Nvidia’s $100 Billion OpenAI Bet Mostly Fuels Its Own Chips

​A Monumental Deal With a Twist
Nvidia’s headline-grabbing $100 billion investment in OpenAI isn’t just about equity. Most of that cash will circle right back to Nvidia as OpenAI leases its powerful GPUs instead of buying them outright.

Leasing, Not Owning
Each AI data center could cost $50 billion, with about $35 billion earmarked for Nvidia’s chips. By leasing rather than purchasing, OpenAI spreads the cost over five years, while Nvidia still gets paid — and takes on some risk if the gear doesn’t live up to expectations.

Early Tranches, Big Ambitions
The first $10 billion tranche is coming soon to jumpstart OpenAI’s initial gigawatt-scale data center. While some of the equity may fund hiring or operations, insiders say the bulk will go to compute power — meaning GPUs.

Oracle and Other Partners Step In
At its new Texas data center, OpenAI CFO Sarah Friar pointed to partners like Oracle, which is leasing the facility, alongside Nvidia’s chip equity. She stressed that demand for compute far outweighs current supply, making these partnerships critical.

Bubble Fears Linger
Skeptics warn of a financial loop: OpenAI raises money from Nvidia, spends it leasing Nvidia chips, and in turn boosts Nvidia’s earnings. Critics call it “circular,” while OpenAI CEO Sam Altman insists real customer demand will pay for the chips in the long run.

The Bottom Line
​It’s a bold symbiosis: Nvidia fuels OpenAI with cash and chips, while OpenAI helps cement Nvidia’s dominance. But whether this cycle reflects unstoppable AI demand or just hype-fed recycling is the trillion-dollar question.