NVIDIA Corporation

Fiscal 2025 was extraordinary. Full-year revenue reached $130.5 billion — triple last year. Data center revenue for Q4 was $35.6 billion, up 93% year over year. The age of AI infrastructure build-out is accelerating, not decelerating. Blackwell architecture shipped $11 billion in its first quarter — the fastest product ramp in our history. Every hyperscaler, CSP, and major enterprise is deploying Blackwell. The demand signal remains extraordinarily strong. Our software business is becoming increasingly important. CUDA ecosystem now has 5 million developers. NVIDIA AI Enterprise is being adopted broadly across industries.

Q4 revenue was $39.3 billion, above our guidance of $37.5 billion. Gross margin was 73.5%. Q1 FY2026 guidance: revenue of $43.0 billion (plus or minus 2%), gross margin of approximately 71%. The slight gross margin step-down reflects Blackwell manufacturing ramp costs, which we expect to normalize by mid-year.

Jensen, how should we think about competition from AMD and custom silicon?

Competition makes us better. But consider the CUDA moat — 5 million developers, 4 million applications, 20 years of software investment. Custom silicon solves a narrow problem. CUDA solves every AI problem. Our system-level advantage compounds with each architecture generation.

The $43B guide is remarkable. Is there any visibility into when demand might normalize?

We continue to see multi-year infrastructure investment cycles from all of our major customers. The sovereign AI trend is adding a new demand vector. We don't see demand normalization in the near term.