What if satellites did not have to send everything they collect back to Earth before figuring out what matters?

That is the problem Nairobi- and Sunnyvale-based space-tech startup Satlyt is trying to solve.

The company has raised $8 million in seed funding to develop software that allows satellites to process data and run artificial intelligence models directly on computers already onboard spacecraft.

The round was led by non sibi ventures, with participation from TLCOM, Antler, Slauson & Co., Launch Africa Ventures, Enza Capital, Askya Investment Partners, Axian Investment, Gaingels and other investors.

Founded in 2024 by Rama Afullo, a Kenyan-American engineer who previously worked in Google's cloud business and at SpaceX's Starlink unit, Satlyt is taking a different approach from companies building dedicated orbital data centres.

It is not building rockets.

It is not building satellites.

It is building the software layer that can make satellites more useful once they are already in orbit.

Satlyt Wants Satellites to Process Data Before Sending It Home

Satellites generate enormous amounts of information, including imagery, sensor readings, telemetry and system logs.

Traditionally, much of that information has to be transmitted back to Earth for analysis.

That creates a bottleneck.

Communications bandwidth is limited, transmitting information from space can be expensive, and operators may have to wait before receiving the data they actually need.

Satlyt's approach is to move some of that computing closer to the source.

Instead of sending every piece of raw information to Earth, an AI model running onboard can analyse the information and determine what is important enough to transmit.

That is essentially edge computing in space.

And it could be particularly useful for satellite health monitoring, Earth observation, imagery analysis and other applications where large volumes of raw data are generated.

The Startup Doesn't Need to Build a Satellite

This is where Satlyt's business model becomes particularly interesting.

Companies such as SpaceX and other emerging space-computing startups are pursuing purpose-built infrastructure for computing in orbit.

Satlyt is taking a software-first route.

Its software can be deployed on spacecraft operated by other companies, allowing satellite owners and application developers to use the computing resources already available onboard.

Founder Rama Afullo has compared the strategy to software companies that allow customers to use existing hardware without having to build the underlying infrastructure themselves.

The long-term ambition is even bigger.

Satlyt wants its software to eventually coordinate computing resources across satellites operated by different organisations, creating something resembling a distributed cloud in orbit.

But that is still a future goal.

Today, the company's work is primarily focused on computing within individual spacecraft.

Satlyt Has Already Put AI in Orbit

The company is not starting from a laboratory demonstration.

Satlyt has already flown its software on spacecraft.

In September 2026, the company deployed Google's Gemma AI model onboard a Momentus spacecraft.

The system was used to analyse spacecraft telemetry and software-error information in orbit.

According to reporting on the demonstration, Satlyt reduced the size of a transmission concerning onboard software errors by more than 60%. The company has also reported reductions of up to 97% in telemetry data under certain workloads.

Those numbers are significant because less data travelling from a satellite to Earth can mean lower communication requirements and potentially lower operating costs.

However, they should be understood as results from Satlyt's reported demonstration rather than a universal performance benchmark for every satellite or AI workload.

The actual benefit will depend on the spacecraft, processor, application and type of data being processed.

The Next Step Is More Software in Orbit

Satlyt's latest deployment involves software running on a spacecraft built by TakeMe2Space, an Indian startup developing computing infrastructure for satellites.

The deployment is designed to demonstrate that a spacecraft can host software from multiple customers, including research and commercial applications.

That distinction is important.

Satlyt does not need to own the spacecraft for its business to work.

Its opportunity grows as more satellites are launched with increasingly capable processors.

Every new spacecraft could potentially become another place where Satlyt's software is deployed.

That creates a very different economics from building and launching an entire satellite network.

The $8 Million Will Fund the Software Layer

Satlyt says the new funding will accelerate development of its virtual AI data-centre vision while expanding its engineering and customer-delivery teams.

The company also plans to deploy its software across more third-party spacecraft.

That means the immediate challenge is not proving that AI can run in space.

It is proving that Satlyt can make the technology reliable, easy to deploy and commercially useful across many different spacecraft.

Satellite hardware is highly diverse.

Different operators use different processors, operating systems, communications systems and mission architectures.

Building a software layer that can work reliably across that fragmented environment is therefore a much bigger challenge than simply getting one AI model to run on one satellite.

Why Investors Are Betting on the Idea

Satlyt's investors are effectively betting that future satellites will need more computing power onboard.

As spacecraft become more capable, sending every piece of raw information back to Earth may become increasingly inefficient.

Processing data closer to where it is generated could allow operators to respond faster and reduce the amount of information that needs to travel through the communications link.

The opportunity also extends beyond satellite operators.

Software developers could eventually have a way to deploy applications into orbit without having to build their own spacecraft.

That creates the possibility of an ecosystem where spacecraft provide the hardware while companies like Satlyt provide the software infrastructure.

It is a familiar model from cloud computing  except the infrastructure is hundreds of kilometres above Earth.

Satlyt's African Connection Matters

Although Satlyt operates across the United States and Africa, its Nairobi base gives the company a notable connection to Africa's growing deep-tech ecosystem.

Its U.S. headquarters is in Sunnyvale, California, while its Africa headquarters is in Nairobi, Kenya. The company says it has a substantial engineering presence in Nairobi and a Kenyan subsidiary, while its broader team works across Africa and the United States.

That puts African engineers inside a highly specialised global market involving artificial intelligence, distributed computing and space infrastructure.

It also demonstrates a different path for African technology companies.

Instead of building products exclusively for local consumers, Satlyt is developing infrastructure for a global industry while maintaining significant engineering operations in Africa.

The Bigger Race Is for Computing in Orbit

Satlyt's fundraising comes as interest in space-based computing accelerates.

Companies including SpaceX and Google are exploring the possibility of putting more computing infrastructure in orbit.

Google's Project Suncatcher, for example, is exploring space-based computing infrastructure, while other startups are pursuing dedicated orbital data-centre hardware.

Satlyt's argument is that the industry does not necessarily need to wait for huge purpose-built data centres.

There are already thousands of satellites being launched or operating in orbit.

If their existing computing hardware can be used more effectively through software, some of that infrastructure could become a distributed computing network.

That is the bet behind Satlyt.

The Hard Part Is Connecting the Satellites

Satlyt's biggest ambition is not simply making one satellite smarter.

It wants computing resources across different spacecraft to eventually work together.

The company plans to demonstrate computing across multiple satellites as it develops its platform. If successful, that could move Satlyt from an onboard AI software provider toward something closer to a cloud infrastructure company for space.

But there are major technical challenges.

Satellites have limited power, limited computing resources and intermittent communications links. Different spacecraft also operate under different constraints.

A system that coordinates computing across them would have to deal with delays, connectivity interruptions, hardware differences and the harsh conditions of space.

Satlyt therefore has a long road between its current demonstrations and its vision of a shared orbital computing layer.

Satlyt's $8 million round is interesting because the company is not trying to win the space race by building another rocket or satellite.

It is betting that the next layer of value in space will be software.

Satellites are already collecting enormous amounts of data. The question is increasingly whether they should send all of that information back to Earth before doing anything with it.

Satlyt's answer is: let the satellite think first.

If the company can make its software reliable across different spacecraft, the opportunity could extend far beyond AI-powered satellite monitoring.

It could eventually provide an infrastructure layer where satellites become distributed computing nodes  turning existing spacecraft into something closer to a cloud network in orbit.

For an African-founded company with engineering operations in Nairobi, that is an ambitious position to take in one of the world's most technically demanding industries.

And with $8 million now behind the vision, Satlyt has more resources to find out whether space computing can become the next major software platform.