Musk: Orbital Compute, the Only Way to Scale AI

Musk: "We Can't Keep Expanding on the Ground"

On August 14, Elon Musk posted a short line on his X account. By around 2029, he said, orbital computing will be the only way to keep scaling AI. He gave two reasons: power is becoming increasingly hard to secure, and permitting delays keep slowing down every attempt to build data centers on the ground. Musk wrote, "Power and permitting issues make orbital computing the only way to scale."

The remark wasn't a new claim on its own — it came as a citation of someone else's analysis. The post Musk linked to was a thread by an account named Owen Lewis, introducing a computing infrastructure simulation model from Mach33 Financial Group. Lewis singled out the model's orbital data center section as its most important part, highlighting the line: "What's dismissed as a niche today — orbital data centers — will become a necessity for AI scaling in the 2030s."

Mach33's "Keystone" Model Projects 2040

On August 13, Mach33 Financial Group released a market simulation model called "AI Compute Keystone." The model forecasts supply and demand for AI computing infrastructure from 2026 to 2040, and notably covers both ground-based and orbital infrastructure together. According to Mach33, the model ran 5,000 simulations across thousands of combinations of assumptions to account for uncertainty.

Metric	Mach33 Forecast (2026–2040)
Cumulative token throughput	Approximately 800 quadrillion tokens
Dedicated infrastructure scale	Approximately 500GW
Infrastructure layer revenue	Approximately $3.2 trillion (excluding model and app layers)

The company noted that these figures represent revenue projections for the infrastructure layer alone, excluding revenue from the model or service layers built on top of it. At 500GW, the scale is substantial even compared to current total global data center power consumption, and whether power at this scale can be sourced from the ground alone is at the heart of the debate.

The Ground-Level Bottleneck — Power and Permitting

There's context behind Musk specifically citing "permitting issues" as a reason. Data centers run by his own company, xAI, have already drawn controversy over unpermitted turbines. It has been confirmed that SpaceX pushed back by another year its plan to remove all unpermitted turbines installed at xAI's data center — a case that illustrates how the pace of securing power for AI data centers is outrunning permitting procedures. The point that power has become the core bottleneck in the AI infrastructure race is also evident in this case.

Bringing new gigawatt-scale power online on the ground requires clearing three layers of process: expanding power generation, connecting to transmission grids, and obtaining local permits. Given that each stage often takes years, the math suggests it would be physically difficult to meet hundreds of gigawatts of demand within the 2030s. The orbital computing scenario stems from exactly this speed limit built into ground infrastructure.

Editor's Take

This single tweet is hard to dismiss, because of the timing relative to Musk's other ventures. SpaceX has already deployed a satellite communications network in Earth orbit; that same SpaceX recently delayed the removal of xAI's unpermitted turbines and has also completed its acquisition of Cursor. The fact that the person calling orbital computing "the only way" also happens to found the company that can launch payloads into space most cheaply invites reading this not simply as a forecast, but as a move positioning assets he already controls.

Viewed over time, this isn't a new idea. The concept of orbital data centers has circulated for years among academics and startups, on the logic that solar power generation is far more efficient in space than on the ground and that cooling can take advantage of the vacuum of space. But until now, it remained a "someday" idea, blocked by launch costs and communication latency. What's different this time is that a specific date — 2029 — has been attached to that "someday." As entities like Mach33 put out simulations with concrete numbers — 500GW, $3.2 trillion — the discussion is shifting from speculation to calculation.

Practically speaking, there isn't much domestic companies can do right now. Orbital computing requires an entire stack of satellite manufacturing, launch, and communication infrastructure, leaving little room for ordinary companies to participate. What this debate does suggest, though, is that power should be the very first item checked in any AI business plan. It's no longer just about how many GPUs you can secure — it's about when you can get the power and site permits to run those GPUs, and that timeline is what's actually determining service launch schedules.

In the coming weeks, this tweet is likely to be cited in a wave of funding announcements from orbital data center startups. At the same time, friction between big tech and local regulators over ground-based power and permitting is expected to keep making headlines. Whether orbital computing actually materializes by 2029 remains uncertain, but this remark has once again confirmed that, until then, power is the real bottleneck in the AI industry.