Situation report: Nvidia is rethinking the revenue-sharing deals it has with cloud companies, according to The Information. The details made public so far are thin. Still, the direction matters. The world’s most valuable AI chip supplier is changing how it splits money with the companies that rent its hardware to everyone else.
This isn’t a pricing tweak. It’s about who owns the customer relationship in AI compute and who keeps the profit from it.
Current Situation
The Information’s report frames this as a rethink, not a finished restructuring. Here’s what that means right now:
- The news itself: Nvidia is reconsidering the terms of revenue-sharing deals with cloud firms.
- The open questions: It isn’t yet clear which partners are affected, how the terms might change, or when.
- Why to read carefully: Until the new terms are public, treat any specific numbers you see elsewhere with caution.
Background
To see why this matters, look at how Nvidia has moved past just selling chips over the last few years.
- DGX Cloud. Nvidia launched its own AI cloud service. It ran on hardware hosted inside the big cloud providers’ data centers. Nvidia sold access directly to enterprises, and the hosting partners took a cut. That put Nvidia in an awkward spot: it supplied its biggest customers and competed with them at the same time.
- Neocloud ties. Nvidia built close financial links with newer GPU cloud providers like CoreWeave and Lambda. Those links included investments and agreements to rent back capacity. Critics call this circular: Nvidia helps fund the buyers of its own chips.
- Hyperscaler tension. Microsoft, Google and Amazon all build their own AI chips to cut their reliance on Nvidia. Any deal where Nvidia shares in their cloud revenue, or vice versa, carries strategic baggage.
Why It Matters
What stands out here is the timing. AI compute demand is still huge, but the market is maturing. Pricing for GPU rentals has gotten more competitive, and investors are looking harder at how sustainable AI infrastructure spending is.
A rethink could go a few ways:
- Nvidia pulls back. It cedes more of the customer relationship to cloud partners, keeps its focus on selling hardware and software, and eases tension with its largest buyers.
- Nvidia pushes harder. It asks for better economics and treats its own cloud offering as a bigger profit center.
- Nvidia rebalances by partner. It sets different terms for hyperscalers and neoclouds, depending on how much leverage each side has.
Each path changes the economics for the cloud companies that depend on Nvidia’s supply. Neoclouds would feel it most, since their business models lean heavily on Nvidia’s support.
What It Means for You
If you buy or build on AI compute, this one is worth watching.
- GPU rental pricing. Changes to revenue splits can show up in what cloud providers charge for Nvidia-powered instances.
- Where you buy compute. If Nvidia changes how it sells cloud access directly, the line between going to Nvidia and going to your cloud provider could shift.
- Neocloud stability. Smaller GPU clouds that rely on Nvidia ties may face pressure if the terms tighten. If you run important workloads on them, keep a backup provider in mind.
- Hyperscaler chip push. Any friction here gives Amazon, Google and Microsoft more reason to promote their in-house chips. Expect more aggressive pricing on alternatives like TPUs and Trainium.
What to Watch
- Partner statements or contract disclosures that show which deals are changing.
- Commentary on Nvidia’s next earnings call about cloud services revenue.
- Pricing moves from neoclouds over the coming quarters.
- Any change to Nvidia’s positioning of DGX Cloud.
Assessment: Nvidia controls the most important chokepoint in AI. When it rewrites the money terms with its distribution partners, the effects spread through the whole compute supply chain. The specifics will decide who wins. For now, the signal is that Nvidia is reassessing where it wants to make money in the AI stack. The Information has the full report for readers who want more detail.