Rama Afuro says he sees orbital data centers coming.
The Kenyan-American engineer worked in Google’s cloud computing business before taking a short stint in SpaceX’s Starlink communications network in 2024. His experience convinced him that there was untapped potential for computers in the universe.
“When I was at SpaceX, I tried to sell this internally and they said no. When I was at Google, I tried to sell this internally and they said no,” he told TechCrunch.
So he left SpaceX and co-founded a new company, Satlyt, headquartered in Sunnyvale, Calif., and Nairobi, to develop satellite software that allows many satellites to share the work on large-scale computing jobs in space. The company has raised $8 million in a seed round, Afullo exclusively tells TechCrunch.
The timing seems good for him. Since then, SpaceX has gone all-in on orbital data centers, and Satlyt’s software is scheduled to launch on a SpaceX rocket on Thursday with Project Suncatcher, the first prototype of Google’s space data center business.
Unlike SpaceX, Google, or startups like Starcloud and Cowboy Space Company, Satlyt has no plans to build its own spacecraft. Afullo’s team is building software to operate AI models on satellites, and he compares his approach to VMware and Snowflake, companies that offer platforms that allow users to run complex software without worrying about the underlying infrastructure.
“People doing orbital data centers like SpaceX, if they were the iPhone, we would build Android as a horizontally integrated open ecosystem,” he said. In other words, Satlyt wants to provide software that works on many companies’ satellites.
Satlyt has already flown the software in two demonstration missions. The initial focus is on finding operational efficiency for the spacecraft, and typically relies on ground-based flight controls to resolve any problems that arise. Sending data back to Earth, known as a downlink, is expensive and often time-consuming, so using AI to make decisions onboard a spacecraft can mean real savings when resolving anomalies or processing sensor readings on a spacecraft rather than on the ground.
Earlier this year, Satlyt deployed Google DeepMind’s Gemma AI model to a spacecraft operated by Momentus. This model reduces transmission sizes for onboard software errors by more than 60%, which Affurulo said can save hundreds of thousands of dollars per satellite each year.
The launch will put Satlyt’s software running on a spacecraft built by TakeMe2Space, an Indian startup that develops computing hardware for satellites. This mission involves three customers. NASA is paying a company to test cloud computing protocols in space. Steelerian is a startup focused on space surveillance (tracking objects in orbit) that wants to test image processing workloads. And TakeMe2Space itself has shown that its satellites can host software from other companies.
There are very few high-performance GPUs in orbit at this time. For now, the most likely opportunities lie in the kinds of workloads Satlyt is working on. But the company’s next project is to show it can build a shared computing system (cloud) that spans two different satellites, and it plans to try that next year. If successful, the company will be able to act as a third-party provider to spacecraft manufacturers to build a true computing cloud in orbit.
The way we think about it is that we turn your satellite into a revenue-generating managed service,” Affurulo said.
Satlyt’s seed round was led by Houston-based Non Sibi Ventures, whose partner Bernard Harris, a former NASA astronaut for more than 20 years, helped ground the company’s belief in novel space-based businesses.
One of the reasons the company invested is that Satlyt doesn’t rely on the most high-octane vision of a space data center, which faces uncertain economics and launch bottlenecks. “Rama doesn’t need a data center in space for it to be a big success, right?” non-Sibi partner Kent Lucas told TechCrunch. “It could simply be driven by the number of satellites.”
Afflullo’s hope is to have 20% of satellites operating by the end of the decade, by which time he expects virtually all spacecraft manufacturers will have equipped their satellites with GPUs or similarly advanced processors.
“If you’re putting up a satellite without a GPU on board, you’re at least doing yourself a disservice, right?” he says.
This article was originally published on October 1st. Satlyt is one of dozens of other startups vetted by TechCrunch that will appear at Disrupt this week as part of the broader Battlefield competition in downtown San Francisco. It all starts on Tuesday. Click here for more information.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
