ENN’s last update in this thread covered Elon Musk’s 18 November post that Grok Bot can do your accounting (read it here)1. Before that, our report on the coming general release also noted his pitch that the agent can save companies millions in corporate expenses2. The new development is on the other side of the ledger. On 18 November, independent coverage returned to Musk’s admission that running Grok has been costing more than it brings in3.
Taken together, this is a revealing week for xAI’s AI teammate. Musk is selling Grok Bot as a tool that finds savings for its users. Meanwhile, the press is looking at what it costs to keep a fast-growing agent running for paying subscribers.
What happened
TechShots reported on 18 November that Musk has said operating xAI’s Grok became financially unsustainable, with running costs heavily outpacing revenue3. According to the outlet, he was responding to complaints from paying subscribers about strict usage limits and quotas that ran out quickly3. TechShots described the issue as the economic pressure of scaling generative AI infrastructure to meet surging user demand3. Its piece points readers to a fuller story at CIO Economic Times3.
The admission itself goes back to the start of the month. A LinkedIn post dated 2 November shared the headline “Elon Musk admits Grok Bot costs too much to run as users hit weekly caps”4, which shows the story was circulating within a day of Musk’s reply.

What Musk said
Rocky Mountain Madman gives the clearest account. It reports that over Halloween weekend, some Grok Bot users found their weekly usage limits maxed out even though they had not been using the agent overnight5. Users on X reported usage jumping from as low as 50 percent to 100 percent5. On Sunday, 1 November, Musk confirmed in a reply on X that this was not a glitch5.
Citing TechCrunch’s 2 November report, the outlet quotes Musk’s explanation:
“The cost of serving Grok Bot was unsustainably high relative to revenue,”
That quote comes from Rocky Mountain Madman’s text, which credits it to TechCrunch5. It is a frank statement. Musk named the problem in public and did not leave paying users guessing about whether a bug had eaten their allowance. The outlet also noted that he had not yet said how subscribers’ limits would change from there5.
The numbers behind the story
The coverage describes a product with a demand problem, not a product in trouble. Rocky Mountain Madman writes that “Grok Bot is not a failing product”5. It reports these figures:
- SpaceX said in a filing last month that AI now makes up 52 percent of company income5.
- Its renamed Super Intelligence division turned a profit of 431 million dollars, which the outlet calls a rarity in this industry5.
- That profit would have been higher without 1.2 billion dollars of one-time costs tied to the acquisition of Cursor5.
- Musk has called Grok Bot the fastest-growing product he has ever seen5.
For an AI business, a profitable division with a product people use so heavily that it strains its own allowances is a strong place to be. In the outlet’s account, the tighter limits were applied to a successful product5.
Why agents cost more to serve
Rocky Mountain Madman explains why this happens. Agentic usage scales costs faster than subscription revenue, it argues, because an agent working overnight burns inference the whole time, whether or not the user ever sees the bill5. With a flat-rate agent product, it says, every extra tool call, retry and long-running task becomes the provider’s cost and not the customer’s5.
The outlet lists the usual defenses: meter by task rather than by seat, cache aggressively, send routine work to smaller models, and put hard per-task budgets into the code5. It also points out that the pressure is not unique to Grok Bot. It cites OpenAI’s reported move to draw ChatGPT and Codex from one shared token pool as “the same pressure from the other side”5. Its conclusion is that when a profitable division at SpaceX’s scale pulls back limits in public, smaller builders should assume their margins are thinner than they look5.

How it fits this week’s pitch
The timing gives the story extra interest. On the same day as the TechShots piece, Musk posted that Grok Bot “can save you millions in corporate expenses!”2 He wrote that it “will research every expenditure, evaluate its necessity and find better deals on the Internet or by calling suppliers on the phone directly”2. He followed that with a shorter post: “Grok @Bot can do your accounting”1.
These are Musk’s descriptions of what the agent can do. The sources do not include independent tests of those savings, so ENN reports them as his pitch. Still, the two threads fit together. The agent he describes, one that researches expenses and phones suppliers, is the kind of long-running, multi-step work that Rocky Mountain Madman says drives serving costs5. A product capable of that much work is also one users will lean on hard, which matches the demand TechShots describes3.
Independent coverage and the critical view
The coverage is not all praise. TechShots’ headline, “Running on Empty,” frames the cost admission sharply, and its text calls Grok’s operation financially unsustainable3. Rocky Mountain Madman notes that the throttle left paying users with a product whose allowance can be revised downward without warning5. A LinkedIn commenter sharing the story argued that “the future of personal AI bot/actors is local”4.
Those are fair points to record. The same reporting also supplies the wider context: a division that turned a profit, a product Musk calls the fastest-growing he has seen, and a candid public explanation from the top instead of silence5. The pressure it describes, AI products being priced on what they cost to serve, affects the whole industry and not only xAI5.
What to watch next
The open question is the one Rocky Mountain Madman flagged: how subscriber limits will change5. None of the sources here show a new limit, price or plan. ENN’s previous coverage reported that Musk expected general release within a week or two after a soft rollout. When that release comes, watch whether it brings new allowances or task-based pricing of the kind the coverage describes. Also watch whether Musk backs his savings pitch with examples of Grok Bot cutting real company bills2. ENN will update this thread when he does.



