Elon Musk confirmed on Sunday, November 1, that Grok Bot users are hitting their usage limits sooner because the cost of serving the agent is too high. Replying to an X user whose allowance had run out four days early, he wrote that the cost of serving Grok Bot was “unsustainably high relative to revenue.” TechCrunch reports that over the Halloween weekend some users found their weekly limits maxed out even without overnight use, and that Musk confirmed it wasn’t a glitch.
What users saw. On Cursor’s community forum, users said their weekly meter flipped between 77% and 100%, and Cursor reported a problem with billing and its usage dashboard on Sunday, according to Forbes. That matters because xAI’s subscription documentation says Grok and Cursor draw from a shared weekly allowance under its unified plans.
The backdrop. On SpaceX’s Q3 earnings call on October 20, Musk said Grok Bot was seeing the fastest growth of any product he had ever seen, and President Gwynne Shotwell said it reached 800,000 paid weekly users earlier that month, double September’s level. Musk also said usage per user was growing over time, not shrinking. One reading, offered by commentators, is that heavy users on flat monthly plans raise serving costs without paying more. The sources don’t confirm that explanation.
It isn’t a statement about SpaceX’s whole AI business. TechCrunch reports that SpaceX’s filing says AI makes up 52% of company income and that its renamed Super Intelligence division turned a $431 million profit. Musk’s comment concerns the cost of serving Grok Bot relative to its revenue, not the division as a whole.
What isn’t known. Musk’s posts didn’t say how much Grok Bot costs to run or whether xAI has formally lowered the limits. The reporting doesn’t say which subscription plans, regions, or tasks are affected, whether every limit has changed, when earlier allowances might return, or whether xAI plans to change prices or add higher-capacity tiers. A former xAI employee publicly questioned the shift, asking how a success story two weeks earlier could now be unsustainable.
A contrast worth noting. OpenAI doubled its Work and Codex allowances on several Pro plans days earlier and reversed a planned Pro 200 cut. The two companies’ choices show how differently flat-rate agent plans can be managed, though neither has published its costs.
Bottom line: Musk attributes the tighter caps to serving costs outrunning revenue, and the usage meters changed before any formal announcement. For anyone paying for an agent plan, the takeaway is that the stated allowance is a policy that can change, so watch your meter and check whether your tool shares a pool with another product.
