Nvidia chief executive Jensen Huang used his appearance at the Goldman Sachs Communacopia + Technology conference on Thursday to repeat a number that would be extraordinary for almost any company at this scale: about 70% year-over-year revenue growth next year. Analysts expect Nvidia to close its current fiscal year at roughly $400 billion, which would put next year’s figure near $680 billion.
It was the second time Huang has put that figure on the record. He first offered the 70% outlook last month alongside another record quarter, and on Thursday he framed it not as a stretch goal but as a reading of the order book. “I think we could grow 70% year over year. We’re confident about that,” he said.
Also read: AI Agents Are Driving Sharp Rises in Public Service Requests Worldwide, Researcher Finds
Why Huang says he can see the demand
The pitch rests on Nvidia’s position as what Huang calls a foundational platform. “Nvidia runs every model. Every single lab can use us,” he said, naming Anthropic, OpenAI, and Google, plus open-weight models that anyone can run.
That ubiquity gives Nvidia unusual visibility into construction pipelines. Huang said the company tracks every gigawatt of land, power, and shell — the empty building before computers go in — around the world, and hears from neoclouds, OEMs, cloud providers, and AI-native startups as they plan capacity. “We’re working with everybody, and so we kind of know where everything is,” he said.
Also read: OpenAI Pauses New ChatGPT Pro Sign-Ups as Astra Demand Strains Its Systems
He also pushed back on the idea that Nvidia is still a chip company in the consumer sense. One current GPU system, he said, is not a $399 graphics card but an $8.5 million assembly of roughly 2 million parts drawing 250,000 kilowatts, linked by NVLink. “You need airplanes to ship what we build,” he said.
The clearest near-term number he offered was for a single product: a computer system combining 36 Grace CPUs with 72 Blackwell GPUs, which he said is growing 27% month over month.
The circular-deals question keeps coming back
Nvidia’s habit of investing in companies that then buy its hardware has drawn comparisons to the vendor-financing arrangements that helped sink suppliers during the dot-com build-out, most famously Lucent Technologies. Huang’s answer was blunt and slightly mischievous. “Well, it’s not circular because we put a little bit of money in, and a lot of money comes back,” he said, adding, “I look at the spreadsheet, we put in $1 and $100 comes back in. Is that circular? If that is, let’s do more of that.”
Behind the joke, he said Nvidia requires evidence of real customer contracts before writing a check, and that he has reviewed about $100 billion worth of such contracts. “I’m not taking any risks. … I need a sure thing.”
Competition is real, and Huang acknowledges part of it
The roster of rivals building alternatives is long and growing. Amazon, Microsoft, and Google each design their own AI accelerators for internal workloads; Anthropic and OpenAI are pursuing silicon of their own; newly public Cerebras and startups such as Etched target niches where general-purpose GPUs are inefficient.
None of that has dented Nvidia’s order book yet, but Huang conceded a structural point that matters for the longer arc. Much of AI’s current growth comes from AI-native startups raising large sums and spending most of that cash on their own compute. As the industry matures, buyers typically get better at squeezing more work out of the same infrastructure and tokens — a dynamic that historically compresses demand growth for hardware vendors.
What to watch from here
Two things will test the 70% claim. The first is whether hyperscaler capital spending keeps expanding at the pace implied by current data center announcements; the second is whether Nvidia’s Grace-Blackwell ramp holds its current monthly trajectory through the next two quarters. Both will show up in supplier commentary — memory makers, cooling vendors, and power providers report earlier than Nvidia does.
For investors, the gap between Huang’s $680 billion and a more conservative consensus is now the central debate in AI hardware. Anything close to the higher number would reshape earnings expectations across the semiconductor supply chain; a miss would sharpen questions about how much of the current build-out is durable demand versus a cyclical peak. This is not financial advice, and the semiconductor and AI infrastructure markets are volatile and inherently uncertain — forecast figures should be treated as management guidance, not guaranteed outcomes.
Huang’s next scheduled appearances and Nvidia’s following quarterly report will be the first real checkpoints. The company’s guidance has proven reliable through this cycle, but the compute market has never been tested at $400 billion of annual revenue, and no vendor in the history of semiconductors has held a position quite like this one for long.
Be the first to comment