SpaceX’s AI business is on track to be able to bring 200,000 of Nvidia’s Grace Blackwell systems online “every month for the remainder of this year” in its Memphis-area data centers, CNBC reported from the company’s third-quarter earnings call on 20 October 2026. Elon Musk said on the call that SpaceX aims to “deploy power, cooling infrastructure at a rate that is vastly faster than anyone else,” according to CNBC.
Musk also said SpaceX would attempt to catch Starship and its booster next month, The Verge reported. Below is what was said on the call, set against the results release SpaceX filed with the SEC the same day and against what the company said on its August call.
Business Pill · CAPEX AND DEPRECIATION
A short explainer on capex and depreciation: how money spent on long-lived assets such as data centres turns into a cost over years. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, the call put the bottleneck on energising chips rather than buying them: CNBC reported SpaceX is on track to be able to bring 200,000 Nvidia Grace Blackwell systems online each month for the rest of 2026, while nameplate compute stood at 1.7 gigawatts on 30 September, according to the release.
CNBC’s live coverage, by Lora Kolodny, reported both the 200,000-a-month figure and Musk’s line on power and cooling. The same coverage noted that of SpaceX’s $19.6 billion of capital expenditure in the quarter, more than $16.4 billion was committed to the AI business.
The results release puts numbers on where that buildout stood at the end of September. Nameplate compute was 1.7 gigawatts on 30 September, up from 1.4 gigawatts at the end of June and 0.6 gigawatts a year earlier, as SpaceX continued the buildout of COLOSSUS II, according to the release. Capital expenditure was $19,593 million, of which $16,444 million was in the SI segment.
SpaceX nameplate compute, from its third-quarter 2026 results release (Form 8-K exhibit 99.1, 20 October 2026): 1.7 gigawatts on 30 September 2026, against 1.4 gigawatts on 30 June and 0.6 gigawatts a year earlier. The top bar is the year-end aim Elon Musk gave on the August 2026 call (“over 2 gigawatts of compute”). The 200,000-a-month figure is from CNBC’s live coverage of the 20 October call.
What Was Posted on X
Here’s full Elon Musk’s remarks at SpaceX Q3 2026 results:
0:00 – Inflection moment for SpaceX 0:50 – Starship Flight 14 to orbit 1:40 – Flight 15 catch both stages 2:50 – Louisiana spaceport announced 4:00 – Weekly then twice weekly flights 5:20 – Complete reusability like airplane 6:30 – Starlink 26 Tbps = 10 Falcon 9 7:50 – 100x bandwidth, 100k satellites 9:10 – 1000x demand from AI agents 10:20 – SI constraint: power & cooling 11:30 – Memphis 200k GB300s monthly 12:40 – Aiming 20 GW next year 13:20 – 440k Vera Rubin cluster Q2 14:00 – GrokBot fastest growth ever 14:35 – $500 B-$1T ARR by end 2027
On 20 October the DogeDesigner account (@cb_doge) posted a video it introduced as “full Elon Musk’s remarks at SpaceX Q3 2026 results”, with a chapter list of its own. The poster’s chapters include “Memphis 200k GB300s monthly”, “Aiming 20 GW next year”, “440k Vera Rubin cluster Q2” and “$500 B-$1T ARR by end 2027”. The chapter labels are the poster’s own.
Against What SpaceX Said in August
On the August call, Musk said: “We expect to end this year with over 2 gigawatts of compute.” He added that cumulative compute online by the end of 2027 may be “closer to 10 gigawatts of compute than 5 gigawatts of compute,” according to the transcript.
By our arithmetic, reaching more than 2 gigawatts from 1.7 gigawatts at the end of September means adding more than 0.3 gigawatts in the fourth quarter, about as much as SpaceX added in the third (1.4 to 1.7 gigawatts). On the same August call, SpaceX said new cloud agreements put it on a trajectory, including contribution from Cursor, “to reach $100 billion of ARR” by the end of this year, according to the transcript.
Starship: A Full Catch Next Month
Musk said SpaceX would attempt to catch Starship and its booster next month, with the goal of recovering the entire orbital launch system for the first time, The Verge’s Andrew J. Hawkins reported. “This is a profound technology milestone in the history of civilization consciousness,” Musk said, according to The Verge.
He also allowed for a failure. “Success is not guaranteed on these first flights, but I’d say it’s more likely than not that this succeeds next flight,” Musk said, The Verge reported. The release says Starship Flight 14 was SpaceX’s first revenue-generating Starship mission, delivering 26 Starlink V3 satellites.
The demand picture showed up elsewhere on the same evening: our piece on Supermicro set out the server maker’s preliminary first-quarter revenue estimate of $18.5 billion to $19.5 billion, which it put “primarily due to increased demand for AI platforms.” And our piece on Amazon’s AI spending reported, citing Business Insider, that some AWS sites had room for more racks but not enough available power to turn them on. As we read it, that is the same constraint Musk’s power-and-cooling line addresses.
The Structural Read
As we read it, the rate matters more than the stock: Musk said on the call SpaceX aims to “deploy power, cooling infrastructure at a rate that is vastly faster than anyone else,” according to CNBC.
As we read it, the August aim is the yardstick: Musk said then that SpaceX expected “over 2 gigawatts of compute” by year end, and by our arithmetic that needs more than 0.3 gigawatts in the fourth quarter, about what was added in the third.
As we read it, Starship is the other half of the plan: Musk said SpaceX would attempt to catch the ship and its booster next month and called success “more likely than not,” according to The Verge.
Elon Musk on SpaceX’s third-quarter call, 20 October 2026, as reported by CNBC
“deploy power, cooling infrastructure at a rate that is vastly faster than anyone else”
Three Implications
200,000 A MONTH SpaceX’s AI business is on track to be able to bring 200,000 Nvidia Grace Blackwell systems online each month for the rest of the year, CNBC reported.
1.7 GW NOW Nameplate compute reached 1.7 gigawatts on 30 September, against Musk’s August aim of more than 2 gigawatts by year end, according to the release and the transcript.
$16.4B OF SI CAPEX Of $19.6 billion of capital expenditure in the quarter, $16.4 billion was in the SI segment, according to the release.
The framework sizes the cycle this way: “The total cycle: $725B Big Four + $50B Oracle + $13B Apple + $60B Neoclouds + $80B China + $60B Sovereigns + $48B Other = ~$1.04 trillion in 2026 compute capex, the first trillion-dollar year of compute capex in human history.”
As we read it, a buyer bringing 200,000 Nvidia systems online a month, as CNBC reported of SpaceX, is the demand side of that cycle, and power and cooling are where it binds. The framework was published in May 2026 and does not discuss SpaceX’s call.
What Is Not Established
SpaceX had not posted a call transcript when this was written; the quotes here come from CNBC’s and The Verge’s reporting of the call, and we will check them against the transcript when it is posted. Posts on X relaying the call attributed further revenue and compute targets to Musk; we leave them out until a news outlet reports them in his words. The results release itself gives no formal revenue or capex guidance in the text we read.
Business Engineer Framework
The AI Capex Map
A Business Engineer analysis of how the AI build-out is financed, where it cracks and who can absorb it.
SpaceX’s AI business is on track to be able to bring 200,000 Nvidia Grace Blackwell systems online each month for the rest of the year, CNBC reported from the call, with nameplate compute at 1.7 gigawatts at the end of September against Musk’s August aim of more than 2 gigawatts by year end. As we read it, the call moved the question from whether SpaceX can raise money for compute to whether it can energise the chips as fast as it buys them.
Elon Musk predicted on the call that SpaceX’s annual recurring revenue by the end of 2027 will be “somewhere between $500 billion and $1 trillion,” TechCrunch reported, describing it as a particularly bullish stance on the Super Intelligence segment. It is Musk’s prediction, not a figure in the results release. In August, SpaceX described a trajectory to $100 billion of annualized revenue run rate by the end of 2026.
Chief financial officer Bret Johnsen said the company would spend $35 billion to $40 billion on new capital investments in the fourth quarter, and forecast total revenue in the range of $33 billion to $37 billion in the first quarter of 2027, according to TechCrunch. Third-quarter capital expenditure was $19.6 billion, per the release; by our arithmetic, the fourth-quarter range is about 1.8 to 2 times that.
“People are starting to realize [that] the constraint on deployment of of super intelligence is more in power and cooling and infrastructure than it is in chips,” Musk said, per TechCrunch, adding: “This is a bit like the New York Yankees playing in Little League. It’s not fair, but the net result is being able to deploy power cooling infrastructure at a rate that is vastly faster than any anyone else.” TechCrunch also reported that Musk said he was spending four days a week in Memphis, Tennessee, the site of Colossus II, and that he predicted the growth of AI agents would lead to more demand for connectivity.
Update, 21 October: From the Transcript
The Motley Fool’s transcript of the 20 October call records Musk saying he was “fairly confident of being able to bring an incremental 10 gigawatts of compute online in 2027,” in addition to the 2.5 to 3 gigawatts expected by the end of this year. “I would be surprised if we brought on less than 13 gigawatts of compute by the end of next year,” he said, per the transcript.
On a larger figure, Musk described “a non-zero possibility” of “SpaceX having over 20 gigawatts of Super Intelligence compute online next year,” adding: “A small possibility, but it used to be zero, it’s not zero anymore.” He also said SpaceX is planning “the most powerful training cluster online by the end of Q2, which will be 440,000 Vera Rubin computers deployed for optimal training,” the transcript shows. These are Musk’s targets, not company guidance.
CFO Bret Johnsen said: “At current pricing, new compute deployments are generating a payback of less than a year.” He said pricing “is currently tracking slightly above the high end of the $30 to $50 per watt range we were discussing just last quarter,” and that SpaceX remains “on track to deliver our previously stated goal of $100 billion in ARR by the end of this year,” per the transcript.
On Starlink, Musk said that “probably within three years” SpaceX expects “about 100x the bandwidth of the current Starlink constellation,” with “something on the order of 100,000 satellites,” according to the transcript. On compute supply, he said: “I think we’re going to face a shortage of super intelligence compute. And that shortage is going to remain for a long time.”
A note on the GPU pace: the transcript records Musk’s opening remarks as “roughly 200,000 GB300s online per month,” the figure in this article, while a later answer is transcribed as what is “needed to turn on 20,000 GB300s a month.” The two figures in that transcript differ by a factor of ten. A second transcript, published by Investing.com, records the opening as “roughly 200,000 GB300s online per month” and the later answer as what is “needed to turn on 220,000 GB300s a month”; CNBC’s live blog also reported 200,000. SpaceX has not yet posted its own Q3 transcript on its investor site (checked 21 October).
A note on sourcing. Statements from SpaceX’s 20 October 2026 earnings call are as reported by CNBC’s live coverage, The Verge and TechCrunch; SpaceX had not posted a transcript when this was written. Company figures come from SpaceX’s third-quarter 2026 results release filed with the SEC and its August 2026 call transcript. Percentages labelled as our arithmetic are our calculations from the release. Nothing here is financial or investment advice.
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.
