The question used to be whether data centers could exist in space. That debate ended sometime around January. The question now is who captures the value when they do, and whether the stock already trading on the assumption they will is priced correctly.

Two orbital data center nodes launched to low-Earth orbit on January 11, 2026, establishing the first operational foundation for space-based cloud computing. Starcloud launched Starcloud-1 in November 2025, carrying an Nvidia H100 GPU and became the first company to run and train small-scale language models in space. By February, multiple orbital data center operators were simultaneously running production workloads in space for the first time in history.

None of that moved markets the way SpaceX did. Reuters reported that SpaceX priced its Nasdaq IPO at $135 per share, selling about 555.56 million shares to raise roughly $75 billion and valuing the company at about $1.77 trillion. The orbital compute program was central to that pitch. SpaceX revealed the design of its first orbital data center satellite, called AI1, measuring 70 meters tip to tip and designed to sustain 120 kilowatts of compute output on average and burst to 150 kilowatts at peak draw.

Elon Musk said the first AI satellites, powered by Nvidia chips, will launch in the fourth quarter of 2027 and reach significant scale in 2028, with SpaceX and Nvidia having designed a space-optimized Vera Rubin NVL72 system for that launch. The timeline has moved up twice since the IPO prospectus. The company intends to rent out compute from the satellites, and it has struck major compute-rental agreements with Anthropic and Google for its terrestrial facilities.

The economics behind all of this remain genuinely unsettled. A SemiAnalysis model from June 2026 puts space compute at more than four times terrestrial cost today, at roughly $8.64 versus $2.37 per GPU-hour for a comparable cluster, narrowing to a 30% premium by the early 2030s and reaching full cost parity around 2040 in the base case. The bull argument is not that orbital compute is cheaper today. The real question is whether space compute becomes viable as overflow capacity in a world where terrestrial data center deployment is bottlenecked by long power grid queues while AI compute demand is scaling rapidly.

Radiation damages electronics. The electronics generate enormous amounts of heat, and getting rid of that heat is surprisingly difficult in space. Repairs are extraordinarily expensive, and every pound launched into orbit still carries a significant cost. Microsoft’s Project Natick is instructive. Servers in its underwater data center failed at just one-eighth the rate of a land-based control group, but Microsoft never scaled Natick into a commercial business.

That is the question SpaceX has to answer with AI1: whether a sealed, orbiting asset can keep pace with a chip cycle that moves every 18 months. SpaceX’s design attempts to address this by making the compute provider interchangeable, meaning the spacecraft is not locked to a single chipmaker.

The market has already weighed in. SPCX closed at about $152.71 on September 18, 2026, implying a market cap around $2.0 trillion. Kalshi puts the chance of a one-megawatt orbital data center going live before January 1, 2031 at about 30%.

That gap between hedge fund conviction and prediction market skepticism is where the investment decision lives. Orbital data centers exist today at demonstration scale. Commercial scale, the kind that justifies $2 trillion, requires Starship launch cadence, space-grade chip economics, and thermal engineering to all converge on a timeline Musk keeps accelerating. The stock is pricing in a world where they do. The honest answer is that nobody knows yet whether they will.