Nvidia isn't just selling the picks and shovels of the AI boom anymore. It's financing them, and taking a cut of what they dig up.
Most people still think of Nvidia as a company that designs chips. That's understandable. For the past few years, that's been the entire story: Nvidia builds the GPUs everyone wants, cloud providers buy them as fast as they can, and the company books another extraordinary quarter.
But that description is starting to become incomplete.
The more interesting question today isn't how many GPUs Nvidia can sell. It's how much of the AI economy it wants to finance after those chips have already been delivered.
According to a report from The Information, Nvidia is offering some smaller cloud providers more than just hardware. Nvidia confirmed the shift itself on July 1 in a blog post co-authored by Chief Financial Officer Colette Kress, describing a new AI Compute Partnership that guarantees GPU capacity to startup cloud providers, agrees to rent back unused capacity if their machines sit idle, and takes a share of the revenue that infrastructure generates.
Individually, none of those ideas is revolutionary. Equipment financing has existed for decades. Revenue-sharing agreements are hardly new either. What's new is seeing all of those pieces brought together by the company that already dominates the market for AI chips.
That changes the relationship. Traditionally, Nvidia sold the hardware and moved on. Whether its customers made money with those GPUs wasn't really Nvidia's concern. If demand was strong, they ordered more chips. If demand weakened, they ordered fewer. The transaction ended when the hardware was delivered.
This model is different. Nvidia now has a financial interest in what happens after the sale. The company isn't just supplying infrastructure. It's becoming part of the business built on top of it.
The first announced projects aren't small. Firmus plans to build a data center campus on Batam Island, Indonesia, scaling toward 360 megawatts and housing up to 170,000 GPUs. Sharon AI, an Australian cloud provider, has committed to roughly 40,000 Grace Blackwell GB300 systems as part of a six-year deployment centered on Australia. Together, that's around 210,000 high-end GPUs running under financing arrangements Nvidia helped put together.
More telling than the numbers is who signed off on them. This wasn't a mid-level partnerships announcement. Nvidia's own CFO put her name on it. That makes it hard to dismiss as an experiment. It looks like a strategy.
Nvidia has been moving this way for a while. Last year it backed multibillion-dollar financing deals involving CoreWeave and Lambda. At the time, those looked like pragmatic ways to help two fast-growing AI infrastructure companies expand quickly. Looking at this week's announcement, they read more like a first step.
The attraction for Nvidia is obvious. Selling a GPU produces revenue once. Financing that GPU, guaranteeing some utilization, and sharing in the revenue it generates creates income over a much longer stretch. For a company whose fortunes have depended on hardware sales, that's a real evolution.
Here's the complication, though. Many of the companies Nvidia is now helping to finance compete, at least in part, with Amazon, Microsoft and Google, who also happen to be Nvidia's biggest customers. That's not a problem while demand for AI infrastructure keeps outstripping supply. Everyone is still expanding as fast as they can.
Markets don't stay like that forever.
What I keep coming back to isn't whether this financing model will work. It probably will, at least while demand stays exceptionally strong. The real question is what happens when conditions turn ordinary again. If Nvidia is guaranteeing utilization and helping finance deployments, some of the industry's operating risk shifts back onto Nvidia's own balance sheet. We don't know how large that exposure is, since the commercial terms haven't been made public. It may prove modest.
But investors should recognize that Nvidia is turning into a different kind of company. For years, it's been valued as the business that makes the world's best AI chips. Increasingly, it's positioning itself to earn money from financing, utilization and cloud infrastructure too. That's a bet that could pay off handsomely, or one that quietly loads risk back onto a balance sheet that used to carry almost none.
