Nvidia's $96.2 Billion Quarter: Vera Rubin Hits Full Production as AI Compute Becomes Revenue

Nvidia just reported the largest quarter in its history — and by a wide margin, one of the largest in the history of the semiconductor industry. On August 26, 2026, the company posted revenue of $96.2 billion for its second quarter of fiscal 2027 (ended July 26, 2026), up 18% from the previous quarter and up 106% year over year. Wall Street had expected roughly $92.2 billion. The beat was so decisive that Nvidia’s own prior guidance — $91.0 billion, plus or minus 2% — now looks conservative in hindsight.

The headline numbers are straightforward: GAAP and non-GAAP gross margins both landed at 75.0%, GAAP operating income hit $63.7 billion (up 124% Y/Y), GAAP net income reached $59.7 billion, and non-GAAP diluted EPS came in at $2.22 versus the $2.09–$2.10 consensus. It was the company’s 15th consecutive quarter of beating profit projections, according to the Wall Street Journal’s live coverage. But the more interesting story is what the numbers reveal about where the AI infrastructure buildout actually stands midway through 2026.

“Compute is revenue”

The most quoted line from Jensen Huang’s earnings commentary was also the most consequential: “AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.”

That framing matters because it answers the question that has hung over the AI trade since early 2025: is all this GPU spending generating returns, or is it a speculative capital expenditure loop waiting to crack? Huang’s argument is that the customer base has fundamentally broadened. “This time last year, one lab alone was driving the buildout; today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online — with strong momentum across the U.S. and around the world,” he said.

The “one lab” is widely understood to be OpenAI, whose record procurement commitments previously dominated Nvidia’s order book. The diversification claim is now visible in the quarter’s partnership announcements: sovereign AI programs with Korea (SK Telecom, NAVER, Brookfield building gigawatt-scale infrastructure), Japan’s national AI infrastructure program, and a sprawling new financing architecture — partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR designed to mobilize over $500 billion of third-party capital for AI infrastructure over time.

Data center: $89 billion and accelerating

Data center revenue was $89.0 billion, up 18% sequentially and up 117% year over year. Kiplinger’s live coverage noted data center growth of 116.6% Y/Y against an ACIE (AI cloud) segment that more than doubled — evidence that hyperscalers and neoclouds alike are still expanding capacity aggressively rather than digesting prior purchases.

The Vera Rubin platform is the engine of the next leg. Nvidia confirmed Rubin is “ramping into full production” with racks already running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius. Spectrum-6 switch systems supporting both pluggable and co-packaged optics are arriving at what the company calls “gigascale AI factories.” Huang earlier this year estimated that Blackwell and Vera Rubin platforms would combine for $1 trillion in sales between 2025 and the end of the decade — a figure that, at current growth rates, no longer looks like CEO bravado.

The quarter also introduced Nvidia Vera, described as the first CPU built specifically for AI agents, plus Groq 3 LPX (the interactive inference accelerator, now in full production), Vera BlueField-4 STX with in-silicon security for agentic storage, and the DSX platform — essentially a standardized playbook for designing and operating AI factories at scale.

Guidance: $108 billion next quarter

For Q3 FY2027, Nvidia guided to $108.0 billion in revenue, plus or minus 2% — another double-digit sequential step up, and notably a forecast that assumes zero data center compute revenue from China. Gross margin is expected to dip slightly to 74.0% (plus or minus 50 basis points), which management tied to the Rubin ramp; the Blackwell-to-Rubin transition is tracking much like the Hopper-to-Blackwell crossover did, with early-production costs pressuring margins before yields mature.

Edge computing contributed $7.2 billion (up 27% Y/Y), bolstered by the RTX Spark superchip partnership with Microsoft for Windows PCs and DGX Station for Windows deskside systems. On the physical AI side, Nvidia released Cosmos 3 (an open frontier omnimodel), the Isaac GR00T reference humanoid robot design, Alpamayo 2 Super for robotaxi development, and Halos for Robotics — a full-stack safety system for embodied AI.

Shareholders received about $26.0 billion in buybacks and dividends during the quarter, with roughly $99.0 billion remaining under the repurchase authorization. The next quarterly dividend of $0.25 per share goes out October 1, 2026.

The bear case, briefly

Not everything points up. Gross margin guidance of 74.0% implies the cost of the Rubin ramp, and Yahoo Finance noted the stock had slipped after four straight earnings beats — a sign that expectations embedded in the share price now require flawless execution. The China exclusion removes a one-time catalyst some investors hoped for, and the $500 billion financing partnerships are “subject to definitive agreements,” meaning the capital mobilization is a framework, not a signed check. If AI application revenue fails to keep pace with token-generation costs somewhere in the ecosystem, the financing structure could amplify a downturn as quickly as it accelerates the buildout.

Still, on the evidence of this quarter, the demand signal Nvidia sees remains unambiguous. As Huang put it: “The AI infrastructure buildout is at full steam. Vera Rubin, now in full production, was built to power exactly this moment.”