When the Hawthorne-founded Space Exploration Technologies Corp. filed to go public in April, it said it would put as many as a million satellites in space – to host data centers.

Yes, data centers.

Data centers have become one of the most contested pieces of real estate in America. A growing list of space companies thinks the fix is to get them off the planet entirely.

The impulse makes sense. A 2026 Gallup survey found that 70% of Americans oppose data centers in their communities, citing artificial intelligence’s never-ending hunger for power, a potential uptick in energy bills, and worries surrounding local pollution and long-term climate change damage. Residents in Monterey Park, El Segundo and Alhambra have blocked proposed data centers in their neighborhoods.

While “space data centers” sounds like three buzzwords mushed together, orbital computing systems aren’t as far-fetched as it sounds. As the cost of spaceflight continues to fall, companies are looking into opportunities to bring Earth endeavors – like mining, manufacturing and drug development – to space.

So why not data centers?

“The road to get to low-Earth orbit has gotten much more accessible to more than just governments,” said Robert DeMillo, the chief executive of orbital computing company Sophia Space. “Now it’s less hard to get there, so you can start to have meaningful conversations around commercial endeavors like factories in orbit or medical development in orbit.”

The idea is gaining traction at a time when hostility to data centers is intensifying. Though every industry is rapidly adopting artificial intelligence to bolster cybersecurity, improve manufacturing capabilities and make movies, that growth has come with a backlash on the ground. But as investments pour into orbital computing systems and the space economy rapidly develops, companies focused on space data centers are navigating a vague, unproven market largely dominated by SpaceX.

“Can we actually execute on this? Is this something beyond a proof of concept?” PitchBook’s emerging space research analyst Ali Javaheri said. “I think it all comes back to whether or not people can actually make money off of this.”

Seeding a new sector

In 2020, Mandala Space Ventures, the California Institute of Technology and the Jet Propulsion Laboratory received a $100 million grant to develop an orbital solar power station capable of beaming solar energy back to Earth. After 18 months, the team developed the initial patents and technical theories behind it. It could work, they said, but the economics might not make it worthwhile.

That’s when Leon Alkalai, the founder of Mandala and a decades-long fellow at JPL, posed a simple question: What if they put a server on it?

That was the start of Sophia Space, an orbital data center company founded in Pasadena. The company announced in late July that it secured a patent with Caltech on an orbital data storage and computing design that was modular, scalable, and powered by space.

Legacy companies and new startups have both entered the race to build space data centers. Sophia Space, along with Starcloud, Axiom Space and downtown Los Angeles-based Orbital have joined large players like Jeff Bezos’ Blue Origin, Nvidia Corp. and Alphabet Inc.’s Google to tackle the logistics of putting data centers in space.

The community is growing. Aetherflux, a solar energy company, pivoted to orbital data centers and rebranded as Cowboy Space Corp. in May. SpaceX said in its S-1 in April that it would build orbital data centers with the influx of cash from its initial public offering.

“Launching a million satellites that operate as orbital datacenters is a first step towards becoming a Kardashev II-level civilization,” SpaceX said in its filing. “One that can harness the sun’s full power-while supporting AI-driven applications for billions of people today and ensuring humanity’s multiplanetary future amongst the stars.”

Back to basics

Space may be the answer to the data center contention. Latency certainly isn’t an issue, given that much of the world’s navigation devices, food delivery apps and streaming services are powered by a constellation of satellites without issue. And these orbital computing systems can bask in an abundance of solar power, leaving earthly elements untouched.

But compressing a football-field-sized data center into a payload that can survive launch is no easy feat. For one, data centers on Earth are equipped with massive cooling systems to prevent chips from overheating. In space, there’s no air to carry heat away from the unit, so companies must design novel cooling techniques that often rely on radiation.

“Managing thermal solutions on the satellite is one of the key challenges,” Euwyn Poon, founder of Orbital, told the Business Journal in April. “We don’t have convection or conduction up there. So, redesigning how chips are being cooled for the Orbital use case is a neat and fun engineering challenge.”

Although commercial spaceflight is the cheapest it’s ever been, it’s still not exactly affordable. Every part of the data center payload needs to be refined to be as light, as small, and as energy efficient as possible.

Orbital’s plan is to deploy a constellation of thousands of satellites, each the size of a refrigerator, in low Earth orbit. Each satellite will carry servers and tennis-court-sized solar panels to capture energy, as well as a proprietary radiator to cool the servers.

Sophia Space’s model is designed to be customizable and scalable. Its servers are built so that the central processing unit, graphics processing unit, memory and batteries are pressed against the radiator, allowing heat to immediately transfer away from the server. Four servers sit on a “tile,” or computing unit, and the company can snap tiles together as needed.

While space data centers are touted as an environmentally friendly option, there are significant limitations. AI training data centers, which process big datasets to build the large language model’s intelligence layer, are extremely energy-intensive and remain too cumbersome to put into space. The majority of first movers are instead focusing on inference, whereby an AI program creates a result based on the prompt that instructed it. These data centers use far less energy than their processing counterparts.

“We’re putting up inference engines,” DeMillo said. “What will happen initially is that LLM training will take place on the planet, and then, once the models are established through the training, those models will be moved up to the satellites for use.”

Addressing the Goliath

Though a rush of smaller companies is hard at work to make these AI space data centers a reality, they are up against a formidable competitor: SpaceX.

Launch costs remain one of the most significant barriers to getting large payloads into space. SpaceX, however, has one of the most affordable space delivery strategies via Starship, a rocket made cheaper through mass production and reusable parts. Starship can carry more than 100 metric tons to orbit, allowing it to make fewer trips per mission.

SpaceX also acquired xAI in February, and the company’s Starlink internet service already supports xAI’s distributed inference. In a PitchBook analyst note, senior research analyst Franco Granda said the merger “formalizes Musk’s shift toward a vertically integrated infrastructure conglomerate.”

Because these other data center companies are primarily iterating on the data center systems, they are likely to rely on existing space infrastructure through partnerships with other companies.

“These companies don’t really have much of a defensive vote beyond that they might re-engineer thermal radiation systems,” Javaheri said. “It is going to be really hard for these companies to compete. And one of the only ways they can is if they provide really specialized services.”

Take Rocket Lab Corp., a Long Beach-based launch and satellite manufacturing company which, like SpaceX, is in the reusable rocket space. The organization acquired satellite communications network Iridium Communications Inc. in late June, allowing it to own both launch and spectrum, like SpaceX with Starlink. But because Iridium owns spectrum that is used only in remote environments, the company operates in a different market entirely.

“That’s where these companies can plug themselves in and won’t directly become competitors with what SpaceX wants to do,” Javaheri said. “There’s an opportunity here, but it’s not going to be a robust market where there are all these orbital data center players springing up.”

Today, Sophia Space is starting with edge computing devices – systems that process data at the same location where it is gathered, rather than sending it to an external cloud server. It’s a more immediate application of its orbital data center technology and works for applications that need to analyze data immediately, like tracking environmental catastrophes, air traffic control, or missile tracking.

“It’s all those things that you could do in orbit because you’ve got these great sensors that are looking out at the planet,” DeMillo said. “But there’s no way to do processing in orbit yet.”

Here to stay

When Javaheri first heard about launching data centers in space, he said that he thought, “All right, some guys were in a dorm room thinking, ‘What if we just put data centers in space?’”

Poon echoes that sentiment. He first heard about space-based data centers on a podcast interview with SpaceX Chief Executive Elon Musk and initially dismissed it as a sci-fi dream.

“But I think AI is larger than most people realize. It’s not just chatbots; it’s really powering an entirely new generation of compute. It’s going to be underlying a lot of applications today, both consumer applications, enterprise systems, and running more to the world,” Poon said. “I think that our demand for computing is just going to be insatiable. We find a way to use up all the bandwidth. Humans are very clever at that.”

Establishing concrete use cases for space-based data centers is key to boosting this growing industry, said Javaheri. Industries such as defense, which need to monitor U.S. satellites for national security, are early adopters of frontier space technologies.

“You also have to always compare this to terrestrial data centers,” Javaheri said. “The whole point is that there’s not enough physical space and energy for our data centers on Earth, so we just push into space. But if the economies of scale work on Earth, then the whole space-based data center thing feels like a rug pull.”

DeMillo admits that, initially, orbital data centers will be more expensive. But Los Angeles is littered with companies looking to manufacture parts, mine asteroids, and develop pharmaceuticals in space. As more commercial enterprises reach for the stars, these agencies will need to compute in space.

Though today’s technology is limiting, DeMillo sees a future in which space logistics will be robust enough to host AI training models and perform other background processing tasks that don’t require immediate human moderation, such as mining cryptocurrency and rendering films.

“I’m old enough to remember the birth of cloud computing. Every time, there were the naysayers – ‘it’s just too expensive, who the hell would do that? What’s the point?’” DeMillo said. “I basically ignore the naysayers. Look, this is happening. You can feel the commercial pressure. You can feel the government pressure. You can feel the civil side of this as well. All these vectors are pointing towards just getting this done.”