Nvidia floats double-dipping datacenter financing scheme

AI infrastructure doesn't come cheap. To keep up, rent-a-GPU outfits such as CoreWeave and Lambda have had to borrow billions of dollars from venture capitalists and hedge funds to bankroll their datacenter build outs. So long as their revenues are greater than the interest payments on the loans, they have the potential to make a profit. Unfortunately for entrepreneurs looking to cash in on the AI hype, not everyone with a bright idea can tap into this kind of funding. But don't worry, Nvidia is here to help. In a blog post published this week, the GPU giant floated the idea for a new program that promises to make it easier for emerging AI cloud providers to get the financing they need, although it's not clear that Nvidia itself will be providing the financing - it may only be brokering deals with third-party lenders. Regardless, the GPU provider is expecting a cut of the revenues in exchange. Through the partnership, AI clouds will sell Nvidia-powered cloud services, with Nvidia earning both standard product revenue and a share of the cloud revenue on the supported capacity," the company explained. This structure accelerates adoption of Nvidia platforms among the high-growth, high-conviction AI native sector, and provides Nvidia with a recurring, usage-linked earnings stream." In other words, Nvidia first brings in revenues based on how many of its products are deployed, and later, if the neocloud turns out to be successful, a share of the revenues its hardware generates. It could also provide a bit of insulation against a potential AI bust - if demand for new GPUs falls, Nvidia may still be able to earn a recurring revenue from the GPUs it's already sold, assuming customer demand remains high. Specifics on how this new business model will work in practice are rather thin. Nvidia declined to offer details beyond the contents of its blog post. However, the company has already signed up two customers, Sharon AI and Firmus, to put it to the test. Sharon AI is a sovereign AI cloud provider founded in 2024 based out of Australia, which is looking to deploy as many as 40,000 Grace Blackwell GB300 GPUs in the land down under. Meanwhile, Firmus plans to deploy as many as 170,000 Nvidia GPUs at a 360-megawatt facility in Batam, Indonesia, designed specifically to Nvidia's DSX spec. (R)