AI compute keystone — GPUs + CUDA + networking; the ~90%-share accelerator the whole AI economy depends on
AI exposure 88% — derived from reported segment revenue, not asserted.
Silicon Layer → Semiconductor Devices · figures rebased to Q2 FY2027, reported 2026-08-26
Compute (GPUs/accelerators — H100/H200/Blackwell for AI) leads 3 businesses in this sub-tier that together are 88.4% of revenue, so Semiconductor Devices is effectively the whole company rather than one leg of it. The rest is tagged separately: Network Switching & Hardware at 10%; Robotics & Autonomous Systems at 1.3%.
Also tagged: emerging in Connectivity → Network Switching & Hardware; emerging in Frontier & Emerging Bets → Robotics & Autonomous Systems.
NVIDIA ($96.2B Q2 FY2027, +105.9% YoY; operating income $66.6B on a 69.2% margin). Two market-platform lines: Data Center 92.5% ($89.0B, +117% YoY) = primary B1 Semiconductor Devices on the compute franchise, with AI networking inside it as a B3 tag; Edge Computing 7.5% ($7.2B) = the combined former Gaming, Pro Viz, Auto and OEM lines, with DRIVE and Isaac robotics as a B7 tag. NEW THIS QUARTER: NVIDIA now splits Data Center into Hyperscale $48.7B (50.6% of total revenue) and AI Clouds, Industrial & Enterprise $40.3B (41.9%), recasting prior periods — the first time the customer mix is…
Moat. CUDA software ecosystem + full-stack (GPU+networking+systems) lock-in; ~90% AI-accelerator share. Switching costs are the moat, not just the silicon.
Bottleneck / pricing power. Pricing power is intact at extraordinary scale: operating margin was 69.2% in Q2 FY2027 on revenue that more than doubled year over year. The constraint on shipments is advanced packaging and HBM supply, not demand.
Role in the AI stack. The compute layer — the keystone supplier every hyperscaler depends on.
All 174 companies · What moved this quarter · How the map is built