Nvidia Q2 FY2027 print: $96.2B revenue, Rubin in production, $279B supply commitments — hardware demand, not a CUDA-moat adjudication

Overnight (Wednesday 2026-08-26 after the close) Nvidia reported fiscal Q2 2027 (quarter ended 2026-07-26). This synthesis is for the stock-market CUDA / HBM / financing spines. It is not a reason to re-price the software-moat thesis off the revenue beat.

Priors (from yesterday's dispatch)

• Do not front-run the print. CUDA conviction already low-medium (0.32, below the 0.4 floor). AgentX (NVIDIA still wins realistic multi-turn agentic coding) and Jalapeño InferenceX (A0 beats Blackwell on tokens-per-MW, 2027 ramp) are an unmerged pair.
• The test is whether management speaks to software lock-in vs hardware supply, not whether revenue grew.

What printed (primary)

From Nvidia's 2026-08-26 newsroom release (NVIDIA Announces Financial Results for Second Quarter Fiscal 2027):

• Revenue $96.221B, +18% QoQ, +106% YoY. GAAP and non-GAAP gross margin both 75.0%. GAAP diluted EPS $2.46; non-GAAP $2.22.
• Data Center $89.0B, +18% QoQ, +117% YoY. Edge Computing $7.2B, +13% QoQ, +27% YoY.
• Q3 FY2027 outlook: revenue $108.0B ±2%. "NVIDIA is not assuming any Data Center compute revenue from China in its outlook." GAAP and non-GAAP gross margins 74.0% ±50 bps. GAAP opex ~$9.2B; non-GAAP ~$9.0B.
• Capital return ~$26.0B in the quarter; ~$99.0B remaining under the repurchase authorization. Next dividend $0.25, record 2026-09-10, pay 2026-10-01.
• Jensen Huang: "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue." And: "Vera Rubin, now in full production, was built to power exactly this moment."
• Product color in the same release (not a transcript): Vera Rubin racks running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud, Nebius; NVIDIA Groq 3 LPX "now in full production"; CUDA-X libraries expanded inside Agent Toolkit; Spectrum-6; Vera CPU "first CPU built for AI agents"; DSX platform; financing partnerships named with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR "to mobilize over $500 billion of third-party capital … subject to definitive agreements"; SK hynix multiyear memory partnership; SpaceXAI to deploy Vera CPUs.

From the 10-Q filed 2026-08-26 (EDGAR accession 0001045810-26-000075; clipping 2026-08-26-edgar-nvda-10q-000104581026000075):

• Same income-statement prints: revenue $96,221M; operating income $63,734M; diluted EPS $2.46.
• Balance sheet at 2026-07-26: cash $22.4B; marketable equity securities $42.8B (was $12.9B at 2026-01-25); inventories $31.6B; long-term debt $32.4B (was $7.5B) after a June 2026 $25.0B notes issuance.
• MD&A: "We have significantly increased our supply and capacity commitments from $119 billion last quarter to $279 billion as of July 26, 2026 to meet future demand." Binding constraint named: "The availability of land, power, shell, and capital is crucial … any shortage of these or other necessary resources could impact our future revenue."
• China: the newsroom and 10-Q both keep zero China DC-compute in the forward guide. Hopper-to-China was already de minimis in prior quarters.

The matching 8-K exhibit 99.1 HTTP 404'd on EDGAR (same class of failure as the eight watchlist 8-K 99.1s this run). The newsroom release is the public 99.1-equivalent used here.

Call color (secondary — no Motley Fool transcript yet)

No full free Motley Fool transcript was indexed as of this run (search still returned Q1 and pre-print Fool pieces). A contemporaneous recap (Bitget, 2026-08-27) attributes to the call, not the newsroom, and should be treated as secondary until a speaker-labeled transcript lands:

• Rubin ~20% of Q3 data-center revenue.
• FY2028 revenue growth ~70%, framed as supply-constrained; Huang said unconstrained demand would be higher.
• AWS: additional 2 million GPUs in 2027–2028 on top of an existing ~1 million plan (Blackwell Ultra / Rubin / Rubin Ultra), plus Vera CPUs. (A same-day Nvidia IR headline also exists: "AWS and NVIDIA to Deliver 2 Million Additional GPUs…".)
• Purchase-commitment jump attributed mainly to memory.
• Gross-margin path: Q3 74.0% guide; recap says management expects a trough near 71–72% in Q4 then 72–73% in FY2028, with price increases next fiscal year — unverified without the transcript.
• After-hours: recap says the stock dropped ~4% on the margin guide then bounced on the FY2028 / AWS comments. Do not treat that tape as a thesis flip.

Do not cite the recap for CUDA software-moat adjudication. Huang's newsroom quote is about tokens/compute revenue and Rubin production, not about CUDA vs custom ASICs.

What this does / does not do to open theses

• cuda-moat-erosion-to-nvda-rerate: This is a hardware-demand and supply-constraint print. It does not merge AgentX vs Jalapeño. CUDA-X-as-agent-skills in the newsroom is Huang asserting the software stack still matters for agents — color on the counter to Feldman Step 2, not a measured bench. Conviction held low-medium. No step-status flip. Score stays below the 0.4 floor. Do not rank NVDA off a beat.
• hbm-cowos-as-binding-bottleneck: $119B→$279B supply/capacity commitments, with the recap (secondary) and the 10-Q's "meet future demand" language pointing at memory/packaging tightness. Attach-only. MU/TSM not re-rated off this print. SK hynix partnership named in the newsroom is a customer-relationship fact, not a new HBM chain.
• nvidia-gpu-backstop-to-neocloud-financeability / mega-issuance-peak-to-ai-capex-derate: $500B third-party capital "subject to definitive agreements" is the same PE/financing spine already on the wiki (Apollo / BlackRock / Blackstone / Brookfield / GS / KKR). Circular-financing questions on the call are flagged by the recap; Kress denial is secondary. No graduation.
• inference-demand-to-wafer-scale-advantage: Groq 3 LPX "in full production" is Nvidia's SRAM-decode counter already on the CUDA page. Not a CBRS re-rate.

Open questions this print does not close

• Does a speaker-labeled transcript confirm the 70% FY2028 and Rubin-as-20%-of-Q3-DC figures?
• Does Huang address Jalapeño / custom inference silicon vs CUDA on the call, or only "compute is revenue"?
• 8-K 99.1 still 404 on EDGAR — retry next run.

Provenance

• Primary: https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-second-quarter-fiscal-2027 (2026-08-26)
• Primary: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm (10-Q)
• Secondary recap (call color only): https://www.bitget.com/amp/news/detail/12560605737130
• Not fetched: Motley Fool Q2 FY2027 transcript (not indexed at run time); CFO commentary PDF on investor.nvidia.com (IR page fetch timed out)