SpaceX's Q2 2026 earnings reveal massive AI spending

SpaceX folded xAI into itself in May 2026, and the deal that followed now defines the company's earnings ^[1]. Anthropic agreed to rent all of the Colossus 1 supercomputer's capacity for $1.25 billion a month through May 2029, roughly $45 billion over the contract's life, making it SpaceX's largest AI compute customer ^[2]. Google followed weeks later, agreeing to pay $920 million a month for compute at xAI's data centers from October 2026 through June 2029, close to $30 billion total ^[3]. A third tenant, Reflection AI, began paying $150 million a month for GB300 GPU capacity in July 2026 ^[2]. Stacked together, that is somewhere between $26 billion and $28 billion a year in annualized compute-leasing revenue, roughly twice what established GPU-rental specialist CoreWeave currently generates, prompting one analyst to call the arrangement 'SpaceX the Neocloud!' ^[4]. The shift already shows up in the numbers: the AI segment brought in $2.56 billion for the quarter ^[5], more than double the $962 million generated by SpaceX's original Space segment ^[13]. A rocket company is now, by revenue, primarily an AI compute landlord.

By almost every headline metric, SpaceX beat expectations: revenue of $7.8 billion topped the roughly $6.9 billion consensus ^[6], and the AI division's operating loss came in at $1.26 billion, far narrower than the $2.39 billion analysts had feared ^[7]. Yet the number that dominated the reaction was capex (capital expenditure, meaning money spent building physical infrastructure like data centers): AI-related spending alone hit $15.83 billion, a roughly 2,013% year-over-year jump that blew past the already-aggressive $13.22 billion estimate ^[5]. Shares fell 7-8% in after-hours trading despite the beat ^[8]. Retail investor commentary on Reddit and X captured the same tension from the other direction, comparing SpaceX's narrowed but still-real net loss against profitable chipmakers posting similar revenue, and questioning whether AI capex is outrunning the contracted revenue that is supposed to justify it. The market, in effect, is no longer grading SpaceX on whether it beats estimates, but on whether its spending pace is sustainable.

Two trading days after this report, on August 6, 2026, up to 911.5 million SpaceX shares (worth roughly $116-123 billion) become eligible for sale for the first time since the IPO, the first and largest of five staggered lock-up triggers ^[9]. That timing was not a coincidence; SpaceX's July 21 decision to set its earnings date ^[14]effectively started the countdown to this unlock ^[10]. Ahead of it, short sellers had already built a position of 219.3 million shares, about 34% of the float and worth $24.6 billion ^[9]. That is a mechanical supply-side risk, distinct from anything in the earnings themselves: even a company executing well can see its stock pressured simply because far more shares suddenly become tradable than the market has absorbed before. Several market commentators framed this unlock, not the earnings beat or miss, as the dominant catalyst for SpaceX stock over the following days.

Musk used the earnings call to move up his own long-term targets, saying internal projections for reaching $1 trillion in annual revenue have shifted from 2031 to 2030, with 'a non-zero chance' of getting there by 2029 ^[8]. Management framed a $100 billion annualized revenue run-rate by the end of 2026 as essentially locked in, saying that figure is 'not a question mark' even under a do-nothing scenario ^[6]. Not everyone is convinced the spending behind that trajectory pencils out. Glenn Thum of Phillip Securities initiated coverage with a Sell rating and a $75 price target, implying roughly 35% downside ^[11], and separately forecast that SpaceX will burn more cash than it generates every year through at least 2030, close to $90 billion cumulatively, summarizing his view as: 'AI carries the valuation but not the earnings' ^[12]. The same aggressive buildout that worries Thum is also SpaceX's competitive weapon: its AI compute business now annualizes to roughly twice CoreWeave's revenue, and SpaceX is reportedly chasing a Pentagon AI cloud contract that would undercut CoreWeave on price ^[4]. Whether that scale becomes a moat or a cash-burning liability is the question this earnings report left unresolved.