OpenAI CEO Sam Altman's 2033 250 GW target could cost up to $12.5 trillion

OpenAI is about to test how far ambition can go before it melts the servers. Analysts at Citi said this week that the company will need to spend more than $1 trillion over the next five years to meet the insane level of computing power it has promised to build.

The number is based on OpenAI’s partnerships with Nvidia, Broadcom, and AMD, which together aim to deliver a jaw‑dropping 26 gigawatts of compute capacity, roughly equal to powering the entire state of New York during a summer heatwave.

According to Citi analyst Chris Danely, it costs about $50 billion to bring one gigawatt of compute online. Multiply that by 26, and OpenAI’s bill rockets to about $1.3 trillion by 2030.

The figure doesn’t even include the wilder plan that CEO Sam Altman has been floating behind closed doors. In late September, The Information reported that Sam has talked about ramping up to 250 gigawatts of compute capacity by 2033, a scale that would swallow about $12.5 trillion in spending if ever attempted.

Analysts warn of funding gap and AI market bubble

The problem isn’t just the price tag. Citi estimates that even as OpenAI’s costs climb above a trillion dollars, its revenue might only reach $163 billion by 2030, which is only a fraction of what’s needed to cover that mountain of spending.

That mismatch has made Wall Street nervous, especially as AI stocks keep pushing to new records this year. Investors are betting big on artificial intelligence, but analysts warn the hype could be building another market bubble.

To make matters worse, it’s not clear if OpenAI even has the capital or access to enough power infrastructure to turn all this into reality.

The United States’ electrical grid is already under pressure, and analysts say scaling it up fast enough to power AI data centers could take years.

If that doesn’t happen in time, OpenAI could find itself with data centers ready to build but nowhere near enough juice to run them.

Massive deals and ripple effects across the chip world

OpenAI has already been writing massive checks to secure its place in the global AI build‑out. In September, it signed a $300 billion deal with Oracle for its Stargate project, a 10‑gigawatt AI infrastructure program in the U.S. meant to anchor OpenAI’s next generation of systems.

The company also announced Stargate projects in the United Arab Emirates and Norway, both built in collaboration with Nvidia. On top of that, OpenAI committed $22 billion to purchase compute capacity from CoreWeave, an AI data center company backed by Nvidia.

The spending spree is so massive that even industry veterans are nervous about how interlinked everything has become.

Bernstein analyst Stacy Rasgon wrote on October 6 that Sam Altman “has the power to crash the global economy for a decade or take us all to the promised land,” describing how deeply AI investments have sunk into financial markets.

Despite the risks, chipmakers stand to gain the most if OpenAI manages to fund its plans. Bank of America analyst Vivek Arya said Nvidia could make as much as $500 billion in revenue from OpenAI if the entire deal is executed. Broadcom could walk away with more than $100 billion, according to Rasgon’s estimate.

Join a premium crypto trading community free for 30 days - normally $100/mo.