Nvidia reported revenue of $96.2 billion for its fiscal second quarter ended July 26, 2026, up 18% sequentially and 106% year over year, comfortably clearing the company's own prior guidance of roughly $91 billion, according to its August 26 earnings release and an accompanying SEC filing. Data Center revenue, the segment most tied to AI infrastructure buildout, reached $89.0 billion, up 117% from a year earlier.

What drove the beat

Nvidia attributed the jump primarily to the ramp of Blackwell Ultra infrastructure, with hyperscale customer revenue more than doubling year over year and rising 13% sequentially. GAAP and non-GAAP gross margins both came in at 75.0%, and diluted earnings per share were $2.46 GAAP and $2.22 non-GAAP. The company returned about $26.0 billion to shareholders during the quarter through buybacks and dividends, and had roughly $99.0 billion remaining on its repurchase authorization as of quarter-end.

For the current quarter, Nvidia guided to revenue of $108.0 billion, plus or minus 2%, a figure that assumes no Data Center compute revenue from China — a market where U.S. export restrictions have continued to constrain Nvidia's ability to sell its most advanced accelerators. Analysts had entered the print split on how much uncertainty Blackwell-to-Rubin transition timing would introduce, with consensus Data Center estimates having drifted lower over the prior months even as CEO Jensen Huang described demand as "parabolic" heading into the report.

Why it matters

The results are one of the clearest available signals of how much capital hyperscalers and AI labs are still committing to compute, at a moment when compute-supply deals — including Anthropic's $45 billion agreement with Nscale, disclosed the same week — have made infrastructure spending a central story across the AI industry. A 117% year-over-year jump in Data Center revenue, driven by early Blackwell Ultra shipments rather than the still-unreleased Vera Rubin platform, suggests the current generation of AI accelerators is still under-supplied relative to demand. The decision to guide with zero assumed China compute revenue also underscores that geopolitics, not just technology, is now a standing variable in Nvidia's own forecasts — and, by extension, in the cost base of every AI lab and agent company that depends on Nvidia silicon.