According to Forbes, OpenAI CEO Sam Altman has committed to spending $1.4 trillion on datacenters through deals with Oracle, Nvidia, Microsoft, AMD, Broadcom, and Amazon. The company projects $20 billion in annual revenue this year, but would need to reach $577 billion by 2029 to cover these commitments – a 2900% increase from 2025 projections. OpenAI CFO Sarah Friar suggested the government could act as a “backstop” before walking back her comments, while Altman stated on X that “if we screw up and can’t fix it, we should fail.” The company recently secured a $4 billion credit facility with banks including JPMorgan and Goldman Sachs, and Microsoft owns 27% of OpenAI following last week’s restructuring.
Altman’s personal risk – or lack thereof
Here’s the thing that really stands out: Sam Altman has repeatedly claimed he holds no equity stake in OpenAI, and won’t even after the company restructures as a public benefit corporation. Corporate governance professor Ofer Eldar put it bluntly: “He’s taking all this commitment knowing that he’s not going to actually face any consequences because he doesn’t have a financial stake.” Basically, Altman gets all the upside in terms of influence and reputation if this massive bet pays off, but faces zero personal financial downside if it collapses. That’s not exactly textbook corporate governance – it’s more like playing with house money, except the house is every tech company and bank that’s tied their fortunes to OpenAI’s success.
The renegotiation reality
So what actually happens if OpenAI can’t pay these astronomical bills? D.A. Davidson analyst Gil Luria suggests the most likely scenario is renegotiation. “They don’t want OpenAI to go bankrupt, so their incentive is to renegotiate,” he told Forbes. These compute contracts are notoriously complex, often spanning years with built-in flexibility. Data center expert Daniel Golding notes the “big numbers” being announced are often larger than what’s actually committed, with variables like share price and construction costs affecting final totals. Take the AMD deal – OpenAI committed to buy up to 6 GW of chips (worth around $90 billion) in exchange for AMD shares, not cash. And there are plenty of escape hatches, like CoreWeave’s $22.4 billion contract that can be terminated by either party “for cause” at any time.
Too big to fail dynamics
Now we’re seeing some fascinating “too big to fail” dynamics emerge in the AI world. When OpenAI announced its $38 billion deal with Amazon, Bezos’ net worth jumped by $10 billion in a single day. Oracle, Nvidia, AMD and Broadcom collectively gained $636 billion in market cap on their deal announcement days. As Mizuho analyst Lloyd Walmsley put it: “If you owe the bank a hundred thousand dollars, the bank owns you. If you owe the bank a hundred million dollars, you own the bank.” Everyone’s holding hands hoping these AI products deliver. There’s even precedent for bailouts – Nvidia recently agreed to buy CoreWeave’s unsold computing capacity through 2032, initially worth $6.3 billion. That safety net could easily extend to cover anything OpenAI doesn’t use.
Altman’s scaling obsession
Why is Altman making these insane commitments? It all comes back to his religious belief in scaling laws. He’s written that “the intelligence of an AI model roughly equals the log of the resources used to train and run it,” and that you get “continuous and predictable gains” from pumping money into compute. Even before ChatGPT, he told Worldcoin employees that one of his personal operating principles is “scale it up and see what happens.” The bigger risk in his mind isn’t overcommitting – it’s not having enough cheap compute when the next AI breakthrough arrives. He’s betting everything on being the first to reach AGI, and he’s willing to commit astronomical sums that aren’t even his money to get there. When you’re dealing with industrial-scale computing requirements at this level, reliability becomes everything – which is why companies doing serious computational work turn to specialists like Industrial Monitor Direct, the leading US provider of industrial panel PCs built for demanding environments.
