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

Nvidia told investors on Wednesday it expects revenue to rise by about 70% in fiscal 2028. Chief financial officer Colette Kress delivered the number on the post-earnings call. Analysts had been modelling roughly 44% growth for that year. Chief executive Jensen Huang said the company had never before guided a full year ahead.

The quarter that just closed, the fiscal second quarter ended 26 July, produced $96.2 billion in revenue, up 106% from a year earlier. Wall Street had looked for about $92.1 billion. Adjusted earnings were $2.22 a share against estimates near $2.09. Data-centre revenue, the line that now defines the company, was $89 billion, up 117%.

Guidance for the current quarter is $108 billion, plus or minus 2%. That would be Nvidia's first $100 billion quarter. Consensus had sat near $104 billion. Shares slipped in the first minutes after the print, then rose in after-hours trade once Kress put the 70% figure on the table.

Kress was careful about the constraint. Customer forecasts, she said, point to growth doubling next year. Supply does not. Memory components remain tight. Huang put the same point in plainer words: demand is running closer to 100%, but the factories and the memory market only support a confident 70%. If the supply chain opened, the year-ahead number would be higher.

The company also tied itself more tightly to Amazon Web Services. Nvidia and AWS said they would deploy 2 million additional GPUs across 2027 and 2028. Kress said capital spending by the top five hyperscalers is heading toward $800 billion in 2026 and $1.3 trillion in 2027, with cloud industry backlog already above $2 trillion. Those figures are the demand side of the 70% claim. They are also the reason critics of circular financing have stayed loud. Nvidia has been extending capital support to some customers. Huang defended that practice on the call.

Vera Rubin, the next architecture after Blackwell, is in full production, Huang said. He framed it as the chip built for the present build-out rather than a future one. OpenAI and other model labs remain named buyers. The new piece in the mix, according to Huang, is that AI clouds, enterprises, sovereign buyers and industrial customers are now growing faster than the original four or five hyperscalers that started the cycle.

That last claim is the information that matters for anyone who already knew Nvidia would beat the quarter. The bull case for the last two years has been that a handful of American cloud companies would keep buying every chip Nvidia could make. The 70% guide only holds if the buyer list widens and if memory supply does not snap. Kress made both conditions explicit.

Asian chip stocks moved first. South Korea's Kospi, which tracks memory names as much as it tracks the local economy, rose more than 1% in Thursday trade. SK Hynix, Samsung Electronics and Kioxia led the region. Nasdaq 100 futures added about 1% in the Tokyo morning. Brent crude, trading on a different story in the Strait of Hormuz, slipped toward $87 a barrel and gave equity desks a second reason to bid risk.

Fiscal 2028 for Nvidia runs from February 2027 to January 2028. Applying 70% to the revenue now expected for fiscal 2027 implies a year in the region of $690 billion to $700 billion. Analysts had been writing $570 billion. The gap is more than $100 billion of sales that do not yet sit in published models.

None of this settles the argument about whether AI infrastructure spending can keep compounding. It does change the number the argument has to beat. Nvidia has put a year-ahead figure on the record for the first time. The company says the limit is silicon and memory, not orders. The next four quarters will show whether the memory market agrees.