SpaceX recently adjusted the usage quota for the Grok Bot, significantly reducing user allowances. Musk responded that the operational costs are too high to sustain.Author and source: AIBase
Recently, multiple users reported on social media that SpaceX quietly adjusted the usage quotas for its popular Grok Bot over the weekend, reducing many users’ allowances by half. In response to user complaints and questions, Elon Musk publicly replied on X, acknowledging that “the cost of running Grok Bot is too high relative to its revenue to be sustainable.” Although he did not explicitly admit to implementing cuts, his statement clearly underscores the real-world pressures AI agents face between commercialization and soaring computational costs.
From explosive growth to encountering quota restrictions
The Grok Bot, an advanced feature capable of browsing the web, managing emails, and logging into various accounts on behalf of users, shares the same subscription quota as Grok itself and the popular code editor Cursor. Just two weeks ago, Musk publicly referred to it as his company’s “fastest-growing star product”; by early October, SpaceX President Gwynne Shotwell released impressive data showing that weekly paying users had reached 80,000, more than doubling from the previous month.
However, as user numbers and high-frequency requests surged dramatically, the issue of computational power consumption came to light. During this quota controversy, some users reported that after exhausting their quotas, they had to wait several days before new quotas were refreshed. In response to user complaints, Musk even asked: “Isn’t a higher-tier subscription too expensive for you?”
Industry under collective pressure: High costs of AI agents become widespread issue
In fact, SpaceX is not the only company facing the challenge of an imbalance between computing costs and subscription models. Over the past period, the entire AI industry has been undergoing a wave of concentrated adjustments to tighten affordable plans:
OpenAI has announced that it will begin merging Codex usage with standard chat quotas starting in December and will take strict action against frequent cross-account exploitation.
Anthropic also noted in its calculations that the computational cost required for an AI agent to independently complete complex tasks is often 4 to 15 times higher than that of a simple conversation.
As AI agents from major tech giants rapidly move to the forefront, how to manage persistently high and far-above-normal-dialogue backend computing costs while ensuring user experience has become a serious challenge for all large model providers.

