Nvidia CEO Jensen Huang announced at GTC 2026 that he sees AI hardware revenue reaching at least $1 trillion through 2027, roughly doubling his earlier projection of $500 billion through end of 2026.
The $1 trillion figure covers sales of Nvidia's Grace Blackwell and Vera Rubin chips, and Huang indicated the final result would likely be higher — adding anticipated revenue from Vera, Groq 3, and storage rack products will push the total well beyond that baseline.
Approximately 60% of sales are expected from hyperscalers — Amazon, Google, Meta, and Microsoft — with the remaining 40% from neocloud providers, industrial and enterprise use cases, and sovereign AI initiatives.
The projection reflects surging enterprise demand for computational capacity, driven by the explosive growth of agentic AI workloads. Companies are moving beyond simple chatbot deployments to running fleets of autonomous AI agents that require substantial GPU infrastructure.
Nvidia shares briefly spiked on the announcement, which came alongside a series of major product unveilings at GTC including NemoClaw, DLSS 5, the Vera Rubin Space Module, and partnerships across the robotics and autonomous vehicle sectors.
The scale of the projection underscores how AI infrastructure spending has become one of the largest capital expenditure categories in the technology industry.