SAN JOSE, March 2026 — In a bold declaration that has sent ripples through the tech industry, Nvidia CEO Jensen Huang announced that the company now anticipates at least $1 trillion in revenue through 2027. This updated forecast nearly doubles the “high confidence” demand of $500 billion projected just one year ago, signaling an unprecedented acceleration in the global AI infrastructure build-out.
The Era of Agentic AI
At the heart of this growth is the newly unveiled Vera Rubin platform. Described as a revolutionary leap for “agentic AI,” the system boasts a staggering 40 million times more compute than technology available just a decade ago.
The Vera Rubin suite includes:
- Seven specialized chips designed for next-generation workloads.
- Five rack-scale computers and a dedicated AI supercomputer.
- Sixth-generation scale-up switching, a proprietary technology Huang called Nvidia’s “secret sauce”.
The Great Inflection: From Training to Inference
Huang noted a fundamental shift in the AI market, declaring that the “inflection point of inference” has arrived. While previous years focused on training models, the industry is now racing to deploy them for productive work.
“AI now has to think, to do, to read, and to reason,” Huang explained. “Every time it performs these actions, it has to inference”. This shift is driving an insatiable demand for capacity from major players like OpenAI and Anthropic, who require more tokens to generate revenue and build smarter systems.
Supply Constraints and Skyrocketing Demand
Despite shipping “incredible amounts” of hardware, Nvidia continues to struggle with out-of-the-charts demand. Huang acknowledged that spot pricing is skyrocketing and GPUs remain nearly impossible to find on the open market. To meet this demand, Nvidia is leaning heavily on partners like Samsung, which is operating at maximum capacity to manufacture critical components like the Gro LP30 chip.
The Final Frontier: Space-Based Data Centers
Looking beyond Earth, Huang revealed that Nvidia is already preparing for the next frontier of computing: Space. The company’s Thor processors are already radiation-approved for satellite imaging.
The next phase involves the Vera Rubin Space 1, a specialized computer designed to anchor the first generation of orbital data centers. While admitting the complexity of building infrastructure in space, Huang remains committed to extending Nvidia’s AI reach into the cosmos.
Bottom Line
Nvidia is no longer just a chipmaker; it has become the central architect of a trillion-dollar AI economy. By pivoting from training to the “inference race” and even looking toward orbital infrastructure, Huang is betting that the world’s hunger for compute is only just beginning.
