NVIDIA sees $1T chip market
NVIDIA told analysts it expects up to $1 trillion in AI chip revenue through 2027, a projection highlighted in analyst coverage of GTC 2026 reported. The figure underlines how GTC conversations are centering on massive long‑term chip demand from AI workloads reported.
At NVIDIA’s GTC keynote on March 16, 2026, CEO Jensen Huang said he sees at least $1 trillion in orders for the company’s Blackwell and Vera Rubin AI chips through the end of 2027. (bloomberg.com) Huang contrasted that outlook with last year’s “about $500 billion” high‑confidence demand figure for Blackwell and Rubin through 2026, effectively extending and doubling the company’s near‑term revenue visibility. (techcrunch.com) The $1 trillion call was delivered alongside the public unveiling of the Vera Rubin architecture and new H300 accelerator systems, announced as purpose‑built hardware to scale inference and real‑time AI deployments. (cnbc.com) NVIDIA framed the opportunity around an “agentic AI” and inference pivot—Huang said inference cost per token and real‑time workloads are driving customer orders at scale. (nvidianews.nvidia.com) The company’s scale underpins the projection: NVIDIA reported record Q4 fiscal‑2026 data‑center revenue of $62.3 billion and full‑year revenue of $215.9 billion, figures it cited when outlining the climb to larger AI spend. (nvidianews.nvidia.com) Strategic moves to lock down inference technology include a roughly $20 billion deal to license Groq’s IP and onboard senior Groq engineers in December 2025, a transaction NVIDIA described as bolstering its low‑latency inference portfolio. (cnbc.com) At the same time hyperscalers are diversifying: Amazon announced a multiyear plan to put Cerebras wafer‑scale processors into AWS data centers on March 13, 2026, underscoring that cloud providers are mixing GPUs with alternative ASICs for inference. (bloomberg.com)
