Earnings Call Highlights
NVIDIA's Q1 FY2027 earnings call was a triumphant, almost euphoric affair, with management projecting supreme confidence across every dimension of the AI infrastructure buildout. Jensen Huang's posture was unambiguously aggressive — framing agentic AI as a demand inflection that has made tokens profitable and compute capacity synonymous with revenue and profit, leaving no room for competitive doubt. Analysts largely played along, probing for upside rather than challenging the narrative, with questions focused on share gains, new TAMs, and product ramp trajectories rather than any signs of deceleration. The dominant theme was that NVIDIA has successfully expanded its addressable market far beyond hyperscale GPUs — into CPUs, networking, sovereign AI, and physical AI — while simultaneously tightening its grip on frontier model inference. The tone was one of a company that believes it has structurally won the AI infrastructure era and is now executing on a multi-year compounding opportunity.
- NVIDIA guided Q2 revenue to $91 billion and disclosed visibility to $1 trillion in combined Blackwell and Rubin revenue from 2025 through calendar 2027, with Vera standalone CPUs, LPX, and growing frontier model share explicitly called out as sources of upside above that figure — a direct signal that the TAM NVIDIA is capturing is expanding faster than prior frameworks suggested.
- Jensen announced Vera CPU as a brand new $200 billion TAM, purpose-built for agentic AI workloads, with $20 billion in standalone Vera CPU revenue visibility already this year and every major hyperscaler and system maker partnering on deployment — framing agentic orchestration as a CPU-intensive workload that complements rather than cannibalizes GPU inference.
- Huang explicitly addressed custom/merchant silicon competition by characterizing SRAM-based decode accelerators like LPX as niche products with limited context processing and throughput, stating 'the use case for LPX is not broad' and that it serves only providers with premium, high-token-rate services — a pointed dismissal of the competitive threat from inference-focused custom chips.
- The addition of Anthropic as a strategic partner — described as previously representing 'largely zero' NVIDIA coverage — was cited as a major driver of inference share gains, with NVIDIA securing compute capacity for Anthropic across AWS, Azure, CoreWeave, and others, representing a significant new frontier model relationship that directly expands NVIDIA's inference footprint.
- VeraRubin production shipments are set to begin in Q3 with a ramp into Q4 and Q1 FY2028, with Colette Kress noting demand is already committed via purchase orders from nearly all major customers, and Jensen stating 'every single frontier model company will jump on VeraRubin from the get go' — a stronger adoption signal than Blackwell received at launch.