Reid Hoffman: SpaceX Not AI, xAI a 'Complete Train Wreck'

Reid Hoffman holds stakes in both OpenAI and Anthropic through Greylock Partners, so he is not a neutral observer of the AI race. That positioning is precisely what makes his recent comments on the Pioneers of AI podcast, covered by Fortune, worth parsing carefully.

On xAI, Hoffman was direct: "a complete train wreck." He noted that all of the company's founders have departed and that it is on its third restart, framing these as signs of fundamental instability in the foundational model effort. His assessment of SpaceX was sharper: Hoffman called it "not an AI company," describing its compute infrastructure business as "a premium-priced CoreWeave." The implicit argument is that owning GPU clusters is not the same as doing the model-building work that earns the AI label.

On the competition between OpenAI and Anthropic, Hoffman rejected the zero-sum framing. "There's a lot of room for both of them to win incredibly," he said, arguing the two companies occupy distinct enough positions that both can succeed substantially. The investor's interest in this framing is obvious, but the underlying market-segmentation argument is worth taking seriously on its own terms.

Hoffman also addressed the U.S. government's order forcing Anthropic to pull its Fable and Mythos models. He called the approach "autocratic willy-nilly" and "very sub-optimum," arguing it lacked principled reasoning: "It doesn't look like there's anything that's a particular principled, here's-the-way-that-we're-navigating-through-things... It's more like, 'Hey, we kind of had some contentious interactions with this company anyway, so we're going to hit them with a stick.'" He acknowledged there may be a legitimate cybersecurity basis for concern while arguing the execution was damaging regardless.

The honest caveat is that Hoffman's criticism of regulatory unpredictability is also, in effect, a defense of companies he is invested in. What the reporting does not give you is a clear picture of what a principled regulatory alternative would look like, or how AI investors should actually price the kind of risk Hoffman describes. If ad hoc model bans become normalized, the chilling effect on investment will reach across the whole sector.