NVIDIA just posted one of those reports that doesn’t feel like “earnings season content” anymore and starts to read like infrastructure telemetry. For the fourth quarter ended January 25, 2026, revenue hit a record $68.1 billion, up 20% quarter over quarter and up 73% year over year, and the full fiscal 2026 number landed at $215.9 billion, up 65% from the prior year. The profitability profile stayed strikingly high in the quarter, too, with GAAP gross margin at 75.0% (non-GAAP 75.2%), and EPS coming in at $1.76 GAAP and $1.62 non-GAAP. On the full-year view, gross margin stepped down versus FY2025 (71.1% GAAP, 71.3% non-GAAP), which is worth noticing even if it doesn’t change the bigger picture; when you’re operating at this altitude, mix shifts, ramp costs, and supply chain realities show up as “just a few points,” but those few points are measured on a mountain of revenue.
If you want the cleanest single line that explains why the market keeps treating these prints like a macro event, it’s this: Data Center revenue in Q4 was $62.3 billion, up 22% from the prior quarter and up 75% year over year, and the full-year Data Center number rose 68% to $193.7 billion. That’s the AI boom showing up as an accounting fact. And NVIDIA’s messaging is basically saying the boom is evolving, not fading: Jensen Huang framed this as an “agentic AI inflection point,” called Grace Blackwell with NVLink “the king of inference” on cost per token, and positioned Vera Rubin as the next extension of that lead. The interesting part isn’t the poetry of it (CEOs are gonna CEO), it’s how explicitly the company is shifting the narrative from “training the biggest models” toward “industrial-scale inference” and “enterprise agents,” because that’s the phase where spend gets sticky and budgets turn into multi-year commitments rather than experiments that can be paused.
Guidance did a lot of work here. For Q1 fiscal 2027, NVIDIA guided revenue to $78.0 billion plus or minus 2%, and they made a point of stating they are not assuming any Data Center compute revenue from China in that outlook. Read that twice: they’re effectively saying the growth engine is strong enough that they’re willing to guide without that piece in the model. Margins are expected to remain around the same rarefied zone (74.9% GAAP, 75.0% non-GAAP, plus or minus 50 bps), and operating expenses are guided to roughly $7.7 billion GAAP and $7.5 billion non-GAAP. Also, a notable accounting/communication shift: beginning in Q1 FY2027, NVIDIA will include stock-based compensation expense in its non-GAAP results, which changes how “clean” non-GAAP looks going forward. Honestly, that’s probably a healthy move for comparability and trust, even if it makes headline non-GAAP numbers a touch less flattering.
Capital return stayed aggressive, almost casually so. During fiscal 2026, NVIDIA returned $41.1 billion to shareholders via repurchases and dividends, and it still had $58.5 billion remaining under the repurchase authorization at quarter-end. The next quarterly dividend is $0.01 per share, payable April 1, 2026, to shareholders of record March 11, 2026. Pair that with the cash and marketable securities position shown on the balance sheet, and you get the picture: this is a company generating enormous cash from operations (and, yes, spending heavily to keep the machine fed), while still buying back stock at scale. Put differently, NVIDIA is behaving less like a “high-growth chip company” and more like a platform business that just happens to ship silicon—and it’s doing it while the entire industry reorganizes around accelerated computing, AI-native networking, and the practical economics of inference. That’s the thread tying the quarter together, and it’s why the guidance mattered as much as the beat.