Opinion: Space AI Computing Power Deployment Could Become a Significant Upside Option for SpaceX Valuation, Orbital Approach Not GPU-Cost-Prohibitive and Potentially Only 1/4 Ground Expenses

June 16 — In the latest episode of the BG2 tech investment podcast, Altimeter Capital CEO and co-founder Brad Gerstner joined Andrew Fox (Atreides Management analyst, known by the nickname Foxy) to discuss the economics of space-deployed AI computing, framing the concept as long-term option value for SpaceX’s valuation. BG2 is a prominent Silicon Valley tech-focused investment podcast hosted by Gerstner and former Benchmark general partner Bill Gurley. According to Fox’s analysis, each Starship launch could deliver roughly 5 terawatts of computing capacity. Extrapolating, the capital expenditure to deploy the necessary hardware in space would come to around $5 billion per gigawatt. Fox contrasted that figure with ground-based data centers, noting non-GPU infrastructure costs for on-the-ground facilities (including switching gear, generators, transformers, buildings, power acquisition and cooling systems) run $20–25 billion per gigawatt. That means if orbital computing becomes a reality, some core infrastructure costs could be far lower than terrestrial counterparts. From a total computation construction cost perspective, Gerstner added that today’s 1-gigawatt ground deployment typically costs about $60 billion overall: $35 billion for GPUs and chips, plus $25 billion for land, buildings, power and cooling. He pointed out these ground-related costs face inflation headwinds in the future, while space’s power and cooling are effectively free for economic purposes. That said, Fox emphasized space-based computing is not a requirement for SpaceX’s IPO valuation. Instead, he called the business line significant upside potential—not a core premise for the company’s current valuation.