Summary: On February 25, NVIDIA reported FY2026 fourth-quarter revenue of US$68.1 billion and net income of US$43 billion. Full-year revenue reached US$215.9 billion, up 65%, so the roughly US$85 billion increase came from a base already above US$130 billion. The scale suggests sustained AI infrastructure spending, but it also concentrates schedule, yield, and policy risk across the broader AI industry.
NVIDIA’s fiscal 2026 fourth-quarter results landed on February 25. The company posted $68.1 billion in quarterly revenue — an all-time high, 20% above the prior quarter — along with $43 billion in quarterly net income.
The full-year view is the bigger headline: FY2026 closed at $215.9 billion in revenue, 65% above the year before. Growth of 65% at this scale works out to roughly $85 billion of additional revenue across twelve months. Run the arithmetic in reverse — divide $215.9B by 1.65 and the prior year lands near $131B — and the scale needs no commentary.
The Quarter in Numbers
- Revenue: $68.1 billion for the quarter, the highest in NVIDIA’s history, up 20% from the prior quarter
- Net income: $43 billion for the quarter
- Full-year revenue: $215.9 billion, up 65% from the prior year
One ratio deserves a longer look: netting $43 billion against $68.1 billion of revenue yields a net margin north of sixty percent. Semiconductor history has never seen this scale arrive together with this profitability. That figure carries two messages at once — AI compute remains a scarce resource, and buyers will keep paying for that scarcity at current prices. Margins that fat at that scale tend to draw competition; they also reveal how few credible alternatives buyers believe they have right now.
What 65% Annual Growth Tells You
Roughly $85 billion of new annual revenue admits only one explanation: procurement of large-scale AI infrastructure continues to speed up rather than cool off. The customers behind these GPUs — AI labs, cloud providers, large enterprises — effectively cast a real-money ballot for the fiscal year’s AI capex. The 20% climb over the prior quarter supplies a second data point: supply constraints failed to hold shipments down, and the market soaked up every bit of capacity that came online. The base effect deserves attention, too — 65% is not growth measured from a small number, since it builds on a full year that had already cleared $130 billion. Sustaining that pace from that base is not what a one-time build-out wave looks like; buyers are turning compute spend into a permanent budget line, planned each year and renewed like any other operating cost.
Two Signals the Industry Should Heed
Signal one: the industry’s bill keeps swelling. Downstream costs for training the next model generation and serving inference ultimately flow together into NVIDIA’s record revenue, and that hardware bill has no structural reason to drop in the near term. Signal two: dependence keeps deepening. The entire AI stack now runs on one vendor’s shipment schedule, so any disruption in capacity, yields, or policy lands directly in every lab’s product timeline. The bigger each quarter gets, the more concentrated that failure point becomes. This risk is no longer NVIDIA’s to own alone — the whole industry carries it now.
How Buyers Should Respond
- Build capacity plans around rising competition: as the pool of buyers grows, predictable access to large GPU allocations erodes, so critical workloads should carry schedule buffers
- Make compute cost a first-class metric: an upstream supply chain earning sixty-percent-plus margins will keep pushing cost downstream, so unit-inference-cost monitoring should sit inside day-to-day operations
- Steer clear of single-point dependence: preserving architectural room to switch is the most workable hedge against concentrated supply