Musk says Grok Bot costs are too high to sustain

The cost of AI agents

Elon Musk recently addressed concerns regarding the high operational expenses associated with Grok Bot. He noted that the costs of providing the service were unsustainably high relative to the revenue the company is currently generating. This admission came in response to user complaints about usage limits being reached too quickly.

Grok Bot differs from a standard chat interface because it functions as an AI agent. Instead of just answering questions, it operates a persistent PC in the cloud equipped with a browser, a file system, and a terminal. This infrastructure allows the bot to carry out complex tasks and run scheduled jobs in the background even after a user closes their app.

Because the bot performs active work in a cloud environment, it requires significantly more resources than a typical text-based AI. As XenoSpectrum reported, the imbalance between these high operating costs and the subscription revenue is the primary driver behind the recent adjustments to user limits.

How usage is measured

Determining how much of a weekly allowance remains can be difficult for users of Grok Bot. Unlike standard chat models where the count of questions and answers is the primary metric, Grok Bot's consumption varies based on the effort required for a specific task. Different background tasks will not necessarily consume the same amount of usage even if the number of requests is identical.

The system also automatically selects which AI model to use based on the requirements of the task at hand. This selection process may include models from various contracted providers, and the specific model used for a request is not displayed to the user. Despite the variety of models, the per-token price applied to the user remains the same regardless of which model the bot chooses internally.

Users must also distinguish between their weekly allowance and their monthly usage cap. Raising the monthly cap for extra charges does not increase the weekly allowance included in a subscription plan. This distinction is important for users trying to estimate how much work they can complete before hitting a limit or incurring additional fees.

Impact on different plans

The way usage limits and billing work depends heavily on whether a user is on a legacy plan or a unified plan. Legacy plans apply a weekly allowance specifically dedicated to Grok Bot. In contrast, the new unified plans share a single weekly allowance across multiple products, including Grok and Cursor.

New users start on these unified plans, where the Premium, Super, and Ultra tiers include Grok Bot access. Users on these plans must monitor their shared allowance carefully, as activity in other bundled products will also deplete the amount available for the bot. This means a drop in the remaining allowance might not be caused by the bot alone.

For those on legacy plans, the ability to continue work after the limit is reached depends on whether pay-as-you-go billing is enabled. If it is turned off, work stops immediately once the weekly limit is hit. If it is on, users can continue but will face extra charges for the overage, which can be managed through specific settings in the web or mobile versions.

Managing costs and limits

Users who want to maintain continuous work without interruption may need to accept higher costs. On legacy plans, enabling pay-as-you-go billing allows the bot to keep running, but it will draw from a monthly limit. However, the documentation notes that a bot already running may complete its work beyond a set monthly limit, meaning a cap does not guarantee a hard stop on charges.

Under the unified plan, the system uses a different mechanism for additional costs. When in-house models exceed the shared weekly allowance, charges are paid from additional credits. This is distinct from the pay-as-you-go billing used on legacy plans, and users need to check these payment methods separately to avoid unexpected expenses.

Because the specific amounts of the cost reductions or the exact new limits have not been confirmed, users should monitor their accounts closely. It remains unclear which specific plans were most affected by the recent changes. Users are encouraged to check their specific plan terms to understand how many more jobs they can complete from their remaining percentage.

What to watch next

The primary thing to watch is how the company balances the high cost of cloud-based agents with the need for a profitable subscription model. If the costs remain unsustainably high, further reductions in weekly allowances or increases in subscription prices may occur. The company has yet to provide specific figures on the exact cost per task or the total revenue generated by the bot.

Another key area to monitor is the transparency of the automatic model selection process. Since users cannot see which model is being used for a specific task, it will be important to see if the company provides more clarity on how these choices impact the consumption of the weekly allowance. This transparency is vital for users trying to budget their usage for complex projects.

Finally, the transition from legacy plans to unified plans will continue to affect how different user bases experience these limits. As more users migrate to the unified plans, the shared nature of the allowance across multiple products will become the standard experience. This shift may lead to more complex usage tracking for power users who rely on multiple tools simultaneously.

Next steps

Check your current subscription type to see if you are on a legacy plan or a unified plan, as this determines how your weekly allowance is shared and billed.