Elon Musk's SpaceX IPO Friday values the company at $1.75 trillion, making it the largest initial public offering on record. The headline number masks what the offering really represents: not a transportation company, but an AI infrastructure play backed by massive capital commitments and orbital ambitions.
SpaceX's AI division lost $6.4 billion last year, a figure that signals the company's pivot toward artificial intelligence development rather than traditional rocket launches. The losses reflect spending on computational infrastructure and research that positions SpaceX as a competitor in the AI race alongside OpenAI, Anthropic, and other frontier labs.
The company plans to deploy one million data-center satellites into orbit, a project that would create a distributed computational network spanning the globe. These satellites would function as nodes in an AI training and inference infrastructure, reducing latency and creating redundancy across continents. The scope of this project dwarfs previous satellite internet ambitions.
SpaceX's valuation has more than doubled since December, driven by investor appetite for AI infrastructure plays and confidence in the company's technical execution. The price reflects not skepticism about the rocket business, which remains profitable, but euphoria around the AI opportunity. Traditional space companies trade on revenue multiples. SpaceX trades on growth potential in a sector that barely existed five years ago.
The contrast with Apple's strategy offers perspective. Apple reduced its AI ambitions this year, scaling back generative AI features and focusing on on-device processing. The company bet against centralized AI infrastructure and external data collection. SpaceX bets everything on distributed, space-based infrastructure that feeds continuous AI development.
Both bets assume the future belongs to companies that control their computational layer. SpaceX's approach requires building satellites, launching them reliably, and maintaining networks in space. Apple's approach requires building chips and keeping data local. One path scales through space.