NVIDIA Corporation

Management framed roughly $1T of Blackwell plus Vera Rubin purchase orders through 2027 at GTC 2026, and Q1 FY2027 revenue was guided to $78B plus or minus 2%. Visibility comes from allocation and order-book economics rather than contractual recurring revenue, so durability depends on the AI capex cycle holding.

Data Center generated about $193.7B of FY2026 revenue, roughly 90% of the company total, with Gaming at 7% and Professional Visualization plus Automotive immaterial. Data Center concentration has risen every year since FY2023 as the AI build accelerated, leaving almost all cash flow tied to one demand driver.

Revenue is billed through a mix of US, Singapore, Taiwan, and China entities, with no single country dominant but Singapore serving as a billing hub for downstream customers. End-demand geography is concentrated in US hyperscalers, which limits the diversification benefit relative to the billing breakdown.

GAAP gross margin was 71.1% in FY2026, 75.0% in FY2025, and 72.7% in FY2024, held at a structurally higher level than AMD at roughly 50% and Intel at 35-40% through the same period. The fabless model hands capital intensity to TSMC, so revenue scales on fixed R&D and software investment rather than proportional capex.

Core revenue is transactional hardware sales to a sophisticated installed base, with CUDA and software subscriptions a small share of the total. Mission-criticality is high because frontier AI training cannot run at comparable efficiency on alternatives, though individual purchase cycles remain project-based rather than contractual.