Nvidia’s $96.2B Quarter Says the AI Buildout Isn’t Slowing
Matthew Leo · Published August 29, 2026 · AI
If the AI infrastructure boom is a bubble waiting for companies to stop spending, Nvidia’s latest quarter did not provide much evidence that the stopping part is close.
Nvidia reported $96.2 billion in fiscal second-quarter revenue on August 26, more than double the same period a year earlier. Data-centre revenue alone reached $89 billion. The company also gave an unusually aggressive longer-term outlook: Reuters reported that Nvidia expects revenue to grow about 70 percent in its next fiscal year, far above the growth Wall Street had been modelling.
The numbers are extraordinary. They are also useful because they show where the AI economy has moved beyond hype. Somebody is writing very large cheques for actual compute.
Data centres now dominate Nvidia
Nvidia’s official results show total quarterly revenue of $96.2 billion, up 18 percent from the previous quarter and 106 percent year over year. Data-centre revenue was $89 billion, up 117 percent from a year earlier.
That means the company is increasingly an AI-infrastructure business that happens to have a famous gaming brand, rather than a gaming-chip company that found a lucrative side market.
The shift has been underway for years, but the scale is now difficult to ignore. Data centres accounted for more than nine out of every ten dollars Nvidia generated in the quarter.
The bigger story is what Nvidia thinks happens next
Companies can produce a spectacular quarter at the top of a cycle. What caught the market’s attention was Nvidia’s confidence that demand remains strong well beyond the current reporting period.
Reuters reported that Nvidia forecast roughly 70 percent revenue growth for the fiscal year ending in January 2028, compared with a much lower Wall Street consensus before the announcement.
Nvidia is effectively saying that the current spending wave is not just a rush to build the first generation of giant AI clusters. It expects customers to keep replacing and expanding them.
That matters because AI hardware ages unusually quickly. A data centre built around one generation of accelerator can look inefficient only a few years later when a newer architecture delivers more compute per watt and per rack.
Amazon just committed to millions more GPUs
One reason Nvidia can sound so confident is that its largest customers keep publishing enormous capacity plans.
Alongside the earnings cycle, Nvidia and Amazon Web Services announced an expansion that is expected to deploy two million additional Nvidia GPUs through 2028. Hyperscalers, specialized AI clouds, national governments and frontier-model labs are all competing for the same class of infrastructure.
This is no longer a market where one or two labs determine demand. Nvidia argues that multiple frontier labs, startups, enterprises and sovereign buyers are scaling at once.
That diversification reduces one risk while creating another: the entire ecosystem is becoming deeply dependent on continued AI demand being economically productive enough to justify the capital spending.
There is a physical limit to how fast Nvidia can grow
The bullish revenue forecast comes with a less glamorous constraint. Memory and other components remain tight.
Days before earnings, reports surfaced that buyers of Nvidia-based AI server systems were being warned about price increases above 15 percent, largely because of higher memory costs. Nvidia’s next-generation systems require enormous amounts of advanced memory, networking and packaging capacity in addition to GPUs.
That is why the company can have more demand than it can comfortably supply while still facing pressure on gross margins.
AI spending is becoming a supply-chain story as much as a model story.
China remains a major uncertainty
Nvidia did not include meaningful China upside in its outlook because U.S. export controls continue to restrict which advanced processors can be sold there.
That is significant for two reasons. China is a huge potential market, but Nvidia is currently producing these growth expectations without assuming it gets full access to it. On the other hand, the restrictions create room for Chinese chipmakers to develop alternatives and reduce long-term dependence on Nvidia.
The company therefore faces an odd strategic balance: extraordinary demand in markets it can serve, while one of the world’s largest technology markets is being pushed toward a competing domestic ecosystem.
What this means for Canadian AI spending
Canada is trying to expand domestic compute capacity at exactly the moment the global price of participating is rising.
A stronger Nvidia roadmap can make Canadian AI infrastructure more capable, but it also raises the bar. Universities, startups and public compute projects are competing for hardware against companies that can commit billions of dollars at a time.
Canada’s advantage will not come from outspending U.S. hyperscalers. It has to come from choosing where domestic compute creates strategic value: research access, sovereign data requirements, Canadian startups and workloads that should not depend entirely on foreign cloud allocation.
Nvidia’s results make that policy question more urgent, not less.
The bubble question is changing
It is still reasonable to ask whether AI investment is running ahead of proven returns. Hundreds of billions of dollars are being committed to infrastructure based partly on expectations about future demand that may not arrive exactly as projected.
But calling the entire cycle “hype” now requires explaining $96.2 billion in quarterly Nvidia revenue and customers lining up for the next generation before the current one is fully deployed.
The more useful question is where the eventual returns land.
If AI applications generate enough productivity and revenue to pay for the infrastructure, Nvidia’s growth can remain extraordinary. If they do not, the correction will travel backward from software companies to cloud providers, data-centre developers and chip suppliers.
For now, the money is still travelling in the other direction, and very quickly.
Tags: Nvidia, AI, Earnings, data centres, Analysis