What if the loudest number in AI right now isn’t about model size, benchmark scores, or how many jobs a chatbot might replace, but a single percentage a CEO said out loud on a Thursday?
Nvidia’s Jensen Huang told an audience at Goldman Sachs’ Communacopia + Technology Conference that his company could grow revenue 70% year over year. His words: “I think we could grow 70% year over year. We’re confident about that.” He said it again, in case anyone thought the first time was a slip.
If you’ve ever wondered what the AI boom looks like when you strip away the demos and the hype reels, this is it. A number, said twice, on stage.
The math, in plain language
Analysts expect Nvidia to finish its current fiscal year at roughly $400 billion in revenue. Apply 70% growth to that and you land somewhere near $680 billion.
Those figures are hard to feel, so try this. Nvidia would be adding, in a single year, more revenue than most large companies generate in their entire existence. Not growing from small to medium. Growing from enormous to something we don’t really have a word for.
Huang’s stated reasons are simple enough to fit on a napkin:
- Nvidia dominates the market for AI computing.
- Demand for its products keeps outrunning what it can make.
He also mentioned being constrained by supply, which is a peculiar kind of problem to have. It means the ceiling on growth isn’t customers losing interest. It’s factories.
Why an AI agents site cares about chip forecasts
Here’s my read, and I’ll flag clearly that this is analysis rather than something Huang said.
When you use an AI agent — the kind that books your travel, sorts your inbox, or works through a research task while you do something else — you’re not making one request to one model. You’re triggering a chain. The agent thinks, calls a tool, reads the result, thinks again, tries a different approach, checks its work. Each of those steps burns computing power on a server somewhere.
A single chatbot question is a sip. An agent completing a task is closer to a long drink. Multiply that by millions of people who’ve started handing over whole workflows instead of asking one-off questions, and the demand curve for computing hardware stops looking like a line and starts looking like a wall.
Huang didn’t attribute his forecast to agents specifically. What he described was demand exceeding supply. But if you’re trying to understand why that demand refuses to cool off, the shift from “ask the AI a question” to “give the AI a job” is a reasonable place to look.
How to read a forecast like this without getting swept up
A CEO predicting enormous growth for his own company is not a neutral observer. That’s not cynicism, just basic literacy about who’s talking. Huang has an interest in confidence, and he’s very good at projecting it.
What makes this particular forecast interesting is that it’s specific and falsifiable. He didn’t say “strong growth ahead” or “we see great opportunity.” He said 70%, and he said it in a setting where analysts write it down. That’s a claim you can check later, which is more than most executive optimism offers.
Some useful things to hold in mind:
- Analyst expectations are estimates, not facts. The $400 billion figure is a projection, and the $680 billion follows from it.
- A supply constraint cuts both ways. It signals real demand today, and it means the forecast depends on manufacturing keeping pace.
- Dominance in a market invites competition. Nothing about the current position guarantees the next one.
What this actually means for you
If you’re a non-technical person trying to figure out whether AI agents are a passing fashion or something you should learn to work with, forecasts like this are a useful signal — not because they’re guaranteed, but because of what they reveal about where money is moving.
Companies don’t order hundreds of billions of dollars in computing hardware for a fad. They order it because their customers are already using something, and they’re betting those customers will use more of it. The hardware follows the usage.
So when the man selling the shovels says he expects to sell 70% more shovels, the interesting part isn’t the shovels. It’s the implication that a lot of people are still digging, and that they’ve moved from test holes to serious excavation.
Whether Nvidia hits the number is a question for the earnings calls. What the number tells us about the direction of AI — that agents doing real work are consuming real resources at real scale — is something you can act on now. Learn how these tools work. Try one on a task you actually care about. The infrastructure bet has already been placed.
🕒 Published: