Earnings Call Highlights
SpaceX's Q2 2026 earnings call was dominated by an unmistakably bullish and expansive tone from management, with Elon Musk projecting extraordinary confidence across AI compute, Starlink, and Starship simultaneously. The AI infrastructure narrative was particularly striking, with management framing SpaceX as the fastest-scaling compute builder on Earth, exclusively committed to NVIDIA's Vera Rubin architecture, and already generating sub-one-year payback periods on GPU deployments. Analyst questions were heavily focused on the sustainability of compute pricing, the path to $100 billion ARR, and the mechanics of the AI cloud services ramp, suggesting the buy-side is rapidly repricing SpaceX as an AI infrastructure company, not just a launch provider. Management posture was assertive and forward-leaning, with Musk pulling the $1 trillion revenue target forward a full year to 2030 and suggesting 2029 is plausible, while Bret Johnsen grounded the narrative with concrete near-term contract disclosures. The overall sentiment was one of a company that believes it is operating in a different competitive league entirely.
- SpaceX ended Q2 with 1.4 gigawatts of nameplate compute, up from 1 GW in Q1 and 400 MW a year ago, and is targeting over 2 GW by year-end; Musk stated the ambition is 15-20 GW of power and cooling infrastructure online by end of 2027, with NVIDIA expected to supply a 'very significant percent' of their GPUs to SpaceX next year.
- SpaceX has gone exclusively NVIDIA, with Musk stating 'we've decided to build exclusively on NVIDIA because we think the Vera Rubin architecture is the best architecture' — a direct and unambiguous endorsement that forecloses any alternative chip vendor relationship for the foreseeable future.
- AI segment revenue hit $2.6 billion in Q2, up 247% year-over-year, driven by new cloud services agreements with customers including Google and Anthropic; the segment turned adjusted EBITDA positive at $1.1 billion, and in the first weeks of Q3 alone SpaceX contracted an additional $6.7 billion of cloud services revenue over a 6-month period beginning October.
- Musk offered a rough monetization estimate of $30-$50 per watt for Vera Rubin compute, and argued that supply-demand economics strongly favor sustained or rising pricing: 'demand is increasing by 200% a year, maybe higher' against memory supply growing only ~20% annually, making price compression unlikely in the near term.
- The Starmind AI satellite — described as an optimized Vera Rubin NVL72 computer in orbit — is expected to begin launching next year, with Musk noting SpaceX intends to deploy the same NVL72 form factor on the ground as well, framing it as a radical simplification of standard rack design that will cost less and perform better.