It can be hard to wrap one’s head around how big the AI revolution really is, but there are some interesting data points that keep popping up.
Take NVIDIA’s fiscal second quarter, reported this week: the company pulled in $96.2 billion in revenue for the three months ended July 26, 2026. Spread that across the 91 days in the quarter and NVIDIA generated roughly $1.06 billion in revenue every single day, weekends included.
That number beat what Wall Street was expecting by a wide margin. Analysts had modeled $91.9 billion in revenue for the quarter, and NVIDIA cleared that by more than $4 billion. Adjusted earnings per share came in at $2.22, ahead of the $2.09 consensus, while GAAP EPS landed at $2.46. Revenue was up 106% from the same quarter a year ago and up 18% sequentially from Q1, at a point in the AI buildout where most companies of this size would be expected to be slowing down rather than compounding.
Data Center, the segment that has effectively become NVIDIA’s entire business at this point, brought in $89 billion, ahead of the $86 billion estimate and now accounting for over 92% of total revenue. On its own, that segment is generating close to $978 million a day. Gaming, professional visualization, and automotive combined make up what’s left, and increasingly look like rounding errors next to the AI infrastructure business.
Gross margin held at 75% on both a GAAP and non-GAAP basis, matching what analysts had penciled in, which is notable given how much has been written about supply constraints and rising component costs across the industry. NVIDIA is still pricing its chips at a level that keeps margins fat even while shipping at a scale few hardware companies have ever managed.
The guidance for the current quarter is where things get more interesting. NVIDIA told investors to expect $108 billion in revenue, comfortably above the $104 billion consensus, which would put the daily run rate above $1.17 billion. Gross margin guidance did dip slightly to 74%, a touch below the 75% Wall Street wanted, and the company explicitly said it isn’t assuming any Data Center compute revenue from China in that outlook, given how export restrictions have played out over the past year.
Jensen Huang’s comment on the earnings call was more declarative than the usual founder-speak that tends to accompany these releases. “AI has reached its inflection point,” he said, framing the shift as one where AI systems are now producing output that companies pay for directly rather than just promising to someday. Compute, in his telling, has moved onto the revenue side of the ledger for NVIDIA’s customers, and that money keeps flowing back to NVIDIA.
Part of what’s propping up these numbers is the sheer scale of capex commitments coming from the hyperscalers. Amazon, Microsoft, Alphabet, and Meta have all been raising their infrastructure budgets through the year, and combined spending from the top four is on pace to touch $715 billion in 2026. That money doesn’t sit in a vault. Most of it eventually shows up as a purchase order for NVIDIA silicon.
The demand picture also extends well past the traditional cloud giants. NVIDIA’s own commentary pointed to sovereign AI programs, so-called NeoClouds, and enterprise buyers as increasingly important sources of revenue, a shift away from the hyperscaler-dominated mix of a year ago. The OpenAI compute arrangement at the Ohio site, structured around gigawatts of capacity rather than a fixed dollar figure, is one example of how these deals are starting to lock in demand years out instead of quarter to quarter.
None of this comes without risk on the supply side. Memory has become the binding constraint in a lot of these buildouts, with HBM allocation now one of the tightest links in the chain. SK Hynix and Micron have both ridden that shortage to trillion-dollar-plus valuations this year, which says something about how much of NVIDIA’s growth actually depends on partners further down the stack keeping pace.
For now, the math is straightforward even if the scale of it isn’t intuitive. A company built on rendering video games two decades ago now moves more money in a single day than most public companies generate in a full year. Whether that run rate holds through the next few quarters will depend less on demand, which shows no sign of slowing, and more on how much of it NVIDIA and its suppliers can actually fulfill.