Morgan Stanley analyst Adam Jonas wants SpaceX investors to learn a new phrase. In a note shared on October 13, he called the company’s Distributed Inference Cloud the part of the SpaceX story with the greatest growth potential, and one investors are largely ignoring [1]. He told clients to expect the idea, along with “physical AI swarms,” to break into the mainstream investor conversation within a year.
The concept is simpler than it sounds. Inference is the work an AI model does when it answers a question or makes a decision. Instead of doing all of that work in giant data centers, a distributed cloud spreads it across many smaller computers in the real world, such as robots, cars and satellites, all linked together.
#Why Robots Could Become Data Centers
Jonas laid out the numbers in an August note. Morgan Stanley expects roughly 2.2 billion robots of all kinds by 2040. If each carries computing hardware drawing about 500 watts, the combined power requirement would reach roughly 1.1 terawatts. Morgan Stanley argues that such devices could theoretically form a distributed computing network connected through Starlink. In theory, that would create a distributed network with a power footprint comparable to SpaceX's 1-terawatt ambition, although it would not necessarily deliver the same usable computing capacity as dedicated data centers.
The idea is not new to Tesla investors. Elon Musk has often talked about using idle Tesla cars, and computing units placed at Supercharger stations, as one huge shared computer. Tesla’s AI5 chip, intended for use in vehicles and Optimus robots, has a reported power requirement of about 250 watts, and Morgan Stanley expects that figure to rise. As SpaceX and Tesla work more closely together, Jonas thinks SpaceX shareholders will need to understand the same logic.
There is also a practical angle. Jonas noted growing corporate, regulatory and public pushback against new data centers. Spreading computing power across devices people already own, or across satellites in orbit, could ease some of that pressure.
#The Spending Comes First
The vision is long term, but the bill is arriving now. SpaceX’s first earnings report as a public company showed $18.4B of capital spending in the second quarter, with $15.8B of it going into AI infrastructure [3]. AI revenue was $2.6B, up sharply, and the company signed $14.1B of cloud service contracts.
The chart shows the gap clearly. In the second quarter, SpaceX spent about six dollars on AI infrastructure for every dollar of AI revenue, although capital expenditure represents investment in assets intended to generate returns over multiple periods. Management says new computing capacity can pay for itself in under a year because demand outstrips supply. SpaceX ended the quarter with 1.4 gigawatts of computing power and told investors it aims to pass 2 gigawatts by year end, while Musk floated a tentative target of 20 gigawatts by the end of 2027.
A sharper way to see the markets in just 5 minutes.
Same news, different lens. We cut through the noise and hand you the overlooked ideas and the deeper read the crowd misses. Join 38,000+ investors seeing the markets differently.
#What Could Go Wrong
The robot swarm math is a thought experiment, not a forecast of revenue. It assumes billions of robots get built, carry powerful chips, and can be safely and reliably linked into one network. Each of those steps carries serious technical, regulatory and commercial uncertainty, and 2040 is a long way off.
Investors are also still weighing the near-term costs. SpaceX shares are sensitive to AI-linked sentiment. The real tension is timing. Jonas is asking investors to value a network that barely exists yet, while the company spends heavily today to build the data centers that come first. Watch the next earnings report, due early November, for whether AI revenue starts closing the gap with AI spending. That will say more about this story than any 2040 projection.