Project Suncatcher is preparing for its first orbital test after years of research. The mission will launch a prototype satellite to evaluate how Google Tensor Processing Units, or TPUs, perform in space.

What Is Project Suncatcher

Project Suncatcher was announced last year. It is a long-term research effort exploring whether space could eventually host scalable machine learning infrastructure. In low Earth orbit, satellites can access near-constant sunlight. As a result, they can generate up to eight times more solar power than they would on Earth. Over time, multiple constellations of satellites could be linked together. This would allow them to manage larger AI workloads while in orbit.

According to the team, major breakthroughs often happen by working backward from an end goal. In this case, the goal is to ensure that AI’s benefits in areas like healthcare and scientific discovery can reach everyone, well into the future. Much like early research into autonomous driving and quantum computing, exploring compute in space requires years of experimentation before practical systems emerge.

A Basic Question to Answer

Turning that vision into reality starts with a fundamental question. Can AI hardware actually operate in space?

To find out, an initial mission will launch onboard the upcoming Transporter-18 rideshare mission with SpaceX. The mission was developed in partnership with Planet. It is designed to gather in-orbit data on how TPUs handle the physical stress of spaceflight, along with the radiation and thermal extremes of space.

As the team prepares for this early test launch, and works toward its next milestone in 2027, members have been discussing what they hope to learn. Their insights are featured in a new video series exploring the science behind the mission.

Hardware Survival in Orbit

A rocket trip into low Earth orbit lasts about 10 minutes. During that time, the spacecraft experiences intense vibration and sustained acceleration loads of up to 10 times the force of gravity, known as g force. Individual components, including TPU chips, can face even greater forces, ranging from 50 to 100 g.

To prepare for this, the team conducted vibration testing. The satellite was shaken intensely on all three axes to mimic the frequencies of a rocket launch. Since tests like this rarely go exactly as planned, the team was pleasantly surprised that the hardware held up under the force.

Once TPU chips reach space, radiation levels outside Earth’s atmosphere present another challenge. Solar events and cosmic rays can damage electronics. Therefore, the team tested TPUs in a proton beam facility at UC Davis’s Crocker Nuclear Laboratory while running AI workloads. Throughout the test, researchers monitored how errors, such as bitflips, would affect performance.

Initial results show that Trillium TPUs hold up remarkably well. In fact, they can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission.

However, some factors can only be tested in space itself. Putting the first TPUs into orbit next week will help gather data to inform future launches.

Cooling Challenges in Orbit

Cooling orbital data centers remains a crucial research challenge. TPUs generate a large amount of heat in a small area. Without proper cooling, the chips risk overheating.

Unlike on Earth, there is no airflow in space. In a vacuum, heat can only be diffused through radiators. This requires an entirely different approach to cooling electronics.

To address this, the team is testing several approaches, including a combination of heat pipes and radiators. So far, the technology has been tested in a thermal vacuum chamber that simulates both the thermal and vacuum conditions of space. The upcoming mission will reveal how the new TPU cooling system performs in actual orbit, allowing the team to refine its designs further.

Satellite Interconnectivity

Future satellite designs will each carry dozens of TPU chips while orbiting Earth in clusters. To maintain the bandwidth necessary for processing AI, every satellite must know both its own position and its position relative to its neighbors. To achieve this, the satellites will communicate using lasers.

While this technology already exists, most state-of-the-art systems are optimized for low bandwidth across long distances. In contrast, these lasers need to operate at very high bandwidth over extremely short distances. Maintaining that connection requires extraordinary precision, comparable to hitting a coin-sized target from miles away while both points are in motion. This element of Project Suncatcher will be tested in 2027, when two satellites are placed in orbit together.

Just the Beginning

Exploring space as a viable location for scalable AI compute will not happen overnight. Instead, it requires methodical engineering, beginning with proving that hardware can handle the physical and unpredictable realities of operating in orbit.

This first launch is focused on identifying what works and pinpointing potential points of failure. Those findings will then be applied to future missions.

Every transformative technology begins with an ambitious goal, paired with the discipline to work through difficult problems along the way. As Project Suncatcher heads toward the launchpad, the team plans to continue sharing what it learns, along with the science driving the work forward.

Related post

View all