More than a third of Nvidia’s revenue comes from three customers. That’s increasingly precarious.

An interesting reveal from Nvidia’s earnings report yesterday is that 34% of the company’s bonkers $130 billion revenue line last year came from just three customers.

We don’t definitively know who they are because they’re cryptically referred to as “Customers A, B, and C” in Nvidia’s annual report. But it’s a pretty safe bet to say, given who has boasted about capital spending plans and hoarding Nvidia chips, that they’re some of the biggest tech companies on the planet. 

Those customers, part of the vaunted “Magnificent 7” that investors love to pile into, have been shouting some mind-blowing 2025 capex forecasts from the rooftops: $100 billion for Amazon, $80 billion for Microsoft, up to $65 billion for Meta, and the list goes on.

That bodes well in the immediacy for Nvidia — the companies that have been driving its insane growth are saying the firehose will likely remain on this year. 

But Nvidia’s revenue line is getting more concentrated. In fiscal 2023, no Nvidia customer accounted for even 10% of its individual revenue. In 2024, one customer did, and it made up 13% of the top line. This year, three companies topped that 10% threshold: Customers A, B, and C were responsible for 12%, 11%, and 11%, respectively, of all Nvidia revenue.

Is that sustainable or healthy? Think about it: when more and more of your revenue — and thus the profits that drive your stock price — are tied up in fewer customers, and those customers are spending exorbitant amounts of money like they never have before, things could get hairy.

For example, Meta’s anticipated $75 billion of capex in 2025 sounds great if you’re one of the companies it’s writing checks to. Capex, of course, is on the balance sheet and doesn’t flow cleanly through to the income statement. But indulge me the unorthodox financial metric for a second and consider: in its entire life as a public company, Meta has generated $286 billion of profit and has racked up $177 billion of capex. That’s a capex-to-income ratio of 62%.

This year, Wall Street expects Meta to generate $64.9 billion of profit, according to FactSet, which would be its most profitable year ever. But the company says it will dole out $60 billion to $65 billion — roughly that entire year’s worth of profit — on capital spending. Capex-to-income ratio? Right around 100%.

The profits will still be there, but I’m trying to put into perspective the raw size of these spending plans. And of course I should say we don’t know for sure whether Meta is one of the big three customers Nvidia mentioned. But even if it’s not, it’s likely spending in a similar way.

While these types of companies — including, presumably, Customers A, B, and C — are spending boatloads of money to make AI happen, there are still questions about whether the technology will generate significant financial benefits that come anywhere close to the amount of money being invested to develop it. Simultaneously, there’s seemingly a race to the bottom happening, as companies like DeepSeek develop ultracheap AI and the general public seems to coalesce on the opinion that cheaper is “good enough.”

In the tech world, things can change quickly. It was only October 2021 when Facebook changed its name to Meta and Mark Zuckerberg wrote in his founder’s letter, “From now on, we will be metaverse-first, not Facebook-first.” On the company’s most recent earnings call, “metaverse” was mentioned twice. “Facebook” was mentioned 14 times. 

Having a revenue base made up largely of a few big spenders has historically not been a great business idea. Now think about more than one-third of Nvidia’s revenue being made up of spending by similar companies, with similar risk profiles, trying to achieve similar transformative AI outcomes that they hope will be extremely profitable. 

I’m not saying the profits won’t come. But what happens if they don’t?