SpaceX launched Grok Bot in August as an agent designed to use a virtual computer for tasks such as shopping, selling access through plans ranging from free to $200 a month. Musk has portrayed the product as a major breakthrough, calling it the fastest-growing product he had ever seen.¹

That pitch continued as he promoted its practical reach. In one post, Musk said simply that “Grok @Bot can do complex trips,” after a user described using it to arrange a family flight itinerary.Elon Musk (elonmusk)² Other posts highlighted the agent negotiating a lower internet bill, underscoring the appeal: a tool that can act, not merely answer.

Then came the constraint. Over Halloween weekend, users said their weekly allowances had suddenly run out — in some cases jumping from partial usage to fully exhausted. Musk confirmed it was not a glitch: “The cost of serving Grok Bot was unsustainably high relative to revenue.”³ His response casts tighter limits as basic business discipline, even as customers read them as a downgrade in a service they were already paying for.

The episode is awkward for a company that has celebrated Grok Bot’s momentum. SpaceX reported 800,000 paid weekly users, but its own executive’s admission shows that growth does not settle the harder question of whether agents can be profitably offered at consumer prices.¹

That problem extends beyond Musk. OpenAI consumer chief Adam Fry has described the company’s Dots assistants as “on the compute heavy side”; each has its own cloud computer, and access remains limited to $100-a-month Pro subscribers. OpenAI is now routing tasks across models in an effort to cut demand.⁴

There is a counterargument: expensive agents can still earn their keep. In a September procurement example, xAI said Grok Bot identified 43 inactive software seats, producing $14,220 in savings after review and downgrades.⁵ But the Grok limits make the divide plain: value to a user and sustainable margins for a provider are not the same calculation.