SpaceX just dropped its Q2 earnings, and the numbers reveal a massive bet on AI infrastructure paying off - sort of. The company's AI revenue shot up more than 3x to $2.6 billion year-over-year, fueled by major compute deals with Anthropic and Google that position SpaceX as a serious neocloud player. But here's the catch: the AI division still burned through $1.5 billion this quarter, even as SpaceX told investors in its IPO documents that AI represents "most of its value." The earnings paint a picture of a company racing to dominate AI compute while wrestling with the brutal economics of the neocloud business.
SpaceX is making a risky pivot that's reshaping the AI infrastructure landscape. The company's Q2 earnings reveal AI revenue of $2.6 billion - more than triple the $867 million it pulled in during the same quarter last year, according to quarterly results reported by The Verge. The surge comes almost entirely from deals to provide compute capacity to other AI companies, pushing SpaceX into direct competition with specialized neoclouds.
The timing of these deals tells the story of how quickly the neocloud market is consolidating. Anthropic signed on in May, followed by Google in June in a deal worth a reported $920 million monthly, according to CNBC. These aren't small pilot programs - they're massive infrastructure commitments that signal both companies see SpaceX as a viable alternative to traditional cloud providers and newer players like CoreWeave.
But the revenue explosion masks a fundamental tension in SpaceX's strategy. The AI division hemorrhaged $1.5 billion this quarter, an improvement from the $1.6 billion loss in Q2 2025 but still a staggering burn rate for a business SpaceX claims is the "source of most of its value" in documents filed ahead of its planned public offering. The math is brutal: even with revenue tripling, the unit economics suggest SpaceX is either pricing aggressively to win market share or facing the same capacity constraints plaguing the entire neocloud sector.
The competitive dynamics are getting nasty. CoreWeave went public earlier this year on the promise of being the go-to infrastructure provider for AI training and inference, and suddenly SpaceX is eating into that narrative with deeper pockets and existing relationships with the biggest names in AI. Google's decision to pay SpaceX for compute is particularly striking - it signals that even hyperscalers with their own massive data center footprints need external capacity to meet AI demand.
SpaceX's IPO filing language is doing a lot of heavy lifting here. Calling AI "the source of most of its value" when the division is losing $1.5 billion a quarter requires investors to believe in a path to profitability that isn't visible in these numbers yet. The slight improvement in losses year-over-year suggests economies of scale are starting to kick in, but the company needs those margins to improve dramatically before public market investors will buy the growth story.
The deals with Anthropic and Google also reveal something about the state of AI infrastructure scarcity. Both companies are sophisticated buyers with options - Anthropic could tap Amazon Web Services (which invested $4 billion in the company), while Google obviously has its own Google Cloud Platform. Their willingness to pay SpaceX suggests genuine capacity constraints across the industry, or SpaceX is offering pricing that's hard to refuse even at the cost of massive losses.
The neocloud model itself is under pressure. CoreWeave has taken on billions in debt to build out data centers, and now SpaceX is signaling it's willing to spend heavily - "the increased presence in AI is also driving the company to spend more," the earnings note - to capture market share. This sets up a potential race to the bottom on pricing just as capital expenditures are exploding across the sector.
What makes SpaceX's position unique is its diversification. Unlike pure-play neoclouds, SpaceX has its core launch business and Starlink to fall back on, giving it financial flexibility to absorb AI losses longer than competitors. But that cuts both ways - investors in the IPO will need to decide if they're buying a rocket company that dabbles in AI or an AI infrastructure play that happens to launch satellites.
The Q2 numbers also raise questions about what SpaceX is sacrificing to hit these revenue figures. Tripling revenue while only marginally improving losses suggests pricing that prioritizes land-grab over profitability. That's a defensible strategy if SpaceX believes it can flip customers to higher-margin contracts later, but it's a gamble that depends on AI demand staying red-hot and competitors not matching its pricing.
SpaceX's Q2 earnings expose the central paradox of the neocloud boom: explosive revenue growth that still can't overcome the brutal capital intensity of AI infrastructure. The company's betting that $2.6 billion in AI revenue and marquee customers like Anthropic and Google will convince public market investors to overlook $1.5 billion quarterly losses, but the path to profitability remains murky. For the broader AI infrastructure market, SpaceX's aggressive push validates that compute scarcity is real and lasting, but it also signals a potential price war that could squeeze margins across the sector. The real test comes when SpaceX hits public markets and has to prove this AI pivot is more than an expensive distraction from its core rocket business.
