Space Based Data Centers: How Orbital AI Compute Works

Space based data centers sound like a science fiction premise until you check the numbers on both ends: the terrestrial bottleneck is measured in hundreds of billions of dollars, and the orbital advantage is measured in physics. The pitch currently carrying this idea to a mass audience is Dylan Jovine’s “Elon’s 100X AI Stock” from Behind the Markets. Here is how orbital AI compute actually works, what is real, and what the presentation adds on top.

Why Anyone Would Put a Data Center in Orbit

The push is the AI power squeeze. “The four biggest names in tech, Microsoft, Amazon, Google and Meta, are pouring a combined $725 billion into AI infrastructure in 2026 alone,” the pitch says, with 16 gigawatts of new U.S. capacity opening this year. The wall is electricity: Morgan Stanley projects “the U.S. will be short 49 gigawatts of electricity for data centers by 2028,” Google “now waits up to 12 years to get a new data center hooked up to the grid,” and local resistance has turned into policy, with Denver, Oklahoma City, and Tulsa pausing new data centers and “Maine’s legislature voted to ban them outright.”

The pull is orbital physics. “A solar panel in space generates roughly eight times the energy of the same panel on the ground”: no night, no weather, no atmosphere thinning the light. “Space is 250 degrees below zero, and it never warms up”: a permanent cold sink for the radiator side of the thermal system. The pitch’s own Musk quotes carry the thesis: “The cheapest place to put AI will be in space within 36 months or less,” and “Space-based AI is obviously the only way to scale.”

The Concrete Machine: AI1

The pitch’s artifact is the AI1 satellite: “Its solar wings stretch 230 feet from tip to tip. That’s longer than the wingspan of a Boeing 747.” The design logic follows the physics: giant solar wings for the eight-times power advantage, radiator surfaces sized for heat that can only leave by radiation (a vacuum carries nothing away by convection), and a compute core built around power-efficient silicon, because in orbit, efficiency is not a preference; it is what determines how much AI a given mass of satellite can run. Musk’s own scale claim, quoted in the pitch: “as you can scale a trillion times more than you could on Earth.”

The Real Projects Underway

The idea is not one company’s fantasy. The pitch names a real field, and the names check out: the Nvidia-backed Starcloud has trained the first AI model in space, the proof-of-concept moment for orbital compute; Bezos is “pouring $1 billion of his own money into Blue Origin every year” and has spoken publicly about space-based data centers; Google’s Project Suncatcher is pursuing orbital solar power; Meta has a space-solar arrangement in motion; and Musk’s AI1 plus the FCC request for one million satellites (versus roughly 10,000 in orbit today) is the largest stated ambition in the field. We have tracked orbital data centers as an investment theme across several campaigns, and the participant list has only grown.

The Investment Angle Jovine Adds

The genuinely distinctive move in this pitch is where it points the buyer. Not the satellite operator, not the rocket company: the chip supplier. “He has the rockets. He has the launch sites. He has the AI… There’s just one thing he can’t build himself. The chip that runs them.” The payment structure: “Elon doesn’t own that chip. He licenses it. That means, every time a Tesla rolls off the line… Every time SpaceX launches an AI1 satellite… This company gets paid.” Jovine dates the supplier relationship to 2012 and argues that a 100X satellite fleet means a proportionally larger royalty bill, hence “Elon’s 100X AI Stock.”

The company is unnamed; its name and ticker are inside Report #1 of the subscription. The clues point toward a chip-IP licensor on a royalty-per-unit model, whose history, the pitch says, includes a blocked NVIDIA acquisition attempt. That identification is unconfirmed; we treat it as a fingerprint, not a fact. What is verifiable is the model itself: architecture licensing with per-unit royalties is a real, well-documented business, and the switching-cost argument for licensor stickiness (rewriting “every line of software in every Tesla and every satellite he owns”) is economically sound.

The Date Attached to the Story

The pitch’s urgency is December 8th, the final SpaceX lockup tranche: “The last block of roughly 800 million insider shares comes loose, all at once. That’s more than $100 billion at the IPO price.” Jovine’s argument is that freshly liquid insider capital chases the suppliers insiders already know. We examined the event and its mechanics in our SpaceX lockup expiration coverage.

Our Honest Read

How orbital AI compute works is no longer in question: the physics is real, the first training run has happened, and the terrestrial bottleneck deepens every quarter. In question are timelines (“36 months or less” is Musk’s, not a construction schedule) and the specific supplier claims, which cannot be checked before subscribing. The offer is $49 for a year of Behind the Markets with a six-month full-refund guarantee and four research reports. If you want the argument in full, the presentation is here.

NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.