Jensen Huang had told an audience roughly twelve months ago that Nvidia had $500 billion in high-confidence demand and purchase orders for its Blackwell and Rubin chips through 2026. On Monday at the GTC conference in San Jose, he stood in front of more than 30,000 developers at the SAP Center and said the number was now at least $1 trillion. He did not say it with the caution of someone hedging a forecast. He said it and then promised to spend the next two hours explaining why it makes sense.
“Now, I don’t know if you guys feel the same way, but $500 billion is an enormous amount of revenue,” Huang told the crowd. “Well, I’m here to tell you that right now where I stand — a few short months after GTC DC, one year after last GTC — right here where I stand, I see through 2027, at least $1 trillion.” He then added, without embellishment, that computing demand will be “much higher than that,” and that supply is the binding constraint. “In fact, we are going to be short,” he said.
Nvidia’s actual reported revenues give that projection at least some grounding in observable reality. The company’s fiscal year 2026, which ended January 25, produced $215.9 billion in revenue, a 65% year-over-year increase and the company’s eleventh consecutive quarter of revenue growth above 55%. The chipmaker has announced that year-over-year revenue will surge approximately 77% this quarter alone to roughly $78 billion, which means the cumulative math toward $1 trillion is not as astronomical as it sounds on first hearing.
The central product argument at GTC was the Vera Rubin platform, Nvidia’s most complex AI system to date and the successor to Grace Blackwell. It is not a single chip but a complete supercomputer platform comprising seven chips, five rack-scale systems and one consolidated supercomputer designed specifically for agentic AI. The Vera Rubin NVL72 rack integrates 72 Rubin GPUs and 36 Vera CPUs through the next-generation NVLink 6 interconnect, which reduces data transfer latency to the microsecond level. The platform also delivers 10 times more performance per watt than its predecessor, a claim that matters enormously given the data centre energy crisis.
Huang also unveiled the Groq 3 Language Processing Unit, the company’s first chip from the startup it acquired through a $17 billion December 2025 deal, calling it the world’s fastest inference chip with the lowest power consumption. It is expected to ship in Q3, and Huang said it could boost tokens-per-watt performance for Rubin GPUs by up to 35 times. These are claims that will be stress-tested by enterprise customers before Wall Street fully prices them in, but coming from Huang in a room full of engineers, they are not treated as marketing.
The strategic shift at the heart of the entire presentation is the move from training AI to running it. Huang said the AI industry has reached an “inflection point for inference,” meaning that the era of building massive models is giving way to the era of deploying them at scale across millions of enterprise and consumer applications. That shift is what underpins the demand acceleration, because inference computing requires a fundamentally different and larger infrastructure footprint than training. “Finally, AI is able to do productive work,” Huang said, “and therefore the inflection point of inference has arrived.”
Uber announced a partnership under which it will launch a fleet powered by Nvidia’s Drive AV autonomous vehicle software across 28 cities in four continents by 2028, starting with Los Angeles and San Francisco in 2027. Nissan, BYD, Geely, Isuzu and Hyundai are all building level 4 autonomous vehicles on Nvidia’s Drive Hyperion programme. These partnerships represent meaningful revenue diversification beyond the hyperscaler concentration that has historically accounted for around 60% of Nvidia’s revenue.
Emarketer analyst Jacob Bourne’s assessment was measured but positive. “Huang mapping out a $1 trillion opportunity through 2027 underscores the durable demand for Nvidia’s AI infrastructure despite investor concerns,” Bourne said. “It signals Nvidia is sustaining its leadership in the AI chip market while the overall AI industry expands beyond early experimentation into large-scale deployment.” Nvidia’s stock climbed modestly in the sessions following GTC, reflecting a market that already had much of the upside priced in but had not anticipated the scale of the demand revision.
Huang also previewed Feynman, the architecture generation planned for 2028, and revealed that Nvidia is heading to space with a Vera Rubin Space Module designed for orbital data centres and geospatial intelligence applications. For a company that opened its GTC in 2023 selling gaming chips into a struggling consumer electronics market, the scope of where Nvidia now operates is genuinely difficult to comprehend from any single vantage point.