SpaceX’s SI segment and CoreWeave ended the third quarter with about the same AI capacity: 1.7 GW of nameplate compute at SpaceX, according to its 10-Q, and over 1.7 GW of active power at CoreWeave, according to its results release of 10 November. SpaceX’s segment brought in $7,560 million in the quarter, 2.1 times CoreWeave’s $3,661 million, by our arithmetic.
An evergreen Business Engineer explainer on capital spending and depreciation, the two lines that separate these AI clouds. It teaches the general idea only and does not discuss SpaceX or CoreWeave.
The key insight: As we read the two filings, SpaceX’s SI segment and CoreWeave run about the same 1.7 GW, but SI booked about $4.45 billion of revenue per gigawatt in the quarter against about $2.15 billion at CoreWeave, and spent $2.18 of capital per dollar of revenue against $2.58, by our arithmetic.
CoreWeave reported third-quarter revenue of $3,661 million, operating income of $207 million and an adjusted EBITDA margin of 62%. It said it “Added more than 200 MWs of active power to reach over 1.7 GW, doubling our active footprint since the start of the year”. Our piece on its results covers its $132.5 billion backlog.
SpaceX’s SI segment reported revenue of $7,560 million, operating income of $431 million and segment adjusted EBITDA of $4,953 million, its results release shows. Its nameplate compute rose to 1.7 GW from 1.4 GW at the end of June and 0.6 GW a year earlier, 2.8 times in a year, by our arithmetic.
The two businesses are not the same. The 10-Q says SI revenue includes “the sale of cloud services” alongside subscriptions for Grok, Cursor and X, data licensing, API access and X advertising. Without the $538 million of advertising, SI Solutions & Infrastructure brought in $7,022 million, 1.9 times CoreWeave’s revenue, by our arithmetic.
Revenue, adjusted EBITDA, capital spending and depreciation per gigawatt in the third quarter of 2026, CoreWeave against SpaceX’s SI segment, from the two companies’ filings; dividing both by 1.7 GW is our arithmetic, and the two define capacity differently.
Capex per Dollar of Revenue
SpaceX spent $16,444 million on SI capital expenditures in the quarter, $2.18 for every dollar of SI revenue, by our arithmetic. CoreWeave’s purchases of property and equipment, including capitalized internal-use software, were $9,441 million, $2.58 per dollar of revenue.
Per gigawatt, SpaceX’s SI capital spending was about $9.67 billion in the quarter, against about $5.55 billion at CoreWeave, by our arithmetic. Over nine months, SI capital spending of $39,995 million was 1.7 times CoreWeave’s $23,558 million.
Depreciation and amortization took 51.6% of CoreWeave’s revenue and 32.7% of SI revenue, by our arithmetic: $1,889 million against $2,473 million.
Same Operating Margin, Different Balance Sheets
Both kept 5.7% of revenue as operating income, by our arithmetic: $431 million at SI and $207 million at CoreWeave, whose release rounds it to 6%. On adjusted EBITDA, a non-GAAP measure, SI kept 65.5% and CoreWeave 61.6%.
Below the operating line the two diverge. CoreWeave’s net interest expense was $846 million, 4.1 times its operating income, by our arithmetic, and it reported a net loss of $709 million. SpaceX as a whole reported interest expense of $675 million and interest income of $873 million in the quarter, its segment note shows.
Chief executive Michael Intrator said CoreWeave’s growth “is supported by multiple sources of funding, including financing backed by customer contracts, operating cash flow, and approximately $9 billion in customer prepayments and deposits expected in the next three months alone”.
Capital spending, adjusted EBITDA, depreciation and operating income per dollar of revenue in the third quarter of 2026, CoreWeave against SpaceX’s SI segment; the ratios are our arithmetic on the two companies’ filings.
What to Watch
Capacity: CoreWeave said it continues to expect “more than 1.85 gigawatts of active power at year-end”, with full-year capital expenditures of “$37 billion to $39 billion”. SpaceX’s chief financial officer, Bret Johnsen, guided to fourth-quarter capital expenditures of “$35 billion to $40 billion” on the 20 October call.
Backlog: CoreWeave reported about $132.5 billion, 2.1 times SpaceX’s company-wide $62,959 million, which covers all three of its segments, by our arithmetic. Run-rate: CoreWeave raised its annualized run-rate revenue guidance to “$20 billion to $21 billion to exit 2026”.
The Structural Read
As we read the cost lines, depreciation is where the two differ most: it took 51.6% of CoreWeave’s revenue and 32.7% of SI revenue, yet both kept 5.7% as operating income, by our arithmetic.
As we read the financing lines, the gap opens below operating income: CoreWeave’s $846 million of net interest expense was 4.1 times its operating income, by our arithmetic, while SpaceX as a whole reported more interest income, $873 million, than interest expense, $675 million.
Michael Intrator, chief executive, CoreWeave third-quarter 2026 results release, 10 November 2026
“This was a breakout quarter for CoreWeave, with revenue increasing by more than $1 billion sequentially and adjusted operating margins reaching double digits a quarter ahead of plan”
Three Implications
FOR REVENUE PER GIGAWATT SpaceX’s figure includes Grok, Cursor and X subscriptions and X advertising on top of cloud services, the 10-Q shows; without advertising, SI revenue was 1.9 times CoreWeave’s, by our arithmetic.
FOR CAPITAL INTENSITY Over nine months SI capital spending was $39,995 million against CoreWeave’s $23,558 million, 1.7 times, by our arithmetic.
FOR BACKLOG CoreWeave’s about $132.5 billion is 2.1 times SpaceX’s company-wide $62,959 million, by our arithmetic, and SpaceX’s figure covers launch and Starlink too.
The framework’s thesis: “The structural bottleneck isn’t computing, it’s electrons”.
As we read it, a per-gigawatt comparison is that thesis applied to two incomestatements: with capacity about equal, SpaceX’s SI segment reported about twice CoreWeave’s revenue per gigawatt, by our arithmetic. The framework predates both reports and does not discuss either company.
What Is Not Established
The two companies measure capacity differently. SpaceX’s nameplate compute counts GPUs installed times their all-in power and excludes cooling and facility overhead; CoreWeave’s release, as we read it, does not define active power. The per-gigawatt figures divide by 1.7 GW for both and are approximations. SpaceX does not report a separate cloud-only revenue line for the SI segment, and its backlog is not broken out by segment in the passages we read.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
On roughly the same 1.7 GW, SpaceX’s SI segment reported $7,560 million of revenue to CoreWeave’s $3,661 million, spent $2.18 of capital per dollar of revenue to CoreWeave’s $2.58, and kept the same 5.7% operating margin, by our arithmetic on the two companies’ filings.
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 10-Q for the quarter to 30 September 2026 and its third-quarter results release, both filed with the SEC; CoreWeave’s come from its third-quarter results release, filed with the SEC on 10 November 2026. Johnsen’s guidance quote comes from The Motley Fool’s transcript of SpaceX’s 20 October call. Ratios, per-gigawatt figures and multiples are our arithmetic. Neither company was contacted for this piece.
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.
