Jensen Huang explains why Nvidia will grow an astounding 70% next year

Nvidia’s CEO is telling the Street the AI infrastructure boom has another full year of acceleration left in it. Speaking at Goldman Sachs Communacopia + Technology, Jensen Huang repeated guidance first floated last month: the company thinks it can grow revenue about 70% year over year. With analysts already pointing toward roughly $400 billion this fiscal year, that math lands near $680 billion if the call holds.

The stakes are bigger than one earnings print. Hyperscalers are building their own chips. Anthropic and OpenAI are building their own. Cerebras is public; startups like Etched are circling. Huang’s answer is that Nvidia is no longer “a chip company” in the old sense—and that the market still underprices how deep the platform runs.

If he’s right, capital, power, and talent keep pouring into one stack. If he’s wrong, the circular-investment jokes stop being jokes.

We’re taking Huang at his word on visibility, and still asking what that visibility buys. His pitch is not a better transistor. It’s omniscience: Nvidia runs every major lab’s models, tracks land and power and data-center shells worldwide, and hears from neoclouds, OEMs, hyperscalers, and AI-native buyers. When the person selling the picks and shovels also sits on every dig site, growth guidance stops sounding like hope and starts sounding like a census.

The hardware framing matters. Huang pushed back on the consumer-GPU hangover—$399 cards versus systems that, in his telling, cost millions, pull hundreds of thousands of kilowatts, and ship by the thousands. One Grace-plus-Blackwell system, he said, is still growing sales roughly 27% month over month. That is demand from people building factories, not buying gaming rigs.

I’m watching the circular-deal defense closest. Huang’s quip—that Nvidia puts a little money in and a lot comes back—lands as comedy until you remember Lucent-era supplier financing. He insists investments follow real customer contracts, and says he’s seen on the order of $100 billion of that paper. Fine. The adult test is whether those contracts still clear when AI-native startups stop spending raised capital like water and start optimizing tokens like adults.

Useful AI needs real compute. We’re not rooting against the build-out. We are saying concentration this extreme is a single point of failure for labs, cloud buyers, and national industrial policy. Huang can see the future because he wired himself into it. That is brilliant strategy. It is also why disruption, when it finally arrives, will be violent rather than polite.

For now the read is simple: the party has another year on Huang’s calendar. Treat the 70% call as a claim about platform lock-in, not a vibe about GPUs.

Huang was speaking Thursday at Goldman’s Communacopia + Technology conference, reiterating an outlook Nvidia first aired with its latest record quarter. Competition narrative has shifted from “can anyone catch Nvidia on chips” to “can anyone catch Nvidia on the full stack”—models, interconnect, software, and the physical data-center map.