Nvidia is guided to cross $100 billion in quarterly revenue for the first time next quarter, with a $108 billion Q3 forecast that puts its annualized run rate at roughly $432 billion. By revenue that would place it sixth among all public companies anywhere, ahead of Apple, McKesson and Alphabet, trailing only Amazon, Walmart, State Grid, UnitedHealth and Saudi Aramco. Those five companies grow in the low single digits or shrink; Nvidia is still expanding at a triple-digit clip.
The headline number obscures a structural change inside the business, one that venture capitalist and analyst Tomasz Tunguz flagged in his breakdown of the earnings call. Hyperscale revenue, the cloud giants buying chips to run their own data centers, grew only 13 percent sequentially last quarter, according to figures Nvidia’s chief financial officer Colette Kress gave on the call. Everything else, the category Nvidia calls ACIE and that covers AI-native startups, enterprises and sovereign buyers, grew 25 percent sequentially and 138 percent year over year. For the first time, that non-hyperscaler group supplied the majority of new data center revenue.
Jensen Huang framed the shift as evidence of a maturing market rather than a slowdown, telling investors that a single lab drove last year’s buildout while this year features multiple frontier labs, a “golden age” of new AI startups and a growing open-model ecosystem. The framing is company talk, but the revenue mix backs it up: demand is no longer concentrated in four or five familiar names.
That diversification carries a cost Nvidia is absorbing directly. Days sales outstanding, the average time it takes Nvidia to collect payment, jumped from 45 to 60 days in a single quarter after sitting in a tight 43 to 46 day band for the previous eight. Receivables grew 64 percent while revenue grew 18 percent sequentially, a gap Kress put down to longer settlement windows written into big contracts that run across several quarters, signed with buyers the company describes as investment grade.
The mechanism matters more than the label. Startups and neoclouds, the smaller cloud providers renting out GPU capacity, generally carry weaker balance sheets and thinner cash reserves than Amazon, Microsoft or Google. To keep them buying at the pace Nvidia’s guidance assumes, the company is effectively extending credit: financing terms, supply commitments, power guarantees, leases and equity stakes in the same customers placing the orders. That is a materially different risk profile than selling to hyperscalers who fund purchases from their own balance sheets.
Nvidia’s own margins mask how thin that support system is stretched underneath. At 75 percent gross margin, a $108 billion quarter produces roughly $324 billion in annual gross profit, a figure that would rank second globally. Strong margins buy patience, but they do not eliminate counterparty risk if a financed customer’s revenue does not materialize on schedule.
Nvidia’s own commentary does not disclose how much of the $108 billion guide depends on newly financed customers rather than self-funded ones, which is the more useful number for judging durability. The metric worth tracking next quarter is whether DSO holds near 60 days or keeps climbing. A plateau would suggest the extended terms were a one-time adjustment to close large deals. A continued rise would mean Nvidia is underwriting more of its own demand curve, and that the industry’s custom silicon push, Google’s TPUs, Amazon’s Trainium and Meta’s MTIA chips among them, is starting to bite into the hyperscaler revenue that once anchored the company’s growth. Anyone modeling Nvidia’s 2027 guidance should treat the DSO print as a leading indicator, not a footnote.
Analysis by Tomasz Tunguz, published 30 July 2026 on his blog tomtunguz.com.
