$81.6 billion. In one quarter.
These aren't projections or guidance—money already in the bank. NVIDIA today reported its Q1 FY2027 results for the period ended April 26, 2026, raising the stakes of the AI infrastructure arms race yet another notch.
The numbers, no translation needed: total revenue $81.6B, up 20% sequentially and 85% year over year. Data center revenue alone: $75.2B, up a staggering 92%. GAAP gross margin 74.9%, net income $58.3B.
The percentages have become numbing. Try this instead: one quarter of profit exceeds most tech companies' lifetime revenue.
Jensen Huang's language in the earnings release is worth parsing. He's no longer talking about "training large models"—the phrase is now "agentic AI." He says it's "already doing productive work." Then NVIDIA announced a reporting framework change: going forward, just two market platforms—Data Center and Edge Computing. Data Center splits into Hyperscale and ACIE (AI Clouds, Industrial, Enterprise).
This restructuring isn't the finance team being bored. It reveals NVIDIA's own read of the market: AI demand has moved beyond "a few big companies training models" to "every industry building its own AI factories." ACIE captures non-public-cloud deployments—sovereign AI, enterprise private clusters, industrial scenarios. This segment is growing faster than Hyperscale.
The breakdown: data center compute revenue $60.4B (+77% YoY), data center networking revenue $14.8B (+199% YoY). Networking is growing faster than compute, suggesting cluster sizes are scaling and interconnect bottlenecks are becoming the primary constraint. InfiniBand and Spectrum-X shipping cadence is accelerating.
Vera Rubin—the next-gen architecture—is "on track for the second half of this year, starting in Q3." That's restrained language. In NVIDIA-speak, "on track" usually means "already taping out."
Another number worth noting: NVIDIA authorized an additional $80B in share repurchases and raised its quarterly cash dividend from $0.01 to $0.25 per share—a 25x increase. This isn't charity. It's a message to the market: we're not short on cash.
The company returned ~$20B to shareholders this quarter, with $38.5B remaining under its existing authorization. With the new $80B added, the total sits near $120B.
One concern: PC sales are declining, with NVIDIA specifically naming "RAM shortage and price hikes" as a headwind. But compared to data center growth, this is a drop in the ocean.
My read on this earnings report is straightforward: NVIDIA is no longer a "chip company." It's the tax collector of the entire AI infrastructure stack. Whoever wins—OpenAI, Anthropic, Google, Meta—they all pay NVIDIA. And they're paying more every quarter.
The only question is how many quarters this growth rate can sustain. At $75.2B in data center revenue, the base is so large that every percentage point requires astronomical new demand. But at least until Vera Rubin ramps, the moat remains wide.
Key sources: