Bernstein gave SpaceX an "outperform" rating and a target price of $239, but its core valuation isn't just about rockets; it's about whether AI computing power can ultimately be deployed in space. The firm's model assumes that by 2031, SpaceX will complete approximately 3,600 launches annually, deploying nearly 170,000 satellites and generating about 20GW of orbital computing power.
The logic is that as terrestrial data centers become increasingly constrained by power, land, and cooling limitations, if Starship achieves rapid two-stage reuse and significantly reduces launch costs, orbit could become a new location for computing power. However, this doesn't eliminate resource constraints; it merely shifts the bottleneck to chip supply, satellite cooling, radiation resistance, communication, and regulation. The scale envisioned by Bernstein alone would require approximately 2.8 million wafers annually, equivalent to about five dedicated wafer fabs. Therefore, terrestrial AI data centers remain the main revenue driver for this decade; orbital computing power is more of a long-term option.
