Nvidia's Huang Reaffirms 70% Growth Forecast, Defends Circular Deal Concerns

Jensen Huang did not soften his forecast this week. Speaking Thursday at the Goldman Sachs Communacopia + Technology conference, Nvidia's CEO reiterated that his company expects revenue to grow 70% year over year in its next fiscal cycle. That is not a new number. Huang first floated it last month when Nvidia reported another record quarter. But repeating it on a public stage, unprompted by earnings pressure, signals conviction rather than a one-off talking point.

The math behind the claim is worth sitting with. Analysts project Nvidia will close its current fiscal year near $400 billion in revenue. A 70% increase would push that figure to roughly $680 billion next year. For context, that single-year increment, about $280 billion, exceeds the entire annual revenue of most Fortune 500 companies. Few semiconductor firms in history have sustained growth at this scale off a base this large.

Why It Matters

Huang's confidence rests on a claim of near-total market visibility. "Nvidia runs every model. Every single lab can use us," he said, citing Anthropic, OpenAI, Google, and the open-weight ecosystem as customers. He described the company as tracking "every single gigawatt of land, power, shell around the world," pulling in data from neoclouds, OEMs, cloud providers, and AI-native startups that report activity back to Nvidia. That is an unusual position for a chipmaker: less a vendor, more a real-time sensor network for the entire AI buildout.

The scale of a single product illustrates the shift. Huang noted that one GPU used to retail for $399 in the PC gaming era. Today, a fully configured unit runs $8.5 million, comprising 2 million parts and drawing 250,000 kilowatts. Orders for the GB200 NVL72 system, which pairs 36 Grace CPUs with 72 Blackwell GPUs, are growing 27% month over month. These are not incremental upgrades. They represent a category of compute infrastructure that barely existed five years ago.

Competitive pressure is real, but so far it has not dented demand. Amazon, Microsoft, and Google are building in-house silicon. Anthropic and OpenAI are doing the same. Newly public Cerebras and startups like Etched, which recently hit a $5 billion valuation on $1 billion in sales, are chasing slices of the AI chip market. None of this has slowed Nvidia's order book, at least not yet.

The circularity question deserves scrutiny. Nvidia has taken equity stakes in companies that then purchase its hardware, a structure that draws uncomfortable comparisons to the dot-com era buildout financing that eventually sank suppliers like Lucent Technologies. Huang's response was blunt and, frankly, a little cavalier: "It's not circular because we put a little bit of money in, and a lot of money comes back." He put a number on it, claiming $1 invested returns $100. He also said Nvidia has verified $100 billion in real customer contracts before making these investments, framing the strategy as underwritten rather than speculative. "I'm not taking any risks," he said. "I need a sure thing."

That framing deserves a healthy dose of skepticism, not because Huang is lying, but because every vendor with a circular financing arrangement describes it as fundamentally different from the last one that failed. The contracts may well be real. The concentration risk, Nvidia's fortunes tied to a handful of AI labs burning enormous cash to build their own infrastructure, is real too. Huang himself conceded that much of current AI growth comes from startups raising large rounds and spending most of it on their own compute. That is not a durable demand curve. It is a subsidized one, and subsidies eventually get reassessed.

History offers a caution here. Every dominant infrastructure layer in tech eventually faces margin compression as the market matures and buyers get efficient. Huang acknowledged as much, noting that as the AI industry matures, companies will use infrastructure and tokens more efficiently, likely reducing the raw volume of chips needed per unit of output. That is standard technology lifecycle behavior. It does not mean Nvidia's near-term guidance is wrong. It means the multi-year growth assumption embedded in current valuations deserves a discount for eventual efficiency gains.

The Bottom Line

Nvidia's near-term numbers are impressive and, on the evidence presented, credible. A 70% revenue growth forecast off a $400 billion base is extraordinary, but it is backed by concrete order data: 27% month-over-month growth on flagship systems, embeddedness across every major AI lab, and a claimed $100 billion in verified contracts underpinning its investment strategy. Investors should not dismiss the guidance as hype.

They should, however, price in two distinct risks. First, the circular investment structure, however Huang characterizes it, ties Nvidia's growth to the continued willingness of capital markets to fund AI labs that spend disproportionately on Nvidia hardware. If that funding tightens, the flywheel slows. Second, competitive inroads from hyperscaler in-house silicon and challengers like Cerebras and Etched are early stage today but could compress margins over a two to three year horizon, even if unit volume keeps climbing.

For now, Nvidia's position looks less like a single company and more like critical infrastructure for an entire industry. That is a powerful moat. It is also, by definition, a concentration risk. Watch the contract disclosures closely, watch hyperscaler capex commentary each quarter, and treat the 70% figure as a well-supported target rather than a guaranteed outcome. The AI buildout is still young, and young markets rerate quickly when growth assumptions shift even modestly.