Jensen Huang used the opening keynote of Nvidia’s annual GTC conference in San Jose to dramatically raise the company’s own revenue ambitions, telling an arena packed with more than 30,000 attendees that he expects purchase orders for the company’s Blackwell and Vera Rubin AI processors to reach at least $1 trillion through 2027. Last year’s equivalent projection was $500 billion.
The conference, which runs March 16 through 19 at the SAP Center in San Jose, has become one of the most closely watched events in the technology calendar, less for individual product announcements and more for what Huang’s framing reveals about where Nvidia intends to take its Business next. This year’s central thesis was that AI is no longer primarily a training story. It is becoming an inference story, and Nvidia intends to own that transition as completely as it has owned the training market.
Huang described 2026 as an “inflection point for inference,” a framing that has significant commercial implications. Training large models requires enormous upfront GPU investment concentrated in a small number of hyperscalers. Inference, the process by which those models generate responses for end users, is distributed, continuous, and scales with usage. The market for inference infrastructure is potentially far larger than the training market, and far harder for custom chip efforts from Google, Amazon, and others to fully displace.
Among the specific announcements, Huang introduced NemoClaw, an open-source stack designed to make the fast-growing OpenClaw AI agent platform enterprise-ready. He also announced that Uber will launch an autonomous vehicle fleet powered by Nvidia’s Drive AV software across 28 cities on four continents by 2028, beginning with Los Angeles and San Francisco next year. Automotive partners building level 4 autonomous vehicles on Nvidia’s Drive Hyperion platform now include Nissan, BYD, Geely, Isuzu, and Hyundai.
The company’s fiscal 2026 revenue came in at $215.9 billion, with quarterly data center revenue of $62.3 billion. Those numbers provide the foundation for the $1 trillion projection, though Huang acknowledged during the keynote that the estimate deserves scrutiny. “Does it make any sense?” he said to the crowd, before spending the remainder of the address explaining why he believes it does.
The GTC event also served as a platform for Nvidia to assert its role not just as a chip company but as the infrastructure layer beneath the entire AI economy. Chips, software, systems, robotics, autonomous vehicles, industrial digital twins: Huang framed all of these as pieces of a single connected story in which Nvidia supplies the underlying computation. The leather jacket has become a symbol of something more than personal style; it signals that Nvidia sees itself as the defining company of the AI industrial era rather than just its most profitable hardware vendor.
For investors who have ridden the stock through its extraordinary run, the keynote offered both validation and a challenge. The validation is that demand remains far stronger than most skeptics expected even a year ago. The challenge is that a $1 trillion order pipeline projection is now the baseline, and anything short of that over the next eighteen months will be read as disappointment.
Nvidia shares rose more than 1% in the run-up to the conference and have been among the steadier performers in a market that has otherwise been grinding lower through March.