The numbers come first.
Nvidia reported second-quarter revenue of $96.2 billion, up 106% from the year-ago period and 18% over last quarter, the company announced Wednesday. Analysts had penciled in $92.2 billion. The beat was not marginal — it was structural.
Data center revenue, the engine of Nvidia's AI business, reached $89 billion, clearing the consensus estimate of $85.7 billion. Within that segment, the company broke out $48.7 billion in hyperscale revenue and $40.3 billion from AI clouds, industrial, and enterprise customers — a new reporting breakdown designed to distinguish large cloud buyers from AI-native clouds, sovereign AI deployments, and on-premises enterprise.
On the bottom line, Nvidia posted non-GAAP earnings of $2.22 per diluted share, ahead of analyst expectations of $2.06 to $2.09. One quarter ago, the non-GAAP figure stood at $1.87 per share. The company has begun including stock-based compensation in its non-GAAP results, which limits direct year-over-year comparisons.
For the current quarter, Nvidia guided to revenue of $91 billion, plus or minus 2%. The average analyst expectation heading into the print was $103.9 billion.
'AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue,' said founder and CEO Jensen Huang in a statement. 'And demand is accelerating.'
In after-hours trading, NVDA was flat following a 1.6% decline during the regular session.
---
The market has already voted — and the verdict is that free enterprise, not government mandate, built the infrastructure layer of the AI economy. Nvidia's $96 billion quarter is the clearest evidence yet that when capital is allowed to chase genuine demand, the returns are historic. No industrial policy produced this. No subsidy committee approved Jensen Huang's roadmap. Customers — hyperscalers, sovereign governments, and enterprise buyers alike — are writing checks because the chips deliver measurable productivity.
For investors and executives watching the AI buildout, the message is straightforward: compute is now a revenue line, not a cost center. The companies that treat it as such, and the policymakers who resist the urge to regulate the stack before it matures, stand to benefit most from what Huang is calling an accelerating demand curve. The ones who hesitate will be reading about it in someone else's earnings release.
