On February 25, NVIDIA reported results for the fourth quarter of fiscal 2026. Quarterly revenue came in at $68.1 billion — a record, and up 20% from the previous quarter — with quarterly net income of $43 billion.
The full-year picture is the real headline: FY2026 revenue of $215.9 billion, up 65% year over year. For a company already at this scale, 65% growth means adding roughly $85 billion of revenue in twelve months. Work the math backward — $215.9B divided by 1.65 puts the prior year near $131B — and the number speaks for itself.
Breaking Down the Quarter
- Quarterly revenue: $68.1 billion, an all-time record, up 20% quarter over quarter
- Quarterly net income: $43 billion
- Full-year revenue: $215.9 billion, up 65% year over year
The figure worth pausing on is net income against revenue: $43B on $68.1B is a net margin above sixty percent. At this scale, that combination has no real precedent in semiconductor history. It measures two things at once — how scarce AI compute remains, and what buyers are currently willing to pay for that scarcity. Margin that fat at this scale tends to attract competition; it also tells you how few credible alternatives buyers currently believe they have.
Reading the 65% Full-Year Growth
Adding roughly $85 billion in annual revenue has exactly one explanation: large-scale AI infrastructure procurement is still accelerating, not cooling. The customers behind these GPUs — AI labs, cloud providers, large enterprises — effectively voted with real money on this fiscal year’s AI capex. The 20% sequential jump adds a second data point: supply constraints did not stop shipments from climbing. Whatever capacity came online, the market absorbed. Note the base effect, too: 65% is not growth off a small number — it compounds on a full year already past $130 billion. Holding that pace at this scale means buyers are not riding a one-time build-out wave; they are converting compute into a standing line item, budgeted and renewed like any other operating cost.
Two Signals for the Industry
First, the industry’s bill keeps growing. NVIDIA’s record revenue is a mirror image of downstream costs — the hardware bill for training the next model generation and serving inference has no structural reason to fall anytime soon. Second, dependence is deepening. When the entire AI stack sits on one vendor’s shipment schedule, any wobble in capacity, yields, or policy transmits directly into every lab’s product timeline. The larger the quarter, the more concentrated that single point of failure becomes. That risk no longer belongs to NVIDIA alone; it belongs to the whole industry.
What It Means for Buyers
- Plan capacity assuming more competition: more buyers means less predictable access to large GPU allocations, so critical workloads need schedule buffers
- Treat compute cost as a first-class metric: a supply chain earning sixty-percent-plus margins upstream will keep passing costs downstream, so unit-inference-cost monitoring belongs in daily operations
- Avoid single-point dependence: keeping architectural room to switch is the most practical hedge against concentrated supply
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.