The Orbital Gold Rush: Why Musk, Bezos, and a Growing Chorus of Scientists Are Battling Over Data Centers in Space

Somewhere between the atmosphere and the void, a new front in the data center arms race is taking shape. Elon Musk and Jeff Bezos — the two richest men on the planet, perpetual rivals in rocketry and ego — are now racing to plant computing infrastructure in orbit. The pitch is seductive: unlimited solar power, natural cooling from the vacuum of space, and freedom from the land-use constraints that have turned earthbound data center construction into a political flashpoint.

But a growing number of scientists, astronomers, and environmental researchers are asking a question that the billionaires seem uninterested in answering: Why?

Not why in the abstract. Why now, why this way, and why with so little public scrutiny of the environmental, scientific, and regulatory consequences. The ambitions are enormous. The evidence that orbital data centers make economic or environmental sense remains, at best, thin.

The Case for Computing Beyond Earth

The logic, on the surface, isn’t crazy. Global data center energy consumption is surging — driven by artificial intelligence workloads that demand staggering amounts of electricity and cooling. In the United States alone, data centers are projected to consume up to 12% of total electricity by 2028, according to estimates from the Department of Energy. Utilities are scrambling. Grid operators are warning of shortfalls. Communities near proposed data center sites, from northern Virginia to rural Oregon, are pushing back against the noise, water use, and strain on local power infrastructure.

Space, by contrast, offers an environment where solar energy is available nearly 24 hours a day, unfiltered by clouds or atmosphere. Cooling — one of the most expensive and energy-intensive aspects of running a data center on Earth — could theoretically be handled by radiating heat into the near-absolute-zero temperatures of space. No neighbors to complain. No aquifers to drain.

As Business Insider reported, both SpaceX and Blue Origin have been exploring concepts for orbital computing platforms. Musk’s Starlink constellation already represents the largest commercial satellite network in history, with more than 6,000 satellites in low Earth orbit. Adding computing nodes to that network — or building dedicated orbital server farms — is a logical extension of SpaceX’s existing infrastructure ambitions. Bezos, through Blue Origin and Amazon’s Project Kuiper satellite broadband initiative, is pursuing a parallel track.

Smaller players are in the mix too. Lumen Orbit, a startup backed by Y Combinator, has publicly stated its intention to launch data center satellites. The European Space Agency has funded studies on in-orbit cloud computing. And multiple defense contractors are exploring the concept for military and intelligence applications where data sovereignty and physical security are paramount concerns.

The momentum is real. But so are the objections.

Scientists quoted by Business Insider didn’t mince words. The fundamental problem, several researchers argued, is that the energy economics don’t actually work in space’s favor once you account for the full lifecycle — particularly the colossal energy cost of launching hardware into orbit. A single Falcon 9 launch burns through roughly 440 tons of RP-1 kerosene and liquid oxygen. The carbon footprint of putting a server rack into space dwarfs the emissions from simply plugging that same rack into a terrestrial grid, even one powered partly by fossil fuels.

There’s also the matter of latency. For AI training and real-time inference — the workloads driving most of the current data center buildout — milliseconds matter. The speed-of-light delay inherent in communicating with orbital infrastructure makes space-based computing fundamentally unsuitable for many of the applications that are actually consuming power on the ground right now.

And then there’s the debris problem.

Low Earth orbit is already dangerously congested. The European Space Agency tracks more than 36,000 objects larger than 10 centimeters. Millions of smaller fragments — bolts, paint flakes, shrapnel from past collisions — travel at velocities that can turn a speck of metal into a bullet. Adding thousands of server satellites to this environment increases collision risk, which increases debris, which increases collision risk further. It’s a feedback loop that orbital mechanics researchers have been warning about for decades, known as the Kessler syndrome.

Astronomers have already seen their ground-based observations degraded by Starlink satellites streaking across telescope fields of view. A 2023 study published in Nature Astronomy documented the impact on wide-field surveys critical to tracking near-Earth asteroids — the very objects that pose actual existential risk to civilization. More hardware in orbit means more light pollution in the night sky, more radio frequency interference, and more obstacles for the space science missions that both Musk and Bezos claim to support.

Follow the Money, Not the Mission Statements

Strip away the futurism and what emerges is a familiar pattern. Both SpaceX and Blue Origin are launch companies. They make money — or in Blue Origin’s case, aspire to make money — by putting things into space. Creating demand for orbital infrastructure creates demand for launches. Data centers in space would be, above all else, a spectacular customer for the very rockets these companies build.

This isn’t conspiracy. It’s business strategy.

SpaceX’s Starship, currently in development, is designed to dramatically reduce per-kilogram launch costs. If Starship achieves anything close to its cost targets, the economics of putting large payloads into orbit shift significantly. But “significantly cheaper than before” is not the same as “cheaper than building on the ground.” The baseline cost of terrestrial data center construction — roughly $7 million to $12 million per megawatt of IT capacity, depending on location and tier — remains far below even the most optimistic projections for orbital equivalents.

There’s also the maintenance question. Servers fail. Hard drives die. On Earth, a technician swaps a component in minutes. In orbit, a failed server module either gets replaced by an expensive robotic servicing mission or becomes space junk. The operational realities of running computing infrastructure where no human hand can reach it are, to put it mildly, nontrivial.

Recent reporting adds context. The AI boom has intensified pressure on power grids worldwide, with companies like Microsoft, Google, and Meta signing unprecedented power purchase agreements — some exceeding a gigawatt — to secure electricity for new data center campuses. Microsoft has even signed a deal to restart a unit at Three Mile Island. Google has inked agreements for small modular nuclear reactors. The hunger for power is genuine and growing.

But the response from the terrestrial data center industry has been rapid. Modular construction techniques are compressing build timelines. Liquid cooling systems are slashing energy overhead. And the geographic footprint is expanding — new facilities in Scandinavia, the Middle East, and Southeast Asia are tapping into diverse power sources and cooler climates. The ground-based industry isn’t standing still while the rocket companies pitch orbital alternatives.

So where does that leave the space data center concept? Likely in a niche. There are specific, narrow use cases where processing data in orbit makes sense — Earth observation, where satellites generate enormous volumes of imagery that could be partially processed before downlink, reducing bandwidth requirements. Military and intelligence applications where physical inaccessibility is a feature, not a bug. Perhaps some edge computing for satellite communications networks themselves.

But as a wholesale replacement for terrestrial data centers? The physics, the economics, and the environmental math all argue against it. The scientists quoted by Business Insider aren’t Luddites or anti-space ideologues. Many of them have spent careers supporting space exploration and orbital science. Their objection isn’t to space activity per se. It’s to the notion that orbital data centers represent a sound solution to a problem that has more practical, less risky answers available on the ground.

The broader concern — and it’s one that extends well beyond data centers — is that the commercial space industry has grown powerful enough to reshape orbital environments before regulatory frameworks catch up. The Federal Communications Commission licenses satellite constellations. The Federal Aviation Administration oversees launches. But no single agency has clear authority over the cumulative environmental impact of tens of thousands of new objects in orbit, nor is there an international body with enforcement power over orbital debris mitigation.

The Outer Space Treaty of 1967, the foundational document of space law, was written for an era of government-run space programs launching a handful of missions per year. It says nothing about private companies filling low Earth orbit with commercial computing hardware. And while the United Nations Committee on the Peaceful Uses of Outer Space has issued guidelines on debris mitigation, compliance is voluntary.

Musk and Bezos are not waiting for regulators to figure this out. They rarely do. The pattern — move fast, build the infrastructure, then negotiate the rules from a position of established fact — has defined both men’s careers. It worked for Amazon in e-commerce and for Tesla in electric vehicles. Whether it’s appropriate for an orbital commons shared by every nation on Earth is a different question entirely.

None of this means space-based computing will never happen. Technology costs fall. Launch economics improve. Novel cooling and power systems may emerge that change the calculus. But the current push feels less like a response to genuine technical necessity and more like a solution in search of a problem — one that conveniently generates revenue for the companies proposing it.

The scientists asking “why” deserve an answer that goes beyond press releases and investor decks. So far, they haven’t gotten one.