The one thing
Marvell printed a record quarter after Thursday’s close — revenue $2.739bn, up 37%, data centre up 46% — guided the October quarter to $3.150bn, which is 15% sequential growth, and raised its revenue outlook for both fiscal 2027 and fiscal 2028 for the second consecutive quarter. The stock is down about 8% pre-market. Nothing in the release is bad. The beat was $39m above the guidance midpoint, about 1.4%, and the market has decided that is not enough.
That is the morning’s real information, and it arrives 36 hours after the opposite signal. On Wednesday night Nvidia beat and guided a full year ahead for the first time; Thursday the Nasdaq Composite closed up 1.57% and the S&P 500 up 0.72% — but breadth was negative underneath: Nvidia alone closed up 8.74% and added roughly $442bn of market value, technology was the only advancing S&P sector, and the average stock fell. Two nights, two beats, one narrow winner and one 8% loser. The bar is no longer “beat”; it is “beat by more than the last four weeks of price action already assumed.” Marvell had run about 28% in a month into the print.
The second-order detail is the one worth carrying into the model. Marvell guided October-quarter non-GAAP gross margin to 57.5–58.5%, down from 58.9% just delivered — falling margin on 15% sequential revenue growth. Nvidia said the same thing on Wednesday about memory. Two different points in the chain, both conceding margin to input costs in the same week, while Korean memory names sold off overnight. Asia was mixed and negative in chips; US futures were slightly soft pre-open, with the Nasdaq-100 contract off 0.32% and S&P and Dow contracts close to flat.
Sector pulse — covered names with material news
| Semis & equipment | 3 / 16 |
|
| Cloud & AI infrastructure | 1 / 8 |
|
| SaaS & enterprise software | 1 / 11 |
|
| Ads & adtech | 0 / 5 | |
| Consumer platforms & commerce | 0 / 5 | |
| Hardware & devices | 0 / 3 |
Overnight & last session
Marvell FQ2 2027: record $2.739bn, both fiscal years raised — and the stock down 8%
Net revenue of $2.739bn was up 37% year on year and $39.0m above the midpoint of the guidance given on 27 May. Data centre revenue growth accelerated to 46% year on year and the segment is now around 79% of the business. GAAP diluted EPS was $0.33 on net income of $308.0m; non-GAAP EPS $0.94 on $865.9m. GAAP gross margin 53.1%, non-GAAP 58.9%. Cash from operations $605.5m. October-quarter guidance is $3.150bn ±5% — 15% sequential growth — with non-GAAP EPS of $1.10 ±$0.05. CEO Matt Murphy said AI-related bookings “remain exceptionally robust” and that Marvell is again raising its outlook for both FY2027 and FY2028, citing “a significant acceleration in our Custom business beginning in the second half of fiscal 2027.”
Two things sit under the sell-off. First, non-GAAP gross margin is guided to 57.5–58.5% against the 58.9% just delivered — margin down while revenue accelerates 15% sequentially, which is what a mix shift toward lower-margin custom silicon looks like in the accounts. Second, the gap between GAAP $0.33 and non-GAAP $0.94 is the widest in coverage this week: $558m of the quarter’s earnings sit in the reconciliation, and Marvell now also calculates EPS under the two-class method following the Series A convertible preferred issued on 31 March 2026.
The read: This is the first clean instance of the AI trade repricing a genuinely good semis quarter, and it is worth writing while it is fresh. The publishable question is not “why did it fall” — that is positioning, and everyone will say so. It is what does a falling gross margin on accelerating custom-silicon revenue tell you about who captures the value in a custom ASIC. Marvell is being paid more and keeping less of each dollar, at exactly the moment its custom business inflects. Set that against Broadcom, which reports on 2 September and guided FY2027 AI semis above $100bn at far higher margins, and the two together answer a question CoreValue’s coverage is unusually well placed to ask: is custom silicon a good business, or is it a good business only for the incumbent with 70–80% share? Marvell’s Investor Day on 6 October is where the long-term margin structure gets defended, and that is the date to hold this piece for.
Nvidia’s session: a near-record single-day value gain on negative breadth, and $108.5bn of guarantees in the CFO commentary
Nvidia closed Thursday at $227.98, up 8.74% and adding roughly $442bn of market value. The VanEck Semiconductor ETF rose 3.10%, Broadcom closed up 4.49% and AMD down 0.89%. Technology was the only advancing S&P sector. On the call CFO Colette Kress reframed the customer mix: hyperscalers remain a major growth driver, but the ACIE segment — sovereign, regional neoclouds, enterprise edge and air-gapped data centres — “will represent roughly half of our data center business.” It already does, at $40.3bn of the $89.0bn quarter.
The under-read part of the release is the CFO commentary’s commitments section. Nvidia’s supply and capacity commitments alone rose from $119bn last quarter to $279bn, “primarily related to the procurement of memory,” inside total future commitments of $366bn — and a second, separate table adds $36bn of AI cloud agreements and $20bn of data-centre leases, taking total contractual commitments to roughly $422bn. Days sales outstanding went from 45 to 60 sequentially “due to extended payment terms on large, multi-quarter agreements with certain investment-grade customers,” on receivables of $63.1bn. Inventory rose to $31.6bn from $25.8bn. And Nvidia disclosed $108.5bn of maximum gross guarantee exposure — $3.5bn for AI-cloud partners’ land, power and shell obligations (partly mitigated by $712m held in escrow), plus $105bn on SB Energy’s PORTS Technology Campus in Pike County, Ohio, roughly 4.25 GW hosting Nvidia infrastructure under 20-year leases to an OpenAI affiliate. Nvidia’s own words for why: AI clouds and model makers “are growing faster than their balance sheets and long-term credit profiles can support.”
Two qualifications that matter and are easy to lose. The SB Energy guarantees were entered into in August 2026 — after the 26 July quarter end, so they are a subsequent event rather than a balance-sheet item for the quarter just reported. And they become effective in nine phases as data centres reach service, the first expected in fiscal 2029. The $108.5bn is a maximum gross notional over two decades, not current exposure. Anyone writing this number without those two sentences is overstating it.
The read: Read those four disclosures as one sentence and Nvidia is prepaying its suppliers, extending credit to its customers, and guaranteeing its customers’ landlords — all at once. It has become the working-capital clearing house for the entire build-out, and the $108.5bn guarantee number is roughly 8% of the market value it added in a single session on Thursday. That is the story below, and it is genuinely unowned: the market spent Thursday pricing the 70% growth guide and skipped the balance sheet entirely. Separately, Kress’s “roughly half from non-hyperscalers” line deserves testing rather than repeating — sovereign and neocloud demand is real, but it is also the part of the customer base with the weakest credit, which is precisely why the guarantees exist. The two disclosures explain each other.
Korean memory names lead Asia lower — into the tightest memory market in a decade
The KOSPI opened down 0.95% at 6,846.54 with Samsung Electronics and SK hynix deepening chip-led declines. The policy backdrop is a day old, not overnight: the Bank of Korea raised its rate 25bp to 3.00% on a 6–1 vote on Thursday 27 August, its second consecutive rise, and lifted its 2026 growth forecast to 3.3% from 2.6% — so it was already in the price at Friday’s open, and this morning’s decline is a chip move rather than a rates move. Japan’s Nikkei held above 66,500. Mainland China benchmarks closed higher on AI-hardware strength following Nvidia; Hong Kong lagged. The backdrop is a memory market running at the tightest conditions in a decade: TrendForce has conventional DRAM contract prices rising 13–18% quarter on quarter and NAND 10–15%, with HBM4 for Nvidia GPUs indicated near $31–32/GB against HBM3e at $17–18/GB, and HBM effectively sold out for 2026 under multi-year agreements. The cost is already showing up on the other side of the ledger: Qualcomm, the only revenue decliner in the coverage list at −4%, attributes the pressure to memory and input-cost inflation and has carried it into its own guidance.
The read: The two halves of this item point in opposite directions and that is the point. Memory pricing is the strongest fundamental in the coverage list — Nvidia has just committed $279bn largely to secure it, which is a customer validating the supplier’s pricing power in a filing — while the equities are being sold. A de-rating of businesses whose earnings are inflecting is the setup the Micron and SK hynix initiations were written for, and both are built and unpublished. The specific claim worth making, because nobody is making it in these terms: Nvidia’s commitments line is the largest single third-party validation of the memory cycle in print, and it is in a CFO commentary rather than a memory company’s release, which is why it has gone unnoticed. Micron reports 30 September and is the direct test.
Nvidia names the levered periphery’s problem out loud — and then underwrites it
Nvidia’s disclosure that AI clouds “are growing faster than their balance sheets and long-term credit profiles can support” is the clearest description of the CoreWeave structure yet published, and it comes from the vendor rather than a short seller. CoreWeave’s own second quarter, reported 11 August, had revenue of $2,575m up 112%, backlog of $104.2bn up 246%, net interest expense of $640m against a net loss of $(626)m, and total debt around $35bn. It guides 2026 revenue of $12.4–13.2bn against capex of $35–39bn. Nvidia has separately signed $36bn of AI cloud agreements and $20bn of data-centre leases it expects to reassign to third parties — commitments that exist to give neoclouds access to infrastructure their own credit could not finance.
The read: The AI Bubble Radar’s CoreWeave tripwire now has a vendor-sourced confirmation, which is a materially stronger form of evidence than the third-party credit commentary the radar currently leans on. Worth updating the tripwire language to quote Nvidia’s own sentence. The analytical point for a note: vendor credit support does not remove the risk, it relocates it from the neocloud’s balance sheet to Nvidia’s — and Nvidia is now carrying it in three forms simultaneously (equity stakes, lease reassignment obligations, and $108.5bn of guarantees) while its receivable days stretch from 45 to 60. Oracle sits in the same frame from the other side, funding its build with roughly $75bn of customer prepayments. Oracle’s FQ1 print in September is the next hard datapoint on whether contracted backlog is converting to cash.
Workday lifts the margin guide, barely moves the revenue guide — and backlog is growing slower than revenue
Second-quarter revenue of $2.649bn was up 12.8%, with subscription revenue of $2.471bn up 13.9%. GAAP operating income $313m at an 11.8% margin, against $248m and 10.6% a year ago; non-GAAP operating income $824m at 31.1% against 29.0%. Diluted EPS of $2.57 against $0.84. Full-year FY2027 subscription revenue guidance moved to $9.94–9.95bn, about 13% growth. Read that guidance move carefully. The prior range was $9.925–9.950bn: the floor rose $15m, the ceiling did not move at all, and the midpoint went up $7.5m. That is a narrowing described as a raise. The genuine upgrade was to margin — the non-GAAP operating margin outlook went from 30.5% to 31.0%. The company said AI drove more than 25% of new annual contract value, with over 5,500 customers using at least one of its agents.
The number that does not fit the story: total subscription revenue backlog of $27.403bn grew 8.0% year on year — five and a half points slower than the subscription revenue it is meant to lead. Margin expanded roughly 210bp; growth did not.
The read: Workday is the cleanest seat-based control for the consumption-based AI narrative, and this print gives the comparison teeth. Salesforce, two nights earlier, put cRPO at +14% against revenue at +11% — bookings running ahead of revenue. Workday is the mirror image: revenue at +13% against backlog at +8%, bookings running behind. Both companies are claiming AI is driving demand; only one has a forward-commitment number that supports it. That is a two-name note that writes itself off data already in the workbooks, and it sharpens a question worth asking of the whole SaaS sector: when a seat-based vendor sells AI agents, does it expand the contract or replace the seats it was already being paid for? Workday’s 25%-of-new-ACV disclosure is impressive precisely because it says nothing about whether that ACV is incremental. Watch the same tension at Snowflake and MongoDB when they report next week.
Full-year capex guidance — hyperscalers and largest semis
Where each name stands
| Name | Period | Reported | Key figures from that print |
|---|---|---|---|
| Semis & equipment | |||
| NVDA | FQ2 FY27 | 26 Aug 26 | Revenue $96,221m +106%; Data Center $89,023m +117% (Hyperscale $48,710m, ACIE $40,313m); GM 75.0%; GAAP dil. EPS $2.46 vs non-GAAP $2.22; FQ3 guide $108.0bn ±2% at 74.0% GM, ex-China; FY28 growth guided ~70%. Commitments $119bn→$279bn (mostly memory), total $366bn; guarantees $108.5bn; DSO 45→60 days; $25bn notes issued |
| AVGO | FQ2 FY26 | 3 Jun 26 | Revenue $22.19bn +48%; AI semis $10.8bn +143%; FQ3 guide $29.4bn with AI semis ~$16.0bn; FY26 AI semis ~$56bn; FY27 AI semis “>$100bn”. FQ3 due 2 Sep |
| AMD | Q2 CY26 | 4 Aug 26 | Revenue $11,536m +50%; Data Center $6,718m +107%; Q3 guide ~$13.0bn ±$300m |
| QCOM | FQ3 FY26 | 29 Jul 26 | Revenue $9,947m −4% — the only decliner in coverage; QCT $8,504m −5% (handsets $5,086m −20%, auto $1,588m +61%, IoT $1,830m +9%); FQ4 guide $9.7–10.5bn. Management attributes pressure to memory and input-cost inflation, also reflected in the guide |
| MU | FQ3 FY26 | 24 Jun 26 | Revenue $41,456m +345.7%; GAAP GM 84.6% / non-GAAP 84.9%, from 37.7%/39.0%; Cloud Memory BU $13,769m; FQ4 guide $50.0bn ±$1.0bn at ~86% GM. No DRAM/NAND split is disclosed any more — four business units replaced it. Bhatia to President & COO, DeBoer to President & CTPO (26 Aug). FQ4 due 30 Sep |
| INTC | Q2 CY26 | 23 Jul 26 | Revenue $16,128m +25% — flattered by the Altera deconsolidation, which Intel flags; GAAP GM 40.4%, +12.9pp; Intel Foundry $5,765m +31% on a $2,089m operating loss (derived from the segment table, not a stated line). $20bn common equity raised mid-August |
| TXN | Q2 CY26 | 22 Jul 26 | Revenue $5,463m +23%; Analog $4,365m +26% on operating profit $1,992m +50%; Q3 guide $5.65–6.15bn, EPS $2.23–2.57 |
| MRVL | FQ2 FY27 | 27 Aug 26 | Revenue $2.739bn +37%, $39.0m above guide midpoint; data centre +46% y/y, ~79% of revenue; GAAP GM 53.1% / non-GAAP 58.9%; GAAP EPS $0.33 / non-GAAP $0.94; CFO $605.5m; FQ3 guide $3.150bn ±5% (+15% q/q) at 57.5–58.5% non-GAAP GM and $1.10 ±$0.05; FY27 and FY28 outlooks both raised. Investor Day 6 Oct |
| AMAT | FQ3 FY26 | 13 Aug 26 | Revenue $9,115m +25%, a record; FQ4 guide $10,250m ±$500m with non-GAAP EPS $4.02 ±$0.20. China share is not in the release — it is in the 10-Q filed 20 Aug |
| KLAC | FQ4 FY26 | 28 Jul 26 | Revenue $3,658m +15.2%; FQ1 guide $4.0bn ±$200m at 62.5% ±1.0% non-GAAP GM. The release states no historical gross margin; blank rather than derived. 10-for-1 split effected 11 Jun 26 breaks per-share series |
| LRCX | FQ4 FY26 | 29 Jul 26 | Revenue $6,722.2m +30.0%, a record; China 26% of revenue, down from 34%; Sep-quarter guide $8.10bn ±$400m. Lam labels this “the June 2026 quarter” |
| ASML | Q2 CY26 · EUR | 15 Jul 26 | Revenue €9,326.5m +21.3%; net system sales €6,564.8m on 91 lithography systems; Q3 guide €11.0–12.0bn at 55–57% GM; FY26 raised to €43–45bn. Net bookings not disclosed — only a qualitative line that order intake “remained extremely strong” |
| ARM | FQ1 FYE27 | 29 Jul 26 | Revenue $1.29bn +22% (record); royalty $715m +22%, licensing $574m +23%; non-GAAP operating margin 41.2% vs GAAP 7.1% |
| TSM | Q2 CY26 · TWD | 16 Jul 26 | Revenue NT$1,270.4bn +36.0% (US$40.2bn); GM 67.7%; operating margin 60.3%; ≤7nm = 77% of wafer revenue; HPC 66% of net revenue. 2026 capex raised to $60–64bn; 2026 growth guided slightly above 40% in USD |
| TSM — monthly | Jul 2026 | 10 Aug 26 | NT$467.58bn, +44.7% y/y, +5.6% m/m; Jan–Jul cumulative NT$2,872.06bn +37.0%. August print due ~10 Sep |
| CDNS | Q2 CY26 | 27 Jul 26 | Revenue $1,584m +24.2%; record backlog $8.1bn; cRPO $4.2bn; FY26 raised to $6.26–6.34bn, ~19% growth |
| SNPS | FQ3 FY26 | 26 Aug 26 | Revenue $2,477m +42% vs $1,740m; Design Automation $2,003.0m, Design IP $473.8m (Ansys sits inside Design Automation); non-GAAP EPS $3.91, above the high end of guidance; FY26 raised to ~$9.715bn revenue. Stock fell on the print |
| Cloud & AI infrastructure | |||
| AMZN | Q2 CY26 | 30 Jul 26 | Revenue $200.6bn +20%; AWS $42.2bn +37% (fastest in 18 quarters, ~$169bn annualised run rate); AWS operating income $16.6bn. CY26 cash capex ~$220bn, raised from ~$200bn on memory prices. TTM free cash flow is an outflow of $7.6bn on a $66.1bn y/y rise in PP&E purchases; Q2 net income of $62.6bn includes $53.4bn of non-operating other income, largely the Anthropic stake |
| MSFT | FQ4 FY26 | 29 Jul 26 | Revenue $90.0bn +18% (+17% cc); Microsoft Cloud $59.3bn +27%; Azure and other cloud +43%; Intelligent Cloud $39.3bn +32%; commercial RPO $678bn +84%. CY26 capex ~$175bn, down from $190bn on a 15→25yr useful-life change — not a spending cut. Quarter includes a $3.2bn Anthropic gain |
| GOOGL | Q2 CY26 | 22 Jul 26 | Revenue $119,796m +24%; Google Cloud $24,768m +82%; total revenue backlog $519.5bn, of which $513.9bn is Cloud; Q2 capex $44.9bn; CY26 capex raised to $195–205bn from $180–190bn; $84.75bn equity raise Jun 2026. ~75% of revenue is still ads |
| ORCL | FQ4 FY26 | 10 Jun 26 | Revenue $19.2bn +21% (+20% cc); RPO $638bn +363%, up $85bn q/q (incl. ~$75bn of prepayments and customer-supplied hardware); OCI $5.8bn +93%, total cloud $9.9bn +47%; FY26 capex $55,663m vs $21,215m (+162%) on free cash flow of −$23.7bn; FY27 net cash capex ~$70bn, implying ~$90–95bn gross; ~$40bn of debt and equity planned for FY27. The 10-K also discloses $260bn of additional lease commitments not on balance sheet, commencing FQ1 FY27 to FY2029, plus $19bn of unconditional cloud-infrastructure purchase commitments entered after 31 May |
| NET | Q2 CY26 | 6 Aug 26 | Revenue $696.1m +36%; current RPO +35%; non-GAAP operating income $96.1m at 13.8% margin; non-GAAP GM 73.1%, down from 76.3%. Cloudflare no longer quantifies large-customer count or DBNER in the release — the +27% / 4,698 figures previously carried here are not in this quarter’s disclosure and have been removed. |
| VRT | Q2 CY26 | 29 Jul 26 | Revenue $3,274m +24% reported / +18% organic; adj. operating margin 22.6% (+410bp); adj. FCF $925m +234%; FY26 sales guide $14.0bn. Orders and backlog are no longer disclosed anywhere — absent from both the release and the Q2 10-Q |
| GEV | Q2 CY26 | 22 Jul 26 | Revenue $11.1bn +22% reported / +12% organic; orders $24.2bn +88% organic; backlog $176bn, up $13.0bn q/q. Power: orders $16.7bn +134% organic on revenue $5.5bn +14%. Electrification: orders $6.3bn +66% organic on revenue $3.6bn +68% |
| CRWV | Q2 CY26 | 11 Aug 26 | Revenue $2,575m +112%; backlog $104.2bn +246%; Q2 capex $6,422m; net loss $(626)m against net interest expense of $(640)m; total debt $35.07bn ($31.4bn recourse + $3.7bn non-recourse). FY26 revenue guide $12.4–13.2bn, capex $35–39bn. $2.6bn DDTL closed 10 Aug at SOFR+5.50%, ~5yr vs ~3yr contracts |
| SaaS & enterprise software | |||
| CRM | FQ2 FY27 | 26 Aug 26 | Revenue $11.3bn +11% (incl. $456m Informatica); cRPO $33.5bn +14%; RPO $66.3bn +11%; GAAP op. margin 20.5% / non-GAAP 34.1%; GAAP EPS $4.29 / non-GAAP $5.90; FY27 guide raised to $46.1–46.4bn (a $200m raise, $300m CC); Agentforce ARR >$1.5bn on a widened definition |
| ADBE | FQ2 FY26 | 11 Jun 26 | Revenue $6.62bn +13% (+11% cc); total ending ARR $27.10bn, including ~$480m from Semrush; Creative & Marketing Professionals subscription revenue $4.54bn +13%, Business Professionals & Consumers $1.85bn +16%. Digital Media / Digital Experience segmentation dropped for two customer groups. GAAP EPS includes a $0.17/share goodwill impairment; CFO Dan Durn departed 15 Jun 26 |
| NOW | Q2 CY26 | 22 Jul 26 | Revenue $3,987m +24%; subscription $3,877m +24.5%; cRPO $13.20bn +21%; 658 customers above $5m ACV |
| SNOW | FQ1 FY27 | 27 May 26 | Revenue $1.39bn +33%; product revenue $1.334bn +34%; NRR 126%; RPO $9.21bn +38%. FQ2 due 2 Sep |
| MDB | FQ1 FY27 | 28 May 26 | Revenue $687.6m +25%; Atlas $512.5m +29%; 67,700+ customers; 2,895 above $100k ARR. FQ2 due 1 Sep |
| HUBS | Q2 CY26 | 5 Aug 26 | Revenue $911.7m +20%; subscription $894.0m +20%; 306,446 customers +14%; ASRPC $11,800 +4% |
| DDOG | Q2 CY26 | 6 Aug 26 | Revenue $1.12bn +36%; ~4,720 customers above $100k ARR, +23%; FCF $279m. GAAP operating income $5m (~0%) against 23% non-GAAP |
| WDAY | FQ2 FY27 | 27 Aug 26 | Revenue $2.649bn +12.8%; subscription $2.471bn +13.9%; GAAP op. income $313m (11.8%) / non-GAAP $824m (31.1%); dil. EPS $2.57; subscription backlog $27.403bn +8.0%; FY27 subscription guide raised to $9.94–9.95bn at 31.0% non-GAAP margin; AI >25% of new ACV, 5,500+ agent customers |
| SAP | Q2 CY26 · EUR | 23 Jul 26 | Revenue €9.878bn +9% (+11% cc); cloud revenue €6.281bn +22%; current cloud backlog €22.9bn +27% |
| PANW | FQ3 FY26 | 2 Jun 26 | Revenue $3.0bn +31%; NGS ARR $8.1bn +60%; RPO $18.4bn +36%. CyberArk and Chronosphere contributed $388m — not organic. FQ4 due 1 Sep |
| CRWD | FQ2 FY27 | 26 Aug 26 | Revenue $1,470.9m +26%; net new ARR $332.8m +51% (record); ending ARR $5.84bn +25%; FCF $377.4m; first GAAP profit since Jul-2024 ($5.3m); FY27 ARR guide $6,603–6,612m. 4-for-1 split 2 Jul 2026 breaks per-share series |
| Ads & adtech | |||
| META | Q2 CY26 | 29 Jul 26 | Revenue $60,801m +28% (+27% cc); Family DAP 3.60bn +3%. Income from operations fell 8% to $18,775m on $2.40bn of legal charges and $1.18bn of May-2026 severance; Q2 capex $31.08bn incl. finance-lease principal left free cash flow of just $784m. CY26 capex $130–145bn, narrowed from $125–145bn; Q3 guide $61–64bn. 26 Aug: $16.7bn/10yr settlement with 51 state AGs |
| ZETA | Q2 CY26 | 4 Aug 26 | Revenue $442.8m +43.5%, heavily M&A-assisted — ex-M&A and political, guidance implies 24–25% FY growth; 197 super-scaled customers +17%; FY26 guide raised to $1,811–1,824m |
| TTD | Q2 CY26 | 6 Aug 26 | Revenue $715m +3% — weakest in coverage; adj. EBITDA $241m at 34% margin, down from $271m and 39%; Q3 guide of “at least $650m” implies a year-on-year decline. Netflix joined the marketplace |
| ROKU | Q2 CY26 | 6 Aug 26 | Revenue $1,354.7m +22%; platform $1,221.0m +25% (advertising $672.8m +25%, subscriptions $548.2m +26%); 37.9bn streaming hours +7%. No forward outlook given; household count discontinued since Q1 2025 |
| NFLX | Q2 CY26 | 16 Jul 26 | Revenue $12,560m +13.4% (+12% FX-neutral); operating margin 33.4%, down from 34.1%; FY26 revenue guide narrowed to $51.0–51.4bn with FY operating margin held at 31.5% and ads revenue roughly doubling to ~$3bn. Membership counts not disclosed |
| Consumer platforms & commerce | |||
| ABNB | Q2 CY26 | 6 Aug 26 | Revenue $3.6bn +17%; Nights and Seats Booked +10% (metric renamed from Nights and Experiences Booked); GBV $27.2bn +16%. 52-week high $187.30 on 21 Aug |
| BKNG | Q2 CY26 | 4 Aug 26 | Revenue $7.4bn +8% (~+7% cc); room nights 325m +5%; gross bookings $51.0bn +9%. 25-for-1 split completed 2 Apr 26 — per-share figures retroactively adjusted |
| UBER | Q2 CY26 | 5 Aug 26 | Revenue $14,191m +12% (+11% cc) against gross bookings $58,022m +24%; trips 3,867m +18%; adj. EBITDA $2.82bn +33%. The revenue/bookings divergence is explained in the release: reported revenue growth is depressed roughly 8pp by business-model changes — not an accounting anomaly |
| SHOP | Q2 CY26 | 5 Aug 26 | Revenue $3,583m +34% (+33% cc); GMV $115,567m +31.6% vs $87,837m; FCF $654m at 18% margin, up from 16% |
| WMT | FQ2 FY27 | 20 Aug 26 | Revenue $187.9bn +5.9% (+5.1% cc); Walmart US comps +2.6% ex-fuel, transaction-led; global eCommerce +23%; global advertising +38%. GAAP operating income +28.8% includes non-recurring tariff-refund benefit |
| Hardware & devices | |||
| AAPL | FQ3 FY26 | 30 Jul 26 | Revenue $109.4bn +16%; iPhone $54.3bn +22%; Services $30.7bn +12% (June-quarter record); GM 50.1% including ~2pp of tariff-refund benefit; dil. EPS $2.02 +29%. Product event 9 Sep, 10:00 PT |
| DELL | FQ1 FY27 | 28 May 26 | Revenue $43.8bn +88%; ISG $29.0bn +181%; AI-optimised server revenue $16.1bn +757%; AI orders booked $24.4bn; AI backlog $51.3bn; FY27 AI-server guide ~$60bn. FQ2 due 1 Sep |
| CSCO | FQ4 FY26 | 12 Aug 26 | Revenue $17.3bn +18%; product orders +35% (+25% ex-hyperscalers); AI infrastructure orders $4bn in the quarter and $9.3bn for FY26, ~4.5x FY25; RPO $46.7bn +7% |
On the calendar
- Tue, 1 Sep · 17:30 BRTDell (DELL) FQ2 2027, guided to $44.0–45.0bn. AI server backlog against $51.3bn and AI-server revenue against the ~$60bn FY27 guide — but the line to watch is gross margin against memory cost. Dell has the least pricing power of anyone in coverage to absorb it.
- Tue, 1 Sep · 17:30 BRTPalo Alto Networks (PANW) FQ4 and FY2026. Reads across to CrowdStrike’s net-new-ARR share claim.
- Tue, 1 Sep · 18:00 BRTMongoDB (MDB) FQ2 2027. Three covered names collide on this date.
- Wed, 2 Sep · 18:00 BRTBroadcom (AVGO) FQ3 2026 (14:00 PT). The direct answer to Marvell’s margin question: $16.0bn AI-semis guide, and whether “>$100bn in FY2027” is refreshed.
- Wed, 2 Sep · 18:00 BRTSnowflake (SNOW) FQ2 2027 (14:00 PT), quarter ended 31 July. Consumption-model counterpoint to Workday’s backlog deceleration.
- Tue, 1 SepJohn Ternus becomes Apple CEO. Worth marking on its own — the 9 September keynote will be the first under a new chief executive.
- Wed, 9 Sep · 14:00 BRTApple “Surprise and shine.” (10:00 PT, Steve Jobs Theater) — iPhone 18 Pro line and a foldable, on 2nm silicon.
- ~Thu, 10 Sep · 02:30 BRTTSMC August monthly revenue (13:30 Taipei). Highest-frequency read on AI demand anywhere in coverage; July was NT$467.58bn, +44.7% y/y.
- ~Thu, 10 SepAdobe (ADBE) FQ3 2026, after the close — third-party listing only. Adobe has issued no press release and no 8-K setting the date, and does not historically publish one. Plausible, not confirmable.
- Sep, unannouncedOracle (ORCL) FQ1 2027 — Oracle has published no scheduling release; its investor-news feed has nothing since 10 June. Discard the widely-quoted “12 September” — it is a Saturday. Last year’s analogue was Tuesday 9 September; trackers point to Monday 14 September. Oracle’s “sets the date” release usually lands about a week ahead, so expect it in the first days of September. The single most important print for the financing thread.
- Tue–Thu, 15–17 SepDreamforce 2026 (Salesforce, Moscone Center). Where the Agentforce ARR definition change gets either substantiated or quietly widened again.
- Wed–Thu, 23–24 SepMeta Connect 2026 — Zuckerberg keynote Wed 23 Sep, 16:00 PT (20:00 BRT). The one venue where Meta explains what $130–145bn of capex is buying.
- Tue, 8 Sep · 14:50 BRTVertiv (VRT) CEO and CFO fireside at Goldman Sachs Communacopia (10:50 PT). Worth listening to only because Vertiv has stopped disclosing orders and backlog in both its release and its 10-Q — this is the next chance anyone gets to ask.
- Tue, 6 OctMarvell (MRVL) Investor Day. Where the long-term custom-silicon margin structure gets defended — the natural publication date for the piece below.
- Wed, 30 Sep · 17:30 BRTMicron (MU) FQ4 2026 (14:30 MT). The direct test of the memory-pricing thesis Nvidia has now backed with $279bn of commitments.
- Wed, 14 OctASML Q3 2026 — just outside the six-week window, but the first read on whether the €43–45bn full-year raise holds.
Story worth owning
Nvidia has become the working-capital clearing house for the entire AI build-out — and it disclosed it in four separate paragraphs nobody read together. Wednesday’s CFO commentary contains, in sequence: supply and capacity commitments rising from $119bn to $279bn “primarily related to the procurement of memory”; days sales outstanding stretching from 45 to 60 “due to extended payment terms on large, multi-quarter agreements with certain investment-grade customers”; $36bn of AI cloud agreements plus $20bn of data-centre leases Nvidia expects to reassign to third parties; and $108.5bn of maximum gross guarantee exposure, of which $105bn was signed in August 2026 against OpenAI’s 20-year leases at SB Energy’s PORTS-Pike campus in Ohio.
Put in one sentence: Nvidia is prepaying its suppliers, financing its customers, and guaranteeing its customers’ landlords. Every one of those is a use of Nvidia’s balance sheet to make someone else’s transaction possible. Nvidia gave the reason itself, in unusually plain language for a filing — AI clouds and model makers “are growing faster than their balance sheets and long-term credit profiles can support.”
The market spent Thursday pricing the 70% growth guide and adding $442bn of market value in a single session. It did not price this. And the two are the same object: a company can guide a year ahead precisely because it has underwritten enough of the demand chain that the orders behave like contracts. Note the FY2028 figure is CFO Colette Kress’s, given on the analyst call against a sell-side consensus near 44% — several outlets have attributed it to Huang, who separately framed it as supply-constrained rather than demand-constrained. The distinction matters if you are quoting it.
Why it is unowned. The circular-financing debate has been running since the OpenAI and SK Group deals, and it has been conducted almost entirely at the level of equity stakes — Nvidia invests in a customer, the customer buys Nvidia chips. That framing is exhausted and both sides have made their case. What is new, and what the August quarter discloses for the first time at scale, is that the support has moved from the equity line to the working-capital and contingent-liability lines, where it is far less visible and far less discussed. Nobody has to mark a guarantee to market. A receivable at 60 days rather than 45 is a footnote. A commitment to buy memory is not a liability at all until it is.
And the buyers are in the same position. This morning’s verification pass turned up three disclosures that belong in the same table, none of them new but none of them read together: Amazon’s trailing-twelve-month free cash flow is an outflow of $7.6bn, on a $66.1bn year-on-year increase in property and equipment purchases. Meta generated $784m of free cash flow in the June quarter against $31.08bn of capex. Oracle’s fiscal 2026 free cash flow was negative $23.7bn, and its 10-K discloses a further $260bn of lease commitments not on the balance sheet, commencing between the current quarter and fiscal 2029. The vendor is carrying working capital; the customers have stopped generating cash. Those are two halves of one mechanism.
The deliverable. A single reconciliation for Nvidia, and then the same schema applied across the covered chain: for each name, the cash it has committed forward (purchase commitments, prepayments), the credit it has extended (receivable days, customer financing, vendor equity), and the contingent exposure it has accepted (guarantees, lease reassignments, residual support). Three columns, one row per name — Nvidia, Oracle with ~$75bn of customer prepayments in RPO and $260bn of off-balance-sheet leases, CoreWeave with a five-year facility against three-year contracts, Marvell with a warrant over 58,970,907 shares at $206.58 — roughly 6.5–7% of the company, vesting one tranche per $500m of Google’s own Custom Products purchases — Dell carrying a $51.3bn AI backlog it must fund into. The question the table answers is the one no published model asks: how much of the AI build-out’s growth is being financed by the seller rather than the buyer, and where does that obligation sit when it is not debt?
CoreValue can build this from work already on the shelf — the AI Bubble Radar tripwires, the CoreWeave debt schedule, the Data Centre Atlas capacity data, and master files for nearly every name in the grid. The honest conclusion may well be that none of it is distress; investment-grade balance sheets can carry a great deal. But “the vendor is carrying the chain’s working capital” is a structural description that is true today, sourced entirely to primary filings, and not in print anywhere.
Sources
- Marvell Technology — Form 8-K, Exhibit 99.1, FQ2 2027 results — SEC EDGAR, 27 August 2026. (Revenue, margins, EPS, FQ3 guidance, CEO quote, Investor Day date.)
- Marvell Technology, Inc. Reports Second Quarter of Fiscal Year 2027 Financial Results — Marvell Investor Relations, 27 August 2026.
- Marvell stock tumbles 8% as outlook underwhelms despite Google deal — CNBC, 28 August 2026.
- Marvell Q2 FY2027 earnings: record revenue, raised guidance — Quartz, 27 August 2026. (FY2028 outlook raise, data-centre share of revenue.)
- Marvell Technology — Form 8-K, Google warrant — SEC EDGAR. (58,970,907 shares at $206.58, vesting one tranche per $500m of Custom Products revenue. The commonly quoted “7%” is a media calculation; the filing states no percentage.)
- NVDA daily price history — Stock Analysis. (Close of $227.98, +8.74%, Thursday 27 August 2026. The +7.4% figure circulating in some recaps is an intraday/early print, not the close.)
- NVIDIA — Form 8-K, Exhibit 99.1, Q2 FY2027 results — SEC EDGAR, 26 August 2026.
- Oracle Announces Record Q4 and FY2026 Results — Oracle Investor Relations, 10 June 2026. (RPO $638bn, +363%.)
- NVIDIA — Form 8-K, Exhibit 99.2, CFO Commentary on Q2 FY2027 — SEC EDGAR, 26 August 2026. (Commitments $119bn→$279bn, $366bn total, $108.5bn guarantees, SB Energy PORTS-Pike, DSO 45→60, inventory, $25bn notes, FQ3 outlook.)
- NVIDIA Announces Financial Results for Second Quarter Fiscal 2027 — NVIDIA Newsroom, 26 August 2026.
- Nvidia CFO says about half of its data center business comes from customers beyond hyperscalers — Fortune, 27 August 2026.
- Stock Market Today, Aug. 27: Nvidia Surges on Blowout Results and Surprising Guidance — The Motley Fool, 27 August 2026. (SMH +3.10%, AVGO +4.49%, AMD −0.89%. This source’s NVDA figure of +7.4% is superseded by the close of +8.74% cited above.)
- S&P 500 Today, August 27, 2026: NVIDIA Adds $435bn as the Average Stock Falls — Eastern Herald, 28 August 2026. (Index closes and the negative-breadth observation; this source puts NVDA’s move nearer +8.7%, hence the range stated in the text.)
- Workday Announces Fiscal 2027 Second Quarter Financial Results — Workday Investor Relations, 27 August 2026.
- Workday Announces Fiscal 2027 Second Quarter Financial Results — PR Newswire, 27 August 2026. (Backlog, margins, FY27 guidance, AI share of new ACV.)
- Salesforce — Form 8-K, Exhibit 99.1, FQ2 2027 results — SEC EDGAR, 26 August 2026. (cRPO +14% against revenue +11%, used in the Workday comparison.)
- US Stock Market Today, 28 August 2026 — Nasdaq futures slip as Marvell sinks pre-market — TS2, 28 August 2026. (Pre-market futures levels at ~04:30 EDT.)
- Asia Stock Market Today, 28 August 2026 — TS2, 28 August 2026. (KOSPI open, Samsung and SK hynix, Bank of Korea decision, Nikkei level.)
- Samsung and SK hynix’s HBM4 push puts HBM and general memory pricing in the spotlight for 2H earnings — TrendForce, 13 August 2026.
- HBM pricing and market share, 2026 — Silicon Analysts. (HBM4 vs HBM3e $/GB indications; third-party estimate, not a company disclosure.)
- TSMC July 2026 Revenue Report — TSMC Investor Relations, 10 August 2026.
- CoreWeave Reports Second Quarter 2026 Results — CoreWeave Investor Relations, 11 August 2026.
- CoreWeave Closes $2.6 Billion Loan Facility — CoreWeave Investor Relations, 10 August 2026.
- NVIDIA Partners With Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to Mobilize Over $500 Billion of Third-Party Capital — NVIDIA Newsroom, 10 August 2026.
- Oracle announces Equity and Debt Financing Plan for Calendar Year 2026 — Oracle Investor Relations, 1 February 2026. ($45–50bn gross, roughly half debt and half equity.)
- Oracle Corporation — Form 10-K, fiscal year ended 31 May 2026 — SEC EDGAR, 22 June 2026. (Capex $55,663m, free cash flow −$23.7bn, $260bn of lease commitments not yet commenced, $19bn of unconditional cloud-infrastructure purchase commitments.)
- Amazon.com — Form 10-Q, quarter ended 30 June 2026 — SEC EDGAR, 31 July 2026. (TTM free cash flow outflow of $7.6bn; cash-capex definition; Q2 capex $53.1bn.)
- Meta Platforms Q2 2026 earnings call transcript — Meta Investor Relations, 29 July 2026. (CFO Susan Li on the $130–145bn capex range being narrowed rather than raised, and the explicit refusal to guide 2027.)
- Alphabet Q2 2026 earnings call transcript — Alphabet Investor Relations, 22 July 2026. (CFO Anat Ashkenazi raising CY2026 capex to $195–205bn from $180–190bn.)
- Microsoft FY2026 Q4 earnings materials and call — Microsoft Investor Relations, 29 July 2026. (CFO Amy Hood on the useful-life change moving calendar-2026 capex from ~$190bn to ~$175bn with underlying expectations “otherwise unchanged”; no FY2027 dollar guidance given.)
- Broadcom Inc. to Announce Third Quarter Fiscal Year 2026 Financial Results on Wednesday, September 2, 2026 — Broadcom Investor Relations.
- Micron Technology to Report Fiscal Fourth Quarter Results on September 30, 2026 — GlobeNewswire, 26 August 2026.
- Dell Technologies reports fiscal Q2 2027 results on September 1 — The Motley Fool, 27 August 2026. (Date only; Dell’s FQ1 figures are sourced to its own 28 May release.)
- Apple officially announces its 9 September event — 9to5Mac, 26 August 2026.
Figures are as reported by companies, their investor-relations releases, SEC filings and the news sources listed above, gathered on the morning of 28 August 2026 and cross-checked in a separate verification pass, which produced corrections to five figures before publication. They have not been independently re-derived from audited financial statements. Where sources disagree, the disagreement is stated in the text rather than resolved silently — today’s live example is the attribution of Nvidia’s ~70% fiscal-2028 growth forecast, which several outlets place with Jensen Huang and which the weight of evidence places with CFO Colette Kress. Prepared for internal research use at CoreValue Research. This is not investment advice, not a recommendation to buy or sell any security, and no representation is made as to accuracy or completeness.
