AI Is Now So Expensive That the Chip Maker Has to Finance It

Building AI at the scale of the biggest companies today requires enormous amounts of power, servers, and space. Reports this week say Nvidia — the company that makes most of the chips inside AI systems — is negotiating to guarantee up to $250 billion in financing so OpenAI can lease a data center campus in southern Ohio. That campus, built on the site of a former uranium enrichment plant, could cost over $500 billion in total, once the chips to fill it are included.

The reason Nvidia is needed is simple: OpenAI cannot easily borrow this much money on its own. The company is worth roughly $852 billion on paper, but it is still losing money — projected losses this year are around $14 billion. Banks will not lend tens of billions without a strong credit rating. Nvidia's balance sheet is solid enough to act as a guarantor — if OpenAI cannot make its payments, Nvidia steps in to cover them.

This creates an unusual loop. Nvidia sells the chips that power AI. Nvidia is now also backing the deal that lets its biggest customer buy more of those chips. On top of the $250 billion lease guarantee, Nvidia is separately discussing another $350 billion package to finance OpenAI's chip purchases for the same campus. The chip maker is effectively guaranteeing its own future sales — which works as long as AI demand keeps growing, but puts an enormous amount of risk in one place.

For anyone building products on AI services today, this is background context — not a reason to panic, but a reason to pay attention. The compute you rely on is underpinned by a bet that AI demand keeps rising. If you are building something serious with AI, stay flexible: know which providers you could switch to, and avoid building in a way that permanently locks you to one platform.