Satlyt raises $8M to turn satellites into shared AI computers in orbit

Updated October 1, 2026 3:54 pm

Satlyt, a startup founded by former Google Cloud and SpaceX product manager Rama Afullo, has raised an $8 million seed round to build software that lets satellites run AI workloads in orbit. The company’s bet matters because it could reduce the cost and delay of sending data back to Earth, while helping spacecraft share compute resources instead of depending entirely on ground stations.

The Nairobi- and Sunnyvale-based company is positioning itself as a software layer for orbital computing rather than a satellite manufacturer. Its technology is set to fly on a SpaceX launch this week, alongside Google’s first prototype for Project Suncatcher, as interest intensifies around the idea of turning satellites into distributed data centers.

Why Satlyt thinks the next cloud platform could be in space

Satlyt’s core thesis is that many satellites already generate more data than they can efficiently downlink, and that some of that work should happen onboard. By processing information in space, satellites can make faster decisions, reduce transmission costs and reserve limited bandwidth for the most valuable data.

Afullo says that experience at both Google and SpaceX convinced him that orbital computing was inevitable, even if neither company wanted to pursue the idea internally at the time. His startup’s argument is not that every satellite should become a full data center, but that satellites can be made smarter, more autonomous and more commercially useful if they can run AI models locally.

Afullo said both companies passed on his early pitches for space-based computing, which pushed him to leave and build Satlyt instead. He now believes the same market is maturing in front of him.

The startup’s pitch is especially timely because the industry is shifting. SpaceX has already moved into the orbital data center conversation, while Google is now testing a prototype through Project Suncatcher. Satlyt’s value proposition is different: it wants to be the software platform that works across many companies’ spacecraft.

What exactly does Satlyt build?

Satlyt is developing software that orchestrates AI and compute workloads across satellites. Instead of designing the hardware itself, the company wants to provide the software stack that makes orbital processing practical for satellite operators, payload providers and future space-computing customers.

Afullo compares the company’s role to infrastructure platforms that abstract away complexity for developers and enterprises. In his framing, the big satellite manufacturers and orbital data-center builders are creating the hardware equivalent of the device, while Satlyt aims to supply the operating system-like layer that lets different satellites work together.

How is Satlyt different from SpaceX or Google?

Satlyt’s approach is software-first, and that distinction is central to its business model. SpaceX, Google and several other startups are exploring hardware-heavy orbital compute systems, while Satlyt says it plans to remain compatible with a broader range of satellites rather than tying itself to a single constellation or spacecraft design.

The company says that makes it closer to a cloud platform than a satellite maker. That model could be attractive to operators that want access to AI capabilities without having to redesign spacecraft or build their own in-house onboard compute systems.

Why onboard AI matters for satellites

Satellites often rely on ground control teams to diagnose anomalies, process sensor data and decide what information is worth transmitting. That workflow can be slow and expensive, especially when downlink bandwidth is constrained or contact windows are short. Putting more intelligence on the spacecraft itself can make those operations more efficient.

For many missions, the economics are straightforward: if a satellite can classify images, compress logs or identify faults before sending data down, it can cut transmission costs and speed response times. Satlyt argues that these savings can add up to substantial annual benefits for operators.

The company says one of its early demonstrations showed that an onboard version of Google DeepMind’s Gemma model reduced the size of a transmission related to software errors by more than 60%. Satlyt says that kind of improvement can translate into significant savings per satellite each year.

What kinds of work can be done in orbit?

Satlyt’s current focus is on workloads that improve satellite operations rather than highly compute-intensive public cloud tasks. That includes anomaly detection, image processing, sensor analysis and other forms of local inference that are useful even with modest onboard hardware.

Right now, the market for high-end GPUs in space remains limited. That means the near-term business opportunity is likely to come from practical, lower-latency applications that reduce operational friction rather than from fully fledged orbital supercomputers.

What is launching this week?

Satlyt’s software is scheduled to fly on a SpaceX mission this week aboard a spacecraft built by TakeMe2Space, an Indian startup focused on computing hardware for satellites. The launch is designed to show that Satlyt’s software can operate on real spacecraft under real mission conditions.

The mission has three customers, each testing a different use case. NASA is paying to evaluate protocols for cloud computing in space. Stellerian, a startup focused on space surveillance, wants to test image-processing workloads. TakeMe2Space itself is using the mission to demonstrate that its satellite can host third-party software.

The launch is important because it moves Satlyt from demonstration flights toward a broader product story: not just that AI can run on a satellite, but that software can coordinate workloads across multiple spacecraft.

How is the company funded?

Satlyt’s new seed round was led by Non Sibi Ventures, a Houston-based firm. The round includes backing from a venture team that includes Bernard Harris, the former NASA astronaut and physician who has long advocated for space-related innovation.

That investor mix reflects a belief that the company’s business case does not depend on the most ambitious version of orbital infrastructure. In other words, Satlyt does not need to wait for a future where giant data centers cover low Earth orbit before it can generate revenue.

Non Sibi partner Kent Lucas said the startup could succeed simply because more satellites are launching every year, even if full-scale space data centers remain uncertain.

That view is important in a sector where economics are still being tested. Large orbital data centers face obvious hurdles: launch costs, spacecraft reliability, power generation, thermal management and the challenge of building enough demand to justify the infrastructure. Satlyt is trying to avoid being trapped by that timeline.

Why investors think the model can work now

Satlyt’s investment case is built around a more incremental path. Rather than betting everything on giant hardware projects, the company is targeting software use cases that can ride on satellites already headed to orbit. If that works, Satlyt could scale as the number of spacecraft rises.

The company’s leaders see that as a way to become a managed service provider for satellite operators. In that vision, spacecraft are not just passive sensing tools; they become revenue-producing assets capable of hosting workloads, running applications and sharing compute across networks.

That framing may also make the company easier to finance than a hardware-heavy orbital infrastructure play. Investors can evaluate software adoption, mission performance and customer interest without having to underwrite an entire space data-center buildout from day one.

What is the long-term ambition?

Satlyt wants to become an operating layer for distributed compute in orbit. Its next major step is to demonstrate a shared computing environment spanning two satellites, a milestone the company hopes to attempt next year.

If that succeeds, Satlyt would move closer to a true cloud architecture in space, where separate spacecraft can share workloads, exchange data and support outside customers as part of a broader network.

That would be a significant development because it would show that orbital compute does not need to be confined to one satellite or one vendor’s hardware. Instead, it could be treated more like a service layer with interoperability across spacecraft designs.

Who is Satlyt trying to sell to?

Satlyt is targeting satellite builders, operators and mission owners that want to reduce operational costs or unlock new services from their spacecraft. It also sees potential customers in government and commercial users that need local processing in orbit for surveillance, scientific analysis or cloud experimentation.

The company’s broader ambition is to become infrastructure for a market that is still forming. If satellite manufacturers increasingly install GPUs and other advanced processors onboard, Satlyt wants to be the software layer that makes those chips useful across fleets.

How big is the opportunity?

Afullo believes the market could grow rapidly as satellite launches increase and onboard processing becomes standard. He has said he expects the company to be on 20% of satellites by the end of the decade, an ambitious target that reflects how early the category still is.

That goal may sound aggressive, but it aligns with a larger shift in the space industry: spacecraft are becoming more like edge-computing nodes than passive hardware. As cameras, sensors and processors improve, the incentive to analyze data in orbit grows stronger.

For Satlyt, the next few months will matter more than the long-range forecast. The company needs successful missions, proof that customers will pay for onboard software services and evidence that its distributed approach can work across multiple spacecraft.

Why this startup matters now

Satlyt is notable because it sits at the intersection of two major technology trends: the rise of AI agents and the push to move more computation closer to where data is created. In space, that means less dependency on ground stations and more autonomy onboard satellites.

It is also part of a broader race that now includes giants and startups alike. As major players test orbital compute concepts, the market may open for software companies that can make those systems interoperable, efficient and commercially practical.

For now, Satlyt remains early. But its seed funding, flight-ready software and customer-backed demonstration mission suggest the company is aiming to be more than a concept: it wants to become part of the operating fabric of the next generation of satellites.

Key milestones in Satlyt’s trajectory

What happens next?

If the upcoming launch succeeds, Satlyt will have another proof point that its software can support AI workloads in orbit. The bigger test will be whether it can extend that capability across satellites and turn orbital processing into a repeatable service.

That is the part investors appear most eager to watch. The company is not promising to build the next giant satellite or the next space station. It is promising to help make every satellite smarter, more efficient and potentially more profitable.

In a space industry increasingly defined by software, autonomy and distributed infrastructure, that could be a compelling place to stand.