OpenAI Cuts Spending Plans From $1.4 Trillion To $600 Billion: What It Means For The Capex Cycle

Last last week, CNBC and other outlets reported that OpenAI had "reset" spending expectations, and was now targeting roughly $600 billion in total compute spend through 2030, as the ChatGPT maker lays groundwork for an IPO that could value it at up to $1 trillion. This is especially notable because just a few months prior, in late 2025, OpenAI's Sam Altman had said OpenAI is committed to spending $1.4 trillion to develop 30 gigawatts of computing resources — enough to power roughly 25 million U.S. homes. So is the capex trade now over? 

Well, not so fast: as JPMorgan's Mark Schilsky wrote in an overnight note (available here to pro subs), the reality is that these two numbers are somewhat apples and oranges, to wit "The $1.4T figure refers to infrastructure and capacity over the next 8 years (through 2033), while the $600B figure that was reported by CNBC a few days ago refers to compute spend over the next 4-5 years (through 2030). Apples and oranges." More from the report: