SpaceX reported Q2 2026 capital expenditures of $18.4 billion, a sixfold jump year-over-year, with the majority allocated to AI compute infrastructure. The company is building data center capacity powered by Nvidia GPUs to position itself as an alternative cloud provider, competing with OpenAI and Anthropic by renting compute capacity rather than building frontier models. Despite beating analyst expectations and narrowing losses, the stock fell over 10% in premarket trading and closed well below its $135 IPO price, signaling investor nervousness about the scale and timeline of AI infrastructure ROI.
CFO Bret Johnsen attempted to allay capex concerns by claiming the company achieves “less than a one-year payback” on AI compute deployments, suggesting customers are already willing to pre-pay for capacity. CEO Elon Musk accelerated revenue guidance, now projecting $1 trillion in annual revenue by 2030 (instead of 2031), implying rapid monetization of the compute footprint. Yet the market's skepticism persisted: investors remain unconvinced that a space company can execute at cloud scale or that AI demand will sustain capex burns of $18B+ annually.
For infrastructure planners, SpaceX's bet is directional: it signals that even space-grade operators are pivoting to compute-first business models, and that captive cloud capacity (via Nvidia and custom schedulers) is becoming table stakes for competing in AI services. The fact that investors reacted negatively despite forward guidance suggests the Street is pricing in a higher discount rate on AI infrastructure returns.
Stock trading context: SpaceX shares also face insider lockup expirations on August 8, 2026, which may exacerbate near-term selling pressure.