SpaceX swings to profit on AI infrastructure boom, but capex bill is climbing

SpaceX Corp (NASDAQ:SPCX) is becoming an AI infrastructure company that also happens to launch rockets, according to analysts.

The market rewarded the shift Wednesday, sending shares up more than 10% after third-quarter results beat estimates in each of its three segments.

The numbers explain the reaction. Revenue in the AI infrastructure segment (SI) reached $7.6 billion, versus $2.6 billion a quarter earlier and a Street forecast of $6.5 billion. EBITDA of $5 billion, against $4 billion expected, came at a 66% margin, up from 45% in the prior quarter.

UBS attributed the jump to a full run-rate contribution from the Anthropic deal, an initial contribution from Google and the close of the Cursor acquisition.

Company-wide, revenue grew 161% from a year earlier, and SpaceX swung to a $4.4 billion profit from a $1.1 billion loss.

The forward guidance matters more than the quarter. SpaceX now targets 13 gigawatts of compute by the end of 2027, against the 2.5 to 3.0 gigawatts UBS had modelled and the 1.7 gigawatts in place at quarter end.

Elon Musk said he was “hounded by every hyperscaler” to lease more capacity and that demand could absorb several times what the company has, with upside to 20 gigawatts depending on power generation. He also expects spot rental rates to keep rising, which could cut payback periods to about six months next year.

UBS said the build, split roughly 50:50 between held and leased capacity, would support annual recurring revenue of $0.5 trillion to $1 trillion by the end of 2027, up from an expected $100 billion at the end of 2026.

The cost of getting there is the central debate. SpaceX guided to fourth-quarter capital expenditure of $35 billion to $40 billion, roughly double the $19.6 billion spent in the third quarter and well above UBS’s $19 billion estimate.

UBS calculates that reaching 13 gigawatts requires about $350 billion in total spending, which puts SpaceX alongside Alphabet at roughly $380 billion and Microsoft at roughly $230 billion.

Swissquote’s Ipek Ozkardeskaya cited S&P Global Visible Alpha figures showing capex of nearly $70 billion this year, climbing to almost $197.4 billion in 2028. “Great growth comes at a big cost,” she wrote.

Funding capacity may prove a competitive edge. Ozkardeskaya pointed to stress among smaller AI infrastructure providers, noting that CoreWeave’s 8.5% bond due in 2032 now yields almost 15%, while SpaceX’s 2033 bond yields less than half that. In her view, if credit conditions tighten, SpaceX could absorb business from weaker rivals.

Elsewhere, Starlink added 1.6 million subscribers, short of UBS’s 2.3 million forecast, and average revenue per user fell 17% to $69, though management expects higher fourth-quarter volumes.

Space revenue of $1.77 billion and EBITDA of $357 million beat consensus, which had called for an EBITDA loss of $46 million, even as revenue growth slowed to 6% amid the shift from Falcon 9 to Starship.

Starship Flight 14, the program’s first orbital and revenue-generating mission, deployed 26 Starlink V3 satellites, roughly 10 times a Falcon 9 mission’s downlink capacity.

Management is targeting weekly flights in the first half of 2027, which would imply up to 72 launches in the year against UBS’s estimate of 30 and the Street’s 40. It plans a ship and booster catch attempt on Flight 15 next month.

UBS, which has a Buy rating, raised its price target to $230 on higher segment estimates. The target is based on a sum-of-the-parts valuation that implies 28 times 2028 EBITDA.

Ozkardeskaya was more measured on valuation. “It’s hard to tell whether this is expensive or cheap,” she said, adding that the answer depends on how much investors believe in Musk’s ability to execute.

A Tesla combination adds another variable. Musk said special committees have been formed to evaluate a transaction, but he is recused and limited in what he can say. He said there was “a lot of merit” in combining the companies but that a deal was by no means inevitable.

Ozkardeskaya said investors could reject a deal unless the synergies create value for both sides.