Nvidia guides for 70% revenue growth in fiscal 2028, its first year-ahead forecast

Nvidia told investors on Wednesday that revenue in fiscal 2028, the year ending about 1 February 2028, should grow 70 percent. The company has never before issued a full-year forecast that far out. Analysts, on average, had been modelling about 44 percent growth for the same period. Apply 70 percent to the $396 billion consensus for fiscal 2027 and fiscal 2028 lands near $673 billion. That figure would put Nvidia ahead of Apple and Alphabet by revenue and behind only Amazon among U.S. technology companies.

The guide arrived after a quarter that already beat estimates. Fiscal second-quarter revenue was $96.2 billion, up 106 percent from a year earlier. The company expects about $108 billion in the current quarter, above a Wall Street figure near $104 billion. Shares rose about 5 percent after hours after an initial dip. CFO Colette Kress delivered the 70 percent number on the earnings call. Chief executive Jensen Huang filled in the reasoning.

Demand above the guide

Huang said demand is much greater than 70 percent and that supply is what lets the company promise 70 with confidence. Memory components remain short. He said AI has reached a point where tokens are productive and profitable, and that compute is now revenue. Non-hyperscale buyers, a group that includes sovereign projects, so-called neo-clouds, AI labs and enterprises, now account for about half of the business and are growing 100 percent a year, he said. AI labs alone are expected to contribute roughly a quarter of sales next year.

Neo-clouds such as Nebius and CoreWeave are on course to exit this year with more than eight gigawatts of Nvidia GPU capacity, up from three gigawatts at the end of last year. Huang said the company has purchase orders from every major hyperscaler, AI cloud and system OEM for Vera Rubin, the next platform after Blackwell. Nvidia expects Vera Rubin systems to make up about 20 percent of data-centre revenue in the current quarter, which would be the fastest product ramp in the company's history. Data centre already provides about 92 percent of quarterly sales. On a $108 billion quarter that implies about $20 billion of Rubin-related hardware in three months.

What the number is for

Huang said the firm wanted everyone to have the same set of information. Critics of the AI buildout have argued that revenue is circular, with chipmakers, clouds and model labs buying from one another. A year-ahead guide that sits $100 billion above the old consensus is meant to close that argument with a number rather than a slogan. It is also a constraint. If memory stays tight, 70 percent is a ceiling set by factories, not a floor set by orders.

Yotta Data Services in India said the same week it is preparing to order 50,000 Vera Rubin GPUs and 45,000 GB300 Blackwell units. Cofounder Sunil Gupta put the Rubin package, including networking, at about $7.5 billion, plus $750 million to build a 120-megawatt D4 hall in Delhi. That single private order is larger than any previous Indian GPU commitment. AMI Intelligence, part of the Greenko group, had been the domestic marker with a binding order for 9,000 Rubin GPUs in Hyderabad.

Nvidia's fiscal 2028 figure will only exist if those halls, and dozens like them, actually draw power and pay invoices. Huang's 70 percent is a supply-limited promise. The customers now placing multi-billion-dollar Rubin orders are the other half of the same sentence.