From Rejection to Rocket: Lone Developer's Software Flies on SpaceX!

Rama Afullo, a Kenyan American engineer with prior experience at Google's cloud computing division and SpaceX's Starlink, identified the nascent potential for computing in space, specifically for orbital data centers. After facing internal rejections for his vision at both companies, Afullo departed SpaceX to co-found Satlyt. Headquartered in Sunnyvale, California, and Nairobi, Satlyt aims to develop software that enables multiple satellites to collectively handle large computing tasks in orbit.

Satlyt has successfully raised an $8 million seed round. The timing appears opportune, as SpaceX has since committed to orbital data centers, and Satlyt's software is slated for an upcoming launch on a SpaceX rocket, coinciding with the debut prototype for Google's own space data center initiative, Project Suncatcher. Unlike competitors such as SpaceX, Google, or startups like Starcloud and Cowboy Space Company, Satlyt distinguishes itself by not constructing its own spacecraft. Instead, Afullo's team concentrates on building software to operate AI models on satellites, likening their approach to VMware and Snowflake, which provide platforms for complex software execution without requiring users to manage the underlying infrastructure. Afullo describes Satlyt's offering as an "Android as a horizontally integrated, open ecosystem" for the orbital data center "iPhone" developers.

Satlyt has already proven its software through two demonstration missions. Its primary objective is to enhance operational efficiencies for spacecraft, which traditionally rely on ground-based flight controllers for problem-solving. The process of transmitting data back to Earth, known as downlink, is both costly and time-consuming. By leveraging AI to make autonomous decisions onboard the spacecraft, significant savings can be realized, whether through anomaly resolution or by processing sensor data directly in space rather than on the ground.

Earlier this year, Satlyt successfully deployed Google DeepMind's Gemma AI model on a spacecraft operated by Momentus. This implementation resulted in a more than 60% reduction in the data transmission size for onboard software errors, a level of efficiency that Afullo estimates can save hundreds of thousands of dollars per satellite annually. A forthcoming launch will integrate Satlyt's software onto a spacecraft manufactured by TakeMe2Space, an Indian startup specializing in satellite computing hardware. This mission will cater to three customers: NASA, which is funding tests for cloud computing protocols in space; Stellerian, a space surveillance startup, which will test image processing workloads; and TakeMe2Space itself, demonstrating the satellite’s capacity to host third-party software.

Currently, high-powered GPUs are not widely prevalent in orbit, making the types of workloads Satlyt is focusing on the most viable immediate opportunities. The company's subsequent endeavor involves creating a shared computing system, or a cloud, spanning two distinct satellites, a project anticipated for the following year. Successful execution of this initiative would enable Satlyt to establish a genuine compute cloud in orbit, functioning as a third-party provider to spacecraft builders, effectively transforming satellites into "revenue-generating managed services."

The $8 million seed round was led by Houston-based Non Sibi Ventures. Partner Bernard Harris, a former NASA astronaut with over two decades of service, played a crucial role in validating the firm's confidence in Satlyt's novel, space-based business model. A key factor in their investment was that Satlyt’s success does not hinge on the most ambitious vision of space data centers, which are subject to economic uncertainties and launch bottlenecks. Non Sibi partner Kent Lucas emphasized that Satlyt's growth can be driven simply by the increasing number of satellites. Afullo envisions Satlyt's software being operational on 20% of all satellites by the end of the decade, anticipating that virtually every spacecraft builder will be incorporating GPUs and similar advanced processors into their satellites.