SpaceX · Earnings · 20 Oct 2026
SpaceX Q3 results: $14.6B revenue and a $4.4B profit as AI becomes the biggest business
SpaceX beat Wall Street's revenue estimate by almost $2 billion in the third quarter. It also reported $4.4 billion in net income, against a loss a year ago. AI (the "SI" segment) brought in more than Starlink for the first time. Capital spending also hit a record.
The headline numbers
SpaceX's figures come from its Q3 2026 results release, filed with the SEC. The estimates are the ones from my preview this morning (LSEG and StreetAccount, via CNBC).
| Metric | Q3 2026 (reported) | Estimate | Q3 2025 |
|---|---|---|---|
| Revenue | $14.58B | ~$12.72B | $5.58B |
| Net income (GAAP) | $4.40B | ~$1.57B | –$1.05B |
| EPS | $0.35 diluted (GAAP) | 13c | –$0.36 |
| Operating income | $2.79B | n/a | –$0.27B |
| Adjusted EBITDA | $8.81B | n/a | $2.17B |
| Capex | $19.59B | n/a | $6.88B |
Reuters puts EPS at 45 cents on IBES's basis, against a 13-cent average estimate. SpaceX's own GAAP diluted figure is 35 cents. It also cites a $1.57 billion net income estimate (Reuters via NBC News). Revenue rose 161% on the year and adjusted EBITDA rose 306% (SpaceX). Reuters reported the shares were up about 2.5% in extended trading after the release.
Why each number matters
- Revenue nearly doubled from Q2 ($7.81B) in a single quarter. That is the clearest sign the AI compute contracts are now real, booked revenue rather than potential.
- Net income is helped by items below the operating line. Those include $873M of interest income, $1.03B of other income and a $390M tax benefit (SpaceX). Operating income of $2.79B is the cleaner read on the business itself.
- Adjusted EBITDA leaves out depreciation, which matters for a company buying data centres this fast. It flatters a capital-heavy business, so read it alongside capex.
- Capex of $19.6B in one quarter is more than the revenue. That is why profit and cash are not the same story here.
The segments
- SI (AI): $7.56B, against an estimate of about $6.47B and up from $813M a year ago. Operating income was $431M, after a $1.26B loss in Q2. Reuters says the revenue comes mostly from renting computing capacity to Anthropic, Google and Reflection AI (Reuters). SpaceX says nameplate compute reached 1.7 GW at 30 September. It also credits the growth to cloud agreements and "the acquisition of Cursor" (SpaceX).
- Connectivity (Starlink): $5.25B, against an estimate of about $4.84B. Operating income was $2.37B, still the most profitable segment. Starlink had 13.6M consumer subscribers, up from 7.2M a year ago, at a consumer ARPU of $69 a month, down from $83 (SpaceX). Reuters says analysts expected 12.19M subscribers.
- Space (launch): $1.77B, against a StreetAccount estimate of about $1.41B (preview). The segment roughly broke even, with an operating loss of $7M. SpaceX completed 12 customer launches. It also flew Starship Flight 14, its first orbital Starship flight, which carried 26 Starlink V3 satellites (SpaceX).
Capex and AI
Of the $19.59B in capex, $16.44B went to SI. Space took $1.61B and Connectivity $1.54B (SpaceX). CNBC reports more than $16.4B of it went to the AI business (CNBC). Reuters adds that SpaceX has cautioned the AI unit needs sustained investment before it can consistently make a profit (Reuters).
From the call and Q&A
The webcast streamed on @SpaceX. The quotes below are my own transcription of call clips posted by @SawyerMerritt.
- Revenue run-rate (an aspiration). Musk said: "what this implies for SpaceX annual recurring revenue by the end of '27 is somewhere between $500 billion and a trillion" (clip; CNBC). That is not company guidance.
- Tesla merger. Musk said both companies have formed special committees of independent directors "at the request of shareholders". He said he can't take part because he is "a conflicted party". In his words: "I don't think we should necessarily assume that a SpaceX and Tesla merger is a certainty." He also said "the special committees have not reached agreement on the terms of any transaction", and that Tesla shareholder approval would be "quite a high hurdle". Then he added: "Do I think it would be a good idea? Yes." (clip; CNBC). For the background, see my piece on the Barclays merger note.
- Terafab (aims, not commitments). Musk said the internal target is a first functional test chip at the Austin research fab "in approximately May or June of next year". The aim is volume production at the main Terafab site around Q3 2028, and "at least in a set of possible outcomes for Q3 or Q4 of '28". He called them "very aggressive dates" with "big error bars" (clip; CNBC).
- Tone. "We are in an inflection moment for SpaceX across launch, connectivity and superintelligence," Musk said (Reuters).
What it may mean for the stock
This is analysis, not advice. The bull case is that a beat this size, driven by outside customers renting compute, shows SpaceX can earn money on AI even without Grok winning the model race. Starlink is still a profitable base under it. The bear case is in the same table. Capex is bigger than revenue, ARPU keeps falling as Starlink goes abroad, and part of the net income comes from non-operating items. The things to watch: whether SI stays profitable at the operating level, how capex trends, the spectrum deal's FCC approval, and anything concrete from the merger committees.
My read
This morning I said the split mattered more than the headline, and it did. AI is now the biggest segment, and it made an operating profit for the first time. But it is rented compute, closer to a cloud business than an AI lab. Starlink still does most of the profit work. The ARR and Terafab numbers are Musk's ambitions, and I'd treat them as targets until they appear in the results.
Sources: SpaceX Q3 2026 results release (SEC filing, 20 Oct 2026); Reuters via NBC News: estimates, IBES EPS, customers, shares after hours; CNBC: LSEG and StreetAccount estimates; call clips from @SawyerMerritt. Not investment advice.
— Alex R
