY27 press release","source":{"publisher":"U.S. SEC","url":"https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27pr.htm","document":"Q2 FY27 EX-99.1"},"sourceDate":"2026-08-26","dataAsOf":"2026-07-26","calculation":null,"confidence":"high","counterevidence":"Guidance assumes no China data-center compute revenue.","lastVerified":"2026-09-14","role":"primary"},{"id":"EV-NVDA-002","title":"NVIDIA Q2 FY27 Form 10-Q","source":{"publisher":"U.S. SEC","url":"https://www.sec.gov/Archives/edgar/data/1045810/000104581026000074/nvda-20260726.htm","document":"Q2 FY27 10-Q"},"sourceDate":"2026-08-26","dataAsOf":"2026-07-26","calculation":null,"confidence":"high","counterevidence":"Commitments and guarantees may not convert to losses in an upcycle.","lastVerified":"2026-09-14","role":"primary"},{"id":"EV-NVDA-003","title":"NVDA market multiples snapshot","source":{"publisher":"24/7 Wall St via public coverage","url":"https://247wallst.com/investing/2026/09/10/nvidia-nvda-stock-price-prediction-where-analysts-see-the-stock-going-by-2030/","document":"Forward multiple commentary"},"sourceDate":"2026-09-10","dataAsOf":"2026-09-10","calculation":"Forward PE about 26x cited from third-party coverage.","confidence":"medium","counterevidence":"Third-party multiples vary by EPS definition and price date.","lastVerified":"2026-09-14","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"FY27 Q2 营收","en":"FY27 Q2 revenue"},"period":"FY27-Q2","value":96221,"unit":"USDm","evidenceIds":["EV-NVDA-001"]},{"label":{"zh-CN":"数据中心营收","en":"Data Center revenue"},"period":"FY27-Q2","value":89000,"unit":"USDm","evidenceIds":["EV-NVDA-001"]},{"label":{"zh-CN":"Q3 毛利率指引中点","en":"Q3 gross margin guide midpoint"},"period":"FY27-Q3","value":74,"unit":"percent","evidenceIds":["EV-NVDA-001"]},{"label":{"zh-CN":"担保义务上限","en":"Guarantee obligation cap"},"period":"2026-07-26","value":108500,"unit":"USDm","displayValue":{"zh-CN":"$108.5B","en":"$108.5B"},"evidenceIds":["EV-NVDA-002"]}],"valuationScenarios":[{"id":"VAL-NVDA-BASE","label":{"zh-CN":"FY2028 展望兑现","en":"FY2028 outlook delivered"},"revenueBaseUsdM":384884,"revenueGrowthPct":70,"forwardRevenueUsdM":654302.8,"salesMultiple":8,"impliedEnterpriseValueUsdM":5234422.4,"calculation":"384884 * 1.7 = 654302.8; * 8 = 5234422.4","interpretation":{"zh-CN":"以 Q2 年化基数与 CFO 约 70% FY2028 总营收展望做敏感性，8 倍 sales multiple 为外生假设，不构成目标价。","en":"Sensitivity on an annualized Q2 base with the CFO's roughly 70% FY2028 total revenue outlook and an exogenous 8× sales multiple; not a target price."}},{"id":"VAL-NVDA-BEAR","label":{"zh-CN":"毛利率与报表代价","en":"Margin and statement cost"},"revenueBaseUsdM":384884,"revenueGrowthPct":35,"forwardRevenueUsdM":519593.4,"salesMultiple":5,"impliedEnterpriseValueUsdM":2597967,"calculation":"384884 * 1.35 = 519593.4; * 5 = 2597967","interpretation":{"zh-CN":"增长退到 35%、倍数 5 倍时企业价值约为基准的一半，对应毛利率路径加深与担保/DSO 风险被重新定价。","en":"At 35% growth and a 5× multiple, enterprise value falls near half the base case, mapping gross-margin path deepening and guarantee/DSO risk being repriced."}}],"counterevidenceClaimIds":["CL-NVDA-005"],"falsifiers":[{"id":"FAL-NVDA-001","metric":"Q3 gross margin (GAAP midpoint)","operator":"<","threshold":73.5,"unit":"percent","horizon":"Q3 FY27 earnings","claimIds":["CL-NVDA-006"],"rationale":{"zh-CN":"毛利率跌破指引区间下沿，FY2028 EPS 上修所依赖的毛利假设需要倒带。","en":"Gross margin below the guidance band would unwind the margin assumptions embedded in FY2028 EPS revisions."}},{"id":"FAL-NVDA-002","metric":"days sales outstanding","operator":">","threshold":65,"unit":"count","horizon":"Q3 FY27 earnings","claimIds":["CL-NVDA-005"],"rationale":{"zh-CN":"DSO 继续走高且收入增速回落，报表形态从观察项升级为一阶变量。","en":"DSO continuing higher while revenue growth slows would upgrade statement shape from observation to a first-order variable."}}],"versions":[{"version":"1.0.0","date":"2026-09-14","diff":{"zh-CN":"首版：FY27Q2 超级周期对账，需求硬数与报表代价分列。","en":"Initial edition: FY27Q2 supercycle reconciliation separating hard demand proof from statement cost."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"NVIDIA：增长有硬数，代价进报表","question":"FY2028 约 70% 总营收展望能否在毛利率不破下沿的前提下兑现？","standfirst":"FY27 Q2 营收 $96.2B 与数据中心 +117% 让需求侧不构成分歧；分歧在毛利率路径、$279B 供应承诺与 $108.5B 担保。","whyItMatters":"fwd PE 26 的便宜是周期化定价：市场买的是 FY2028 增长展望，同时折扣报表形态之变。","consensus":"Q3 指引 ±2% 超出财报前共识，Vera Rubin 全面量产。","differentiated":"我们把软肋从需求端移到成本与报表：DSO 45→60、担保 97% 挂单一项目。","valuation":"情景只做企业价值敏感性，不含每股结论与目标价。","update":"首版 asOf 2026-09-14。","evidenceClaimIds":["CL-NVDA-001","CL-NVDA-002","CL-NVDA-003","CL-NVDA-004"],"causalChain":[{"title":"需求半径扩大","body":"hyperscale 与主权/企业各占约一半。","claimIds":["CL-NVDA-001"]},{"title":"代价进报表","body":"供应承诺、担保与 DSO 改变降速时的受力方式。","claimIds":["CL-NVDA-003"]},{"title":"FY2028 展望受检验","body":"毛利率与 DSO 是 11 月下旬 Q3 的首读数。","claimIds":["CL-NVDA-006"]}],"riskClaimIds":["CL-NVDA-005"],"forwardTestIds":["FAL-NVDA-001","FAL-NVDA-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"NVIDIA: Hard growth, costs on the balance sheet","question":"Can the roughly 70% FY2028 total revenue outlook hold without gross margin breaking the lower band?","standfirst":"FY27 Q2 revenue of $96.2B and Data Center growth of +117% settle the demand debate; the dispute is gross-margin path, $279B supply commitments and $108.5B guarantees.","whyItMatters":"A forward PE near 26 is cyclical pricing: the market buys the FY2028 growth outlook while discounting statement-shape change.","consensus":"Q3 guidance of $108B ± 2% beat pre-report consensus and Vera Rubin entered full production.","differentiated":"We move the weak spot from demand to cost and statements: DSO 45→60 and 97% of guarantees on one project.","valuation":"Scenarios output enterprise-value sensitivity only; no per-share conclusion or target price.","update":"Initial edition asOf September 14, 2026.","evidenceClaimIds":["CL-NVDA-001","CL-NVDA-002","CL-NVDA-003","CL-NVDA-004"],"causalChain":[{"title":"Demand radius widens","body":"Hyperscale and sovereign/enterprise each contribute about half.","claimIds":["CL-NVDA-001"]},{"title":"Costs enter statements","body":"Supply commitments, guarantees and DSO change how a slowdown transmits.","claimIds":["CL-NVDA-003"]},{"title":"FY2028 outlook tested","body":"Gross margin and DSO are the first Q3 reads in late November.","claimIds":["CL-NVDA-006"]}],"riskClaimIds":["CL-NVDA-005"],"forwardTestIds":["FAL-NVDA-001","FAL-NVDA-002"]}},"dataGaps":[{"zh-CN":"本财年起数据中心 networking 收入不再单独拆分，只能通过电话会间接跟踪。","en":"Data-center networking revenue is no longer split out this fiscal year and can only be tracked indirectly through calls."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/nvidia-fy27q2-supercycle-2026-09","en":"https://lunartuliplab.com/en/deep-dive/nvidia-fy27q2-supercycle-2026-09"},"machineReadable":"https://lunartuliplab.com/research/nvidia-fy27q2-supercycle-2026-09","dateModified":"2026-09-14"},{"id":"RO-COMP-SNOW-AIDC-001","slug":"snowflake-agentic-data-cloud-2026-09","kind":"company-deep-dive","tickers":["SNOW"],"publishedAt":"2026-09-14","asOf":"2026-09-14","version":"1.0.0","sourceLabel":"Snowflake agentic data cloud · 2026-09-14 edition","claims":[{"id":"CL-SNOW-001","type":"Fact","text":{"zh-CN":"Snowflake Q2 FY27 产品收入 $1,491.9M，同比 +36.8%，第三个季度连续加速；FY27 产品收入指引上修至 $6.07B（+36%）。","en":"Snowflake Q2 FY27 product revenue was $1,491.9M, up 36.8% year over year, accelerating for a third straight quarter; FY2027 product revenue guidance was raised to $6.07B (+36%)."},"evidenceIds":["EV-SNOW-001"]},{"id":"CL-SNOW-002","type":"Fact","text":{"zh-CN":"NRR 126%，$1M+ 客户 828 家（+27%）；CoCo AI coding agent 账户数首次披露超过 9,100 个。","en":"NRR was 126% with 828 customers above $1M (+27%); CoCo AI coding agent accounts were disclosed for the first time at more than 9,100."},"evidenceIds":["EV-SNOW-001"]},{"id":"CL-SNOW-003","type":"Inference","text":{"zh-CN":"CEO 归因「AI 产品约占本轮加速的一半」，但源材料无 AI 收入金额披露；消费型计费下 37% 增速为真实用量，归因比例无法独立核验。","en":"The CEO attributed roughly half of the acceleration to AI products, yet source materials disclose no AI revenue amount; under consumption billing the 37% growth reflects real usage while the attribution share cannot be verified independently."},"evidenceIds":["EV-SNOW-001"]},{"id":"CL-SNOW-004","type":"Derived","text":{"zh-CN":"以 2026-09-04 收盘价反解，现价隐含未来五年营收复合约 37.8%、稳态 FCF 利润率 23.18%（WACC 9.5%、永续 3%）；该复合高于 FY27 +36% 指引，定价要求近乎零减速。","en":"Reverse-solving the September 4 close implies about 37.8% five-year revenue CAGR and a 23.18% steady-state FCF margin (9.5% WACC, 3% terminal growth); that CAGR exceeds the FY2027 +36% guide and prices near-zero deceleration."},"evidenceIds":["EV-SNOW-002","EV-SNOW-001"]},{"id":"CL-SNOW-005","type":"Hypothesis","text":{"zh-CN":"若 Q3 首次披露 AI 产品收入金额且产品收入守住 $1.588B 指引下限，「AI 数据云重加速」叙事可继续；否则 18.5x fwd P/S 的久期假设需要先下修。","en":"If Q3 discloses AI product revenue for the first time and product revenue holds the $1.588B guidance floor, the AI data cloud re-acceleration narrative can persist; otherwise the duration embedded in roughly 18.5× forward sales must come down first."},"evidenceIds":["EV-SNOW-001","EV-SNOW-002"]},{"id":"CL-SNOW-006","type":"Inference","text":{"zh-CN":"消费型模型下行无递延缓冲；RPO 环比回落与 113x fwd PE 叠加，使 Q3 $1.588B miss 会惩罚叙事而非仅惩罚业绩。","en":"Consumption billing offers no deferred cushion on the way down; falling RPO sequentially plus a forward PE near 113× means a Q3 miss below $1.588B punishes narrative, not just earnings."},"evidenceIds":["EV-SNOW-001","EV-SNOW-002"]}],"evidence":[{"id":"EV-SNOW-001","title":"Snowflake FQ2-27 press release","source":{"publisher":"Business Wire / Snowflake IR","url":"https://investors.snowflake.com/news/news-details/2026/Snowflake-Reports-Financial-Results-for-the-Second-Quarter-of-Fiscal-2027/default.aspx","document":"FQ2-27 earnings release"},"sourceDate":"2026-09-02","dataAsOf":"2026-07-31","calculation":null,"confidence":"high","counterevidence":"Management AI attribution is not independently auditable.","lastVerified":"2026-09-14","role":"primary"},{"id":"EV-SNOW-002","title":"Yahoo Finance SNOW snapshot","source":{"publisher":"Yahoo Finance","url":"https://finance.yahoo.com/quote/SNOW/","document":"SNOW quote and multiples"},"sourceDate":"2026-09-07","dataAsOf":"2026-09-04","calculation":"Reverse DCF implied 37.8% five-year CAGR from uniform research file.","confidence":"medium","counterevidence":"Reverse DCF solves one of several consistent paths; not unique.","lastVerified":"2026-09-14","role":"supporting"}],"financialBridge":[{"label":{"zh-CN":"Q2 FY27 产品收入","en":"Q2 FY27 product revenue"},"period":"FY27-Q2","value":1491.9,"unit":"USDm","evidenceIds":["EV-SNOW-001"]},{"label":{"zh-CN":"产品收入同比增速","en":"Product revenue YoY growth"},"period":"FY27-Q2","value":36.8,"unit":"percent","evidenceIds":["EV-SNOW-001"]},{"label":{"zh-CN":"NRR","en":"NRR"},"period":"FY27-Q2","value":126,"unit":"percent","evidenceIds":["EV-SNOW-001"]},{"label":{"zh-CN":"Q3 产品收入指引下限","en":"Q3 product revenue guide low"},"period":"FY27-Q3","value":1588,"unit":"USDm","evidenceIds":["EV-SNOW-001"]}],"valuationScenarios":[{"id":"VAL-SNOW-BASE","label":{"zh-CN":"零减速路径","en":"Zero-deceleration path"},"metrics":[{"label":{"zh-CN":"隐含五年营收 CAGR","en":"Implied 5Y revenue CAGR"},"value":{"zh-CN":"37.8%","en":"37.8%"}},{"label":{"zh-CN":"FY27 产品收入指引增速","en":"FY2027 product revenue guide"},"value":{"zh-CN":"+36%","en":"+36%"}}],"calculation":"Reverse DCF from uniform file; WACC 9.5%, terminal 3%, cap 10y.","interpretation":{"zh-CN":"现价隐含的增长久期高于公司 +36% 指引，差额是未披露 AI 收入口径的期权；非目标价。","en":"The price embeds longer growth duration than the company's +36% guide; the gap is option value for undisclosed AI revenue lines, not a target price."}},{"id":"VAL-SNOW-BEAR","label":{"zh-CN":"回归板块中位倍数","en":"Revert to sector median multiple"},"metrics":[{"label":{"zh-CN":"fwd P/S（现值）","en":"Forward P/S (current)"},"value":{"zh-CN":"约 18.5x","en":"about 18.5×"}},{"label":{"zh-CN":"板块中位 fwd P/S","en":"Sector median forward P/S"},"value":{"zh-CN":"10-12x","en":"10-12×"}}],"calculation":"Multiple compression sensitivity from research note; no per-share target.","interpretation":{"zh-CN":"若 Q3 无法金额化 AI 收入，fwd P/S 向 10-12x 收敛对应三成以上估值收缩——框架敏感性，非点位预测。","en":"If Q3 fails to monetize AI revenue in disclosure, forward sales compressing toward 10-12× maps to more than 30% valuation contraction in framework sensitivity, not a point forecast."}}],"counterevidenceClaimIds":["CL-SNOW-006"],"falsifiers":[{"id":"FAL-SNOW-001","metric":"product revenue YoY growth","operator":"<","threshold":35,"unit":"percent","horizon":"Q3 FY27 earnings","claimIds":["CL-SNOW-005"],"rationale":{"zh-CN":"产品收入同比跌破 35% 会动摇「三季连加速」与零减速定价的地基。","en":"Product revenue growth below 35% would undermine three-quarter acceleration and zero-deceleration pricing."}},{"id":"FAL-SNOW-002","metric":"NRR","operator":"<","threshold":125,"unit":"percent","horizon":"Q3 FY27 earnings","claimIds":["CL-SNOW-002"],"rationale":{"zh-CN":"NRR 跌破 125 说明消费型扩张未能抵消优化周期风险。","en":"NRR below 125 would show consumption expansion failing to offset optimization-cycle risk."}}],"versions":[{"version":"1.0.0","date":"2026-09-14","diff":{"zh-CN":"首版：AI 数据云重加速对账，加速硬数与 AI 归因缺口分列。","en":"Initial edition: AI data cloud re-acceleration reconciliation separating hard acceleration from the AI attribution gap."}}],"renderings":{"zh-CN":{"locale":"zh-CN","siblingLocale":"en","title":"Snowflake：加速有硬数，归因只有一句话","question":"Q3 能否首次披露 AI 产品收入金额？","standfirst":"产品收入 +37% 三季连加速经得起核查；「一半来自 AI」只有管理层归因。","whyItMatters":"消费型计费让增速等于真实用量，也让下行没有缓冲；现价隐含五年复合 37.8%。","consensus":"财报次日 +16.5%，FY27 指引上修至 +36%。","differentiated":"我们把证真开关定为 AI 收入金额化，而非 CoCo 账户数。","valuation":"情景用 metrics 路径表达倍数与隐含 CAGR，不含目标价。","update":"首版 asOf 2026-09-14。","evidenceClaimIds":["CL-SNOW-001","CL-SNOW-002","CL-SNOW-003","CL-SNOW-004"],"causalChain":[{"title":"用量加速","body":"消费型收入三连加速。","claimIds":["CL-SNOW-001"]},{"title":"归因缺口","body":"AI 一半归因无金额口径。","claimIds":["CL-SNOW-003"]},{"title":"Q3 裁决","body":"指引下限与 AI 披露是 12 月初首读数。","claimIds":["CL-SNOW-005"]}],"riskClaimIds":["CL-SNOW-006"],"forwardTestIds":["FAL-SNOW-001","FAL-SNOW-002"]},"en":{"locale":"en","siblingLocale":"zh-CN","title":"Snowflake: Hard acceleration, soft attribution","question":"Will Q3 disclose AI product revenue for the first time?","standfirst":"Three quarters of accelerating product revenue near +37% is verifiable; the half-from-AI line is management attribution only.","whyItMatters":"Consumption billing makes growth equal real usage and leaves no cushion on the way down; the price embeds a 37.8% five-year CAGR.","consensus":"The stock rose 16.5% the day after earnings and FY2027 guidance was raised to +36%.","differentiated":"We set monetized AI revenue disclosure as the falsification switch, not CoCo account counts.","valuation":"Scenarios use the metrics path for multiples and implied CAGR; no target price.","update":"Initial edition asOf September 14, 2026.","evidenceClaimIds":["CL-SNOW-001","CL-SNOW-002","CL-SNOW-003","CL-SNOW-004"],"causalChain":[{"title":"Usage accelerates","body":"Consumption revenue accelerates for three quarters.","claimIds":["CL-SNOW-001"]},{"title":"Attribution gap","body":"Half-from-AI lacks a revenue line.","claimIds":["CL-SNOW-003"]},{"title":"Q3 adjudication","body":"Guidance floor and AI disclosure are the early-December reads.","claimIds":["CL-SNOW-005"]}],"riskClaimIds":["CL-SNOW-006"],"forwardTestIds":["FAL-SNOW-001","FAL-SNOW-002"]}},"dataGaps":[{"zh-CN":"AI 产品无独立收入披露；reverse DCF 基数为 TTM 总营收而非产品收入指引口径。","en":"AI products have no standalone revenue disclosure; the reverse DCF base uses TTM total revenue rather than product-revenue guidance."}],"canonical":{"zh-CN":"https://lunartuliplab.com/deep-dive/snowflake-agentic-data-cloud-2026-09","en":"https://lunartuliplab.com/en/deep-dive/snowflake-agentic-data-cloud-2026-09"},"machineReadable":"https://lunartuliplab.com/research/snowflake-agentic-data-cloud-2026-09","dateModified":"2026-09-14"}]}