Ranjan Roy: SpaceX’s partnership with Anthropic boosts AI compute capabilities, growing skepticism about tech transformation, and the crucial need for model efficiency | Big Technology

Elon Musk’s SpaceX just handed Anthropic a fully built $5-6 billion data center, instantly easing the compute shortage that had been “hurting them recently,” in the words of podcast regular Ranjan Roy. The partnership is as shocking as it is revealing: the same Musk who founded xAI is now renting out capacity to one of OpenAI’s fiercest rivals.

Roy, founder of Margins, called the arrangement “both shocking and mind-blowing.” For Anthropic the deal is a game-changer, dramatically expanding its ability to train and run models. For SpaceX it marks an admission that the rocket company “doesn’t know how to effectively use this compute capacity to train models,” Roy noted, turning an internal asset into a new revenue stream ahead of a possible IPO.

Yet even as partnerships like this intensify the race against OpenAI, Roy’s bigger message is sobering. After watching AI companies post growth rates that “far exceed initial expectations,” he has grown “definitely more skeptical in the last two weeks” about whether the promised enterprise transformation will actually materialize. Dario Amodei, Anthropic’s CEO, recently revealed the firm had budgeted for growing “about 10 times as big this year”; Roy hopes the 80-times pace “doesn’t continue because that’s just crazy and it’s too hard to handle.”