Grok Bot meters jump to 100%, and Musk says the bots cost too much to serve
The flat-rate agent plan had a bad weekend. Starting early on November 1 UTC, Grok Bot users reported their usage meters jumping, often from around 55% straight to 100%, with every bot idle. One user said $10 of on-demand credit disappeared after six exchanges. Another said the in-app meter had been under-reporting against their own tracking and then "jumped to match". xAI's status page logged a five-hour service degradation on the billing and usage dashboard, and later marked it resolved, while some accounts stayed stuck at usage_exhausted and support quoted three to four days for a human reply.
Then Elon Musk replied to the complaints at 20:34 UTC: "The cost of serving Grok Bot was unsustainably high relative to revenue."
xAI has not published a limits change, so calling this a deliberate cut is an inference. But Musk's line says out loud what the meter behavior suggests: bots that run all day eat far more compute than the subscription was priced for, and the fix lands on the user's meter.
Put it next to last week. OpenAI doubled Codex limits, and Anthropic dropped the five-hour window for its top Claude tier. Those were moves to win always-on agent users. xAI is the first of the three to say the economics don't work. Always-on agents don't fit flat-rate pricing, and every lab will have to show which way it's going.
Musk's reply: https://x.com/elonmusk/status/2116992258520064397 xAI status: https://status.x.ai
← Back to all articles
Then Elon Musk replied to the complaints at 20:34 UTC: "The cost of serving Grok Bot was unsustainably high relative to revenue."
xAI has not published a limits change, so calling this a deliberate cut is an inference. But Musk's line says out loud what the meter behavior suggests: bots that run all day eat far more compute than the subscription was priced for, and the fix lands on the user's meter.
Put it next to last week. OpenAI doubled Codex limits, and Anthropic dropped the five-hour window for its top Claude tier. Those were moves to win always-on agent users. xAI is the first of the three to say the economics don't work. Always-on agents don't fit flat-rate pricing, and every lab will have to show which way it's going.
Musk's reply: https://x.com/elonmusk/status/2116992258520064397 xAI status: https://status.x.ai
Comments