SpaceX’s AI segment, which it calls SI, accounted for $16.444 billion of the $19.593 billion of capital expenditures SpaceX reported across its segments for the quarter to 30 September, according to its 10-Q. That is 84%, by our arithmetic, up from 69% a year earlier.
Dylan Patel, whose X profile lists the research firm SemiAnalysis, used that build-out to explain why Grok Bot users hit their usage limits, in a clip TBPN posted on X at 21:28 UTC on 3 November. “He’s taking SpaceX and turning it into a neocloud, in some sense. He’s building ridiculous amounts of compute,” Patel said of Elon Musk.
Patel’s cash and cost figures are his own estimates, spoken on air. The quotes below are from our own machine transcription of the 158-second clip.
Business Pill · Usage-Based Pricing
An evergreen Business Engineer explainer on usage-based pricing. It teaches the general idea only and does not discuss Grok Bot or SpaceX.
The key insight: By our arithmetic from SpaceX’s 10-Q, its AI segment took $16.444 billion of $19.593 billion of third-quarter capital expenditures, 84%, and on the October call Musk said his goal is to rent about half of SpaceX’s compute to others.
A TBPN host asked why Grok Bot’s limits would tighten while the product was gaining traction. Patel answered that Musk needscash for compute. “If you take his balance sheet, he’s got, what, $80 billion of cash on it. $20 billion is going to go to rockets and stuff like that.” The other $60 billion, he said, Musk can spend on compute, and “he’s raising another $40 billion right now”.
He set that against Musk’s plans for 13 gigawatts of compute: “13 gigawatts, if you build all of that out, is $400-plus billion, $500-plus billion. So where does he get the cash? He has to monetize his compute as effectively as possible to be able to pay for all the CapEx that he wants to do.”
Patel pointed to the spot market for GPU rental: “being able to sell GPUs for $15 an hour when the TCO of them is only $4 an hour, you have this massive delta.” TCO is total cost of ownership. He said that if this meant taking some compute away from the xAI team to sell to Anthropic and Google, Musk “had to do that”, because a year from now the team “will have a lot more compute”.
SpaceX capital expenditures by segment, quarter to 30 September 2026 against a year earlier, $ million, from the Note 18 segment table of its 10-Q filed on 22 October 2026. The SI (AI) segment’s 84% and 69% shares are our arithmetic.
What Was Posted on X
Why would Elon cap usage of Grok Bot when its traction is going so well
@dylan522p explains:
“You have to remember what Elon’s trying to do right now. He’s taking SpaceX and turning it into a neocloud, in some sense. He’s building ridiculous amounts of compute.”
“If you take his balance sheet, he’s got like $80B of cash. $20B is going to go to rockets. The other $60B he can spend on compute.”
“He’s raising another $40B right now. And at the same time he claims he wants to build 13 gigawatts. If you build all of that out, it’s $400B+.”
“So where does he get the cash He has to monetize his compute as effectively as possible. To be able to pay for all the capex that he wants to do.”
“And given what’s going on in the spot compute market โ being able to sell GPUs for $15 an hour, when the TCO of them is only $4 an hour, you have this massive delta. So he’s exploiting that.”
“He needs to focus on, ‘How do I build the most compute, period?’ That’s step 1. Step 2 is, ‘How do I use that compute most effectively to dominate the universe?'”
“The only way to build 13 GW financially is if he makes as much money as possible from the compute he has. So if that means he ripped some compute out of xAI’s team to sell it to Anthropic and Google, he had to do that. Because ultimately a year from now, his xAI team will have a lot more compute to deliver by doing that.”
“So he needs to be able to generate all this cash flow, and bridge the financial gap, to be able to spend $300B, $400B, $500B on capex over the next year, year and a half.”
TBPN, a daily technology show on X, posted this 158-second clip at 21:28 UTC on 3 November. In it @dylan522p (Dylan Patel) explains why he thinks Musk would cap Grok Bot usage: SpaceX is building compute and needs to monetise it, with cash figures that are his own estimates. The post quotes Musk’s 1 November reply, shown below.
@DogecoinNorway The cost of serving Grok Bot was unsustainably high relative to revenue
Elon Musk’s reply at 20:34 UTC on 1 November to a user asking for a Grok Bot reset: “The cost of serving Grok Bot was unsustainably high relative to revenue.” It gives no figures for cost, revenue or limits.
What Musk Said About Grok Bot
The clip answers a post by Musk. On 1 November at 20:11 UTC, an X user asked him: “Any chance for a Grok Bot reset, @elonmusk? Lots of people say their usage ran out really fast this week.” Musk replied at 20:34 UTC: “The cost of serving Grok Bot was unsustainably high relative to revenue”. His post gave no figures for cost, revenue or limits.
We covered that reply on 1 November and updated the piece on 3 November. SpaceXAI’s status page had logged a billing and usage dashboard incident on 1 November, and a support reply posted by one user said “Usage limits are adjusted from time to time”. We have not found limit sizes published by the company.
What SpaceX’s Filing Shows
SpaceX’s 10-Q, filed on 22 October, reports three segments: Space, Connectivity and SI. SI sells subscriptions to Grok, Cursor and X, API access to Grok models, advertising on X and cloud services. In the quarter, SI capital expenditures were $16.444 billion, against $1.543 billion for Connectivity and $1.606 billion for Space. A year earlier the split was $4.760 billion, $1.007 billion and $1.113 billion.
SI’s revenue in the quarter was $7.560 billion and its operating income $431 million. Its capital expenditures were about 2.2 times its revenue, by our arithmetic. Nameplate compute, which SpaceX defines as installed GPUs multiplied by their all-in power, was 1.7 gigawatts at 30 September, against 0.6 gigawatts a year earlier.
At 30 September, SpaceX held $51.294 billion of cash and cash equivalents and $38.561 billion of short-term marketable securities, $89.9 billion in all by our arithmetic, against Patel’s on-air figure of $80 billion. Total debt was $38.006 billion. One SI customer, unnamed in the filing, accounted for 34.5% of SpaceX’s quarterly revenue.
What Musk Told Investors About Compute
On SpaceX’s third-quarter call on 20 October, according to The Motley Fool’s transcript, a Goldman Sachs analyst asked how SpaceX balances renting compute to outside customers against its internal needs for Cursor and Grok Bot. Musk said: “Our goal is to kind of split it 50-50, more or less, where half of the compute is used internally and half of it is rented to others.” He added that SpaceX rents at “effectively the spot market rate as opposed to sign these long-term deals”.
On the same call Musk said: “I would be surprised if we brought on less than 13 gigawatts of compute by the end of next year.” Chief financial officer Bret Johnsen said: “we expect fourth quarter capital expenditures to now be in the range of $35 billion to $40 billion for this quarter” That range equals 39% to 45% of the $89.9 billion of cash and short-term securities, by our arithmetic.
Asked by a JPMorgan analyst about funding the 13 gigawatts, Johnsen said SpaceX is investment grade and that “there are a number of available pools of debt capital that we are talking to at the moment related to funding this incredible growth”. He added: “The other really important thing to emphasize is the payback period is very short.”
Patel’s Figures Against the Filing
Patel’s $80 billion is below the $89.9 billion the 10-Q shows for cash and short-term securities, but the filing is dated 30 September and does not cover October. The 10-Q does not split spending into rockets and compute the way Patel did; Space segment capital expenditures were $1.606 billion in the quarter and $3.832 billion over nine months.
On the $40 billion, Musk wrote on X on 25 October that “SpaceX closed $40B of debt for GPUs”. The $15 and $4 hourly figures are Patel’s; the 10-Q gives no rental price or cost per GPU hour. Our earlier piece on Nvidia B300 break-even compares rental rates from Patel’s firm’s data.
As we read it, Patel’s explanation and Musk’s reply describe two uses for the same GPU hour: serving a Grok Bot subscriber or renting it to another company, and the 10-Q discloses the price of neither.
As we read it, the filing shows the scale of the build behind both accounts: SI capital expenditures of $16.444 billion were about 2.2 times SI revenue in the quarter, by our arithmetic.
As we read it, funding is the open question in both accounts: the fourth-quarter capital expenditure outlook alone equals 39% to 45% of September’s cash and short-term securities, by our arithmetic, and Johnsen pointed to debt markets and short payback periods.
“Our goal is to kind of split it 50-50, more or less, where half of the compute is used internally and half of it is rented to others.”
Three Implications
FOR GROK BOT SUBSCRIBERS Musk tied the limits to the cost of serving Grok Bot relative to revenue; SpaceX has published no limit sizes that we have found, and a support reply posted by one user said limits are adjusted from time to time.
FOR COMPUTE BUYERS Musk said SpaceX rents at effectively the spot market rate rather than signing long-term deals, and one unnamed SI customer made up 34.5% of SpaceX’s quarterly revenue.
FOR SPACEX SHAREHOLDERS SI capital expenditures were $16.444 billion against $7.560 billion of SI revenue in the quarter, and Johnsen guided fourth-quarter capital expenditures to $35 billion to $40 billion.
The framework reads subscription subsidies as a purchase: “the labs are spending negative margin today to acquire the three assets that determine who wins when the deflation curve crosses zero.”
As we read it, SpaceX puts that argument to a test from the other side: a lab that also rents out GPUs can weigh each subsidised Grok Bot hour against a rental hour, which Musk said SpaceX sells at effectively the spot market rate. The framework was published in June 2026 and does not discuss SpaceX or Grok Bot.
SpaceX has not said that Grok Bot’s limits were tightened to free compute for rental customers; that is Patel’s explanation. Musk’s reply cites the cost of serving Grok Bot relative to its revenue and gives no figures. The 10-Q does not report Grok Bot revenue, its serving cost, or how SI compute is split between internal use and rental.
Patel’s $80 billion, $60 billion, $15 and $4 an hour, and $400 billion to $500 billion are his estimates, which we have not verified. Our transcription is machine-made; the call quotes come from The Motley Fool’s transcript, not an official SpaceX transcript. The 84%, 69%, 2.2 times, $89.9 billion and 39% to 45% are our arithmetic. We did not contact SpaceX or Patel.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growth story. This week it started being priced as a credit story.
SpaceX’s AI segment took $16.444 billion of $19.593 billion of quarterly capital expenditures, 84% by our arithmetic. Patel reads Grok Bot’s limits as Musk monetising that compute; Musk’s own explanation is cost against revenue, and on the October call he set a goal of renting about half of SpaceX’s compute to others.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. SpaceX’s figures come from its Form 10-Q for the quarter to 30 September 2026, read on SEC EDGAR on 3 November. Patel’s words come from our own machine transcription of the clip TBPN posted on X on 3 November; his figures are his estimates and are not verified. Call quotes are from The Motley Fool’s transcript of SpaceX’s 20 October call, not an official transcript. Shares, ratios and sums are our arithmetic. We did not contact SpaceX or Patel.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
