Nvidia is changing how it gets paid. The chip giant told customers it will take a cut of some cloud providers’ revenues, according to The Information. It’s a shift that moves Nvidia from selling hardware once to collecting money on an ongoing basis, and it could reshape the economics of the entire AI cloud market.
Here’s what The Information reports and why it matters.
What happened
Nvidia has signaled to certain customers that it plans to take a share of the cloud revenue they generate, as detailed in The Information. In plain terms: some companies that buy Nvidia GPUs and rent them out as cloud compute would owe Nvidia a piece of what they earn from those rentals.
That’s a big departure from the standard deal. Until now, the arrangement was simple. You buy the chips, you own them, you rent them out, and whatever you make is yours. Nvidia got paid at the point of sale and walked away.
Why this is significant
What stands out here is the move toward recurring revenue on top of hardware sales. Nvidia already commands the AI chip market, with GPUs like the H100 and Blackwell line powering most large model training. Now it’s looking to capture value not just from the sale, but from the use.
Think about the leverage. When demand for your product wildly outstrips supply, you set the terms. Nvidia allocates chips, decides who gets priority, and now, apparently, can ask for a cut of the downstream business those chips create. That’s the kind of pricing power few companies ever hold.
This matters for a few groups:
- Neoclouds like CoreWeave, Lambda, and Nebius, whose entire business is buying Nvidia GPUs and renting them out. A revenue share hits their margins directly.
- Big cloud providers (AWS, Microsoft, Google) that buy Nvidia at massive scale and may face pressure to accept new terms.
- AI startups renting compute, who could see costs passed down the chain.
The context behind the move
Nvidia has been doing more than selling chips for a while. It’s been investing in the same cloud companies that buy its hardware, backing firms like CoreWeave and pouring money across the AI ecosystem. Critics have raised questions about circular financing, where Nvidia funds customers who then spend that money buying Nvidia chips.
A revenue-sharing model is another turn of that same screw. It ties Nvidia’s fortunes even more tightly to how much money the AI cloud market actually produces, not just how many chips it can ship. If AI compute keeps printing money, Nvidia wants to be there for every dollar, not only the first one.
What to watch next
A few things will tell us how far this goes.
- Which customers agree. Revenue sharing only works if buyers accept it. Watch whether the large hyperscalers, who have their own leverage and are building custom chips, push back.
- Margin pressure on neoclouds. These companies already run on thin margins and heavy debt to fund GPU buys. A cut to Nvidia squeezes an already tight model.
- Custom silicon acceleration. Every move like this gives Amazon, Google, and others one more reason to invest in their own chips and reduce Nvidia dependence.
- Regulatory attention. A dominant supplier taking a share of customer revenue is the kind of thing antitrust watchers notice.
This is a signal about where Nvidia thinks the power sits. Selling the shovels in a gold rush is a great business. Taking a cut of the gold is a better one. Whether customers go along with it is the real question, and their answer will say a lot about how much room anyone has to say no to Nvidia right now.
For the full reporting and specifics on which customers are affected, see the original story at The Information.