The headline number
Nvidia's fiscal second quarter, which ended July 26, 2026, brought in $96.2 billion in revenue, up 106% from a year earlier and 18% from the prior quarter. That beat Wall Street's roughly $92.2 billion consensus estimate. Data center sales, the part of the business built on AI (artificial intelligence) chips, accounted for $89.0 billion of that total, up 117% year over year. The rest of the business, gaming and other non-data-center chips, added $7.2 billion, up 27%.
Profitability kept pace with the top line: GAAP (Generally Accepted Accounting Principles) net income was $59.69 billion, GAAP earnings per share came to $2.46, and non-GAAP earnings per share hit $2.22 against a consensus of roughly $2.09 to $2.10. Gross margin held at 75.0% on both a GAAP and non-GAAP basis, even as operating expenses climbed 55% year over year to $8.41 billion.
Guidance comes with a China asterisk
Nvidia guided next quarter's revenue to $108.0 billion, plus or minus 2%, with gross margin expected around 74.0%. Chief financial officer Colette Kress said margins are likely to keep compressing, down to a 71-72% range by the January quarter, as memory chip costs rise. Notably, the $108 billion guidance assumes zero revenue from China data-center compute, treating any future access to that market as upside rather than something already baked into the forecast.
Looking further out, Kress guided full fiscal 2028 revenue growth of roughly 70%. Chief executive Jensen Huang pushed back on the idea that this reflects softening demand, saying actual demand is running well above that pace and that supply, not orders, is the constraint. By his own estimate, Nvidia is currently able to fill only about 70% of what customers want to buy.
The data center number is doing most of the work
Splitting the total further shows how lopsided Nvidia's growth has become. Data center sales alone, at $89.0 billion, are now close to the company's entire revenue from a year ago, and they grew almost twice as fast as the rest of the business combined. Gaming and other non-data-center chips are still growing at a healthy 27% clip, but at $7.2 billion they are a rounding error next to the AI (artificial intelligence) infrastructure build-out. That imbalance is the clearest evidence yet that Nvidia has become, in practice, a data center company that also happens to sell graphics cards, rather than the other way around.
New hardware moving into production
The quarter also marked the full production ramp of Nvidia's Vera Rubin computing platform, with racks already live at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure (OCI), and Nebius. Alongside it, Nvidia introduced Vera, which it describes as the first CPU (central processing unit) purpose-built for AI agents, paired with new Spectrum-6 networking switch systems. Getting a new platform from announcement to live racks at five major cloud providers in the same quarter it was introduced is a faster turnaround than Nvidia's own prior generations, and it is the clearest sign that the current pace of AI infrastructure spending shows no sign of slowing from the buyer's side, whatever the supply picture looks like from Nvidia's.
Money back to shareholders, and a question from investors
Nvidia returned $26.0 billion to shareholders through buybacks and dividends during the quarter, leaving $99.0 billion of its buyback authorization still unused. The next quarterly dividend, $0.25 per share, is payable October 1, 2026.
On the earnings call, Huang directly addressed investor concern over so-called circular financing, the worry that a cluster of AI infrastructure deals are effectively financed by the same small set of companies investing in each other's capacity, which could make demand look larger than it really is.
Markets responded favorably regardless: Nvidia shares rose more than 8% in the session following the report, adding over $400 billion in market value.
What it means in practice
For anyone building on Nvidia's hardware, the throughline is that the constraint right now is supply, not appetite. Huang's own estimate, that the company can meet only about 70% of demand, suggests lead times and allocation are likely to remain a bigger planning problem than price for cloud providers and AI labs over the next several quarters. Teams planning capacity around Vera Rubin or successor platforms should treat availability, not budget, as the binding constraint for the foreseeable future.
The zero-China assumption in the guidance is also worth watching: any easing of export restrictions would land as upside to an already aggressive forecast, while continued restriction simply keeps the current picture unchanged. That makes the $108 billion guidance a conservative floor in one specific sense, even though it is already well above what the company reported this quarter.
For investors, the circular-financing question Huang addressed on the call is likely to keep resurfacing as more of these infrastructure deals are announced, and it is a fair lens to apply to headline growth figures across the sector, not just Nvidia's. A market that added $400 billion in value on a single earnings beat is also a market with very little room for the next report to disappoint, and the gap between guided 70% growth and Huang's description of even stronger underlying demand leaves Nvidia less room than usual to absorb a supply hiccup without moving the stock.
