OpenAI president Greg Brockman told a federal court in Oakland this week that the company will spend roughly $50 billion on compute in 2026, a sworn disclosure that puts hard numbers on the most expensive infrastructure build-out in the history of software. Brockman gave the figure during his testimony in the Musk-Altman trial, where he also confirmed that his personal stake in the company is now worth about $30 billion.
The compute number is the headline. He said that OpenAI is gonna spend about $50 billion on compute this year. That is, I mean, just compute. That is wild, Jaeden Schafer said on the podcast, framing the figure as a step-change rather than a steady climb.
The trajectory underscores how quickly the economics of frontier AI have escalated. "The number was about 30 million in 2017, and now we're going to 50 billion in 2026," Schafer said. That is a roughly 1,600-fold increase in nine years, and it follows a 2025 bill that was already in the tens of billions.
Key facts
- 01Greg Brockman testified under oath in the Musk-Altman trial in Oakland that OpenAI will spend roughly $50 billion on compute in 2026.
- 02Brockman's personal stake in OpenAI is now worth about $30 billion, according to his court testimony.
- 03OpenAI's compute spend has climbed from about $30 million in 2017 to tens of billions last year and $50 billion this year.
- 04OpenAI has previously disclosed a $600 billion spend pledge through 2030, with total commitments running over a trillion dollars.
The court disclosure fills in a gap in OpenAI's previously announced spending plans. The company has flagged a $600 billion spend pledge through 2030 and total commitments north of a trillion dollars, but had not broken out the in-year figure for 2026. Brockman's testimony, delivered under oath, is now the cleanest public marker of how fast that capital is flowing out the door.
The other side of the ledger is what makes the number matter beyond OpenAI. Samsung crossed a trillion-dollar market capitalisation on the same day Brockman testified, with profit up roughly eight-fold on the back of AI memory demand. Apple is separately in talks with US chip manufacturers, a sign that the scramble for supply is widening past the obvious GPU vendors.
Schafer argued that OpenAI's bill is increasingly someone else's revenue line. "OpenAI is gonna spend this $50 billion. It's going to companies like Samsung that are gobbling up that money to help them add memory into data centers," he said on the show. The memory tier, long a commodity business with thin margins, has become one of the biggest beneficiaries of the model-training arms race.
That dynamic is reshaping how investors read AI capex announcements. Each new spending pledge from OpenAI, Microsoft or Meta now flows through to balance sheets at memory makers, foundries, networking vendors and power suppliers, turning what looks like a cost line for one company into a growth thesis for several others.
It also raises the stakes for OpenAI's revenue ramp. A $50 billion compute year implies the company needs an enterprise and consumer business large enough to service that burn without leaning indefinitely on outside capital. Brockman's testimony was meant to address questions about OpenAI's structure and Elon Musk's claims, but the numbers he put on the record may end up mattering more to Wall Street than to the judge.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
