Elon Musk is pitching SpaceX’s next frontier not as a rocket or a satellite, but an AI system shaped by the people who build them. His promise is sweeping: Grok would absorb the company’s knowledge — even as the details of what employee information is involved remain unclear.

At a SpaceX all-hands meeting posted Tuesday, Musk told staff the company would train Grok on “the sum total of all SpaceX information.” “So in a way, it will be trained on you,” he said, casting employees as the people who would pass their ideas and beliefs to the model.¹

Musk framed that as an advantage, arguing that SpaceX staff — whom he called an exceptional collection of people — could help align increasingly powerful AI with human goals. He also urged workers to use the company’s AI tools and help improve them. But the announcement leaves the practical boundary unresolved: it is not known which employee data, communications or work products SpaceX plans to use, or how those materials would be handled.

The proposal arrives as AI firms seek more specialized data after consuming much of the public internet’s readily available material. Employee-generated information is particularly valuable for AI agents designed to navigate computers and perform real-world tasks — the kind of work SpaceX’s newly launched Grok Bot is intended to do. Similar corporate data-collection efforts have already drawn staff backlash elsewhere.

Musk paired the workforce-as-training-data pitch with an even larger commercial prediction. Calling AI “an extremely important part of SpaceX’s future,” he said its AI revenue would “definitely” exceed all other company revenue as soon as September and widen that lead in the fourth quarter.²

That would mark a dramatic shift for a company best known for launches, spacecraft and Starlink. Musk’s vision is that SpaceX’s human expertise will train the machine — and that the machine will quickly become the business’s biggest engine.