SpaceX or Micron? SpaceX wins as the better 5-year AI bet

The column frames the contest as a clash of business models rather than a head-to-head benchmark. SpaceX's pitch rests on leasing spare AI compute at high prices, an AI coding push after its acquisition of Cursor that Morgan Stanley projects could become a $33 billion annual recurring revenue business by 2030, up from about $8 billion this year, and a fast-growing, profitable Starlink satellite internet arm. SpaceX's infrastructure spending is described as returns within one year through short-term leases, which the author says funds aggressive capacity expansion. Micron's case, by contrast, depends on a memory supercycle: DRAM was 73% of Micron's revenue last quarter, and both DRAM and NAND are in short supply. Rival SK Hynix expects the DRAM supply-demand imbalance to last through at least 2030, partly because Samsung and its peers cannot expand capacity fast enough. HBM is the driver for AI accelerators, but Micron holds only around 18% HBM share in Q2 and draws likely less than 10% of revenue from HBM, leaving it more exposed to conventional memory prices than the bullish HBM narrative suggests. All three DRAM makers chase the same EUV lithography systems from ASML, and HBM consumes upward of three times the wafer capacity of ordinary DRAM, so capacity cannot be added quickly. That gap between narrative and revenue mix is why the author still favors SpaceX, even with Micron cheaper on the surface.