Sophia Space secures funding to test new orbital computers

The "crazy" idea about Space based data centers is gaining traction.

Sophia Space pulled in $10 million to test a bold idea that sounds like science fiction but could become very real for investors.

What if the next data center stock winner isn’t building on Earth at all?

That’s the bet. Sophia wants to prove that high performance chips can run in orbit without the massive cooling systems that eat up power and profits here on the ground. If they’re right, space doesn’t just solve the heat problem. It turns heat into an advantage.

Investors are increasingly asking a bigger question: Should you invest in data centers in space as launch costs fall and orbital infrastructure matures?

The plan is straightforward. Run ground tests, buy a satellite bus from Apex Space, then launch a demo mission by 2027 or 2028.

Cooling is the bottleneck. On Earth, data centers rely on air to move heat away from chips through convection. In orbit, there’s no air. That means no fans pushing heat around. You’re left with conduction through solid materials and radiation into space. Big players like SpaceX, Google, and Starcloud typically use large radiators bolted onto boxy satellites to dump heat.

The idea of orbiting compute isn’t isolated to one startup. Even Elon Musk has floated plans for space data centers that could eventually operate beyond Earth’s grid constraints.

Sophia is taking a different route.

The team’s roots go back to a $100 million solar power satellite program at California Institute of Technology, connected to work at Jet Propulsion Laboratory. Instead of a cube shaped satellite, that program developed a thin, sail like structure. Sophia adapted that concept into flat 1 meter by 1 meter modules called TILES. Chips sit directly against passive heat spreaders. No fans. No pumps. Less energy wasted on cooling.

As a partner of Nvidia, Sophia says 92% of generated power goes straight to processing instead of cooling and overhead. That’s a big claim in a world where energy costs can make or break a data center business model.

The long term vision is thousands of TILES assembled into a 50 meter by 50 meter array delivering 1 megawatt of compute power in orbit.

If Sophia can prove this works, it won’t just be another space story. It could open the door to a new category of space computing stock plays built around power efficiency, not just rockets. That theme is gaining momentum as major players integrate AI and launch capabilities, highlighted by moves like SpaceX acquiring xAI in a deal valuing the company at over $1 trillion, signaling how tightly compute and space infrastructure may become linked.

The disruption behind the news: Data centers are moving into orbit.

This isn’t about a cooler satellite.

But about moving the data center to where the data is born.

That flips the economics of Earth observation, defense, and communications.

Right now, satellites collect more data than they can send back to Earth. Bandwidth is limited. Downlink windows are short and tightly scheduled. So operators discard large portions of raw imagery and sensor feeds. That’s lost value already paid for at launch.

If processing happens in space, you transmit answers instead of raw data. A wildfire alert instead of terabytes of infrared video. A missile track instead of nonstop radar streams. That can cut bandwidth needs by orders of magnitude and increase the value of each satellite without launching more hardware.

The non-obvious incentive shift is that downlink isn’t just a tech constraint, but a priced and scheduled asset. A typical Earth observation satellite might only get about 10 minutes per pass over a ground station. Even at 1 Gbps, that’s roughly 75 GB per pass. If the payload generates 1 TB per orbit, you’re forced to throw away more than 90% of what you collect, or pay for more ground stations, more spectrum coordination, and more operational complexity. Onboard compute changes the unit economics because every 100× reduction in data to transmit is economically similar to adding about 100× more downlink capacity. And you don’t need to buy spectrum, build new stations, or compete for scarce ground network time. That’s why compute in orbit is really an attack on the scarcity rent embedded in downlink, not just a thermal engineering upgrade.

Launch costs have fallen roughly 90% over the past decade. Compute density keeps improving. Solar power in orbit is abundant compared to constrained power grids on Earth. If Sophia can truly dedicate 92% of power to chips, the cost per processed bit in orbit could fall quickly. At the same time, terrestrial data centers are running into grid bottlenecks and rising energy prices.

A 1 megawatt orbital cluster isn’t trivial. That’s the scale of a small data center. Put that directly above Earth’s sensing layer and you eliminate latency and transmission bottlenecks. Defense agencies will likely care first because they value speed and autonomy. Commercial imaging and telecom companies could follow if the economics prove out.

What to watch next

The demo mission will determine whether this is physics or fantasy.

Thermal performance in real orbital conditions is everything.

If passive cooling holds up, capital could move in quickly.

Watch for satellite operators signing early supply agreements for TILES before 2028. Pre commitments would signal real demand, not just technical curiosity. Also monitor regulatory signals around in orbit processing for defense workloads. A supportive posture could accelerate contracts by years.

Finally, track power to compute efficiency metrics. If Sophia delivers anywhere near 90% usable power in orbit, terrestrial hyperscalers will pay attention. The moment orbital compute becomes cheaper per useful workload than a grid constrained Earth data center, the shift begins.

We’re entering a world where data isn’t just created on Earth anymore. It’s being generated in space. And if that’s where the data starts, it makes sense that processing will eventually move there too. The model extends beyond data centers. Companies are already experimenting with orbital manufacturing, such as Space Forge launching semiconductor factories in space, reinforcing the idea that high value compute and chip production could increasingly migrate off planet.

That shift could kick off the next big infrastructure cycle. As space-based data centers scale and prove their economics, the companies that see it early and build for it could be in a powerful position.

This is a megatrend investors shouldn't just dismiss. Yes, it's early. But it's most definitely where the puck is moving next.