Lausanne, Switzerland – 29.04.2026 - "This image is a highly detailed, close-up view of Earth as seen from space." That sentence was generated not on the ground, but by a vision-language model running onboard a satellite in orbit - marking the first public glimpse of Clustergate-2, DPhi Space's newly commissioned compute server for space.
Clustergate-2 is a programmable compute platform offering a cloud-like experience onboard satellites. The payload launched at the end of March aboard Momentus' Vigoride 7 satellite, on SpaceX's Transporter-16 mission. During commissioning, the host satellite passed an image of Earth to the space server, which processed it with a vision-language model from Liquid AI and generated a caption directly in orbit.
On-Orbit LLM Inference
Prompt: Describe this image
Liquid AI LFM2-3b Response: "This image is a highly detailed, close-up view of Earth as seen from space, likely captured by a satellite or space telescope. The Earth is depicted as a large, circular sphere with a predominantly blue hue, indicating the vast oceans that cover most of its surface. The blue is interspersed with swirling white clouds, which are particularly prominent over the landmasses, suggesting the presence of weather systems and atmospheric activity. The overall composition of the image highlights the beauty and complexity of our planet, showcasing the dynamic interplay between the oceans, atmosphere, and landmasses."
Earth view from space, previously captured by Momentus processed by Clustergate-2
A Cloud-Like Experience in Orbit
Clustergate-2 brings a heterogenous processing cluster to orbit, includes an NVIDIA GPU provided by EDGX and supports containerized software applications - allowing developers to deploy software in space much as they would in the cloud. This lets them process data at the edge using large AI models, rather than downlinking raw data and running the same workloads on the ground.
That capability sits at the heart of the problem DPhi Space is solving: Space sensors are producing far more data than satellites can realistically send back to Earth, and the bottleneck is no longer collecting information but deciding what matters fast enough to act on it. By moving compute directly next to the sensor, DPhi Space makes it possible to reduce reliance on downlink and latency, while unlocking AI systems that can interpret scenes, prioritize events, and eventually trigger actions in near real time.
"Clustergate-2 is an early step toward a constellation equipped with compute infrastructure. Our goal is to remove the friction of deploying software and hardware in space. That is why our service is available through public pricing and documentation, so developers can start building for orbit immediately."
Aziz Belkhiria, CEO and Co-Founder of DPhi Space
Hackathon with Liquid AI
On its roadmap to bring AI applications to compute-enabled satellites, DPhi Space is also co-organizing a hackathon with Liquid AI, the device-native foundation model company, to explore what becomes possible when vision-language models run at the edge of space. Early submissions are already targeting real-world use cases like agents detecting maritime anomalies and identifying illegal mining activity. The hackathon runs until May 8 and offers cash prizes and free credits for winning applications deployed on Clustergate-2.
Together, the commissioning of Clustergate-2 and the launch of the hackathon signal a broader shift: orbital infrastructure is becoming programmable, accessible, and intelligent. DPhi Space is building the foundation for that shift, opening space to a broader set of users who can develop applications impacting the physical world.
About DPhi Space
DPhi Space is building AI-native satellites with GPU clusters onboard to run large AI models close to the sensors. Developers and operators can host hardware payloads and software applications and connect them to the space server to extract insights about the physical world in real time.
Learn more about DPhi Space: www.dphispace.com
Media contact: contact@dphispace.com