Google is sending its custom AI chips to space for the first time. Project Suncatcher will launch a prototype satellite carrying Google Tensor Processing Units (TPUs) on SpaceX's Transporter-18 rideshare mission, developed in partnership with Planet. The test aims to determine whether the hardware can survive the extreme physical conditions of orbit - a foundational step toward building orbital data centers that could one day tap near-constant solar power.
The satellite is designed to gather in-orbit data on how TPUs handle vibration during launch, radiation exposure, and the vacuum of space. "Just as early research into autonomous driving and quantum computing required years of experimentation before we got to practical systems, exploring compute in space begins with measured, deliberate steps," the Project Suncatcher team said.
Surviving the rocket ride
A trip to low Earth orbit takes about 10 minutes. During that window, the spacecraft endures sustained acceleration up to 10 times the force of gravity. Individual TPU chips can experience forces reaching 50 to 100 g. The engineering team tested this on the ground by shaking the satellite on all three axes to mimic launch frequencies. The hardware held up.
Once in orbit, radiation becomes the primary threat. Solar events and cosmic rays can corrupt electronics through effects like bitflips. Researchers tested Trillium TPUs at UC Davis's Crocker Nuclear Laboratory, running AI workloads while bombarding the chips with a proton beam. Initial results showed the processors can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission.
Cooling without airflow
TPUs pack substantial heat into a small area. On Earth, fans and airflow handle that load. In a vacuum, heat can only be dissipated through radiators. The team is testing a combination of heat pipes and radiator designs inside a thermal vacuum chamber that replicates the space environment. The upcoming mission will reveal how the cooling system performs in actual orbit, generating data to refine future designs.
Connecting satellites with lasers
Long-term plans call for satellite clusters, each carrying dozens of TPU chips. To process AI workloads across a constellation, every satellite must know its position relative to its neighbors with extreme precision. The satellites will communicate via laser links operating at high bandwidth over short distances - a configuration that demands accuracy comparable to hitting a coin-sized target from miles away while both points move. Google plans to test this inter-satellite connectivity in 2027 with a two-satellite mission.
Why this matters for science and research professionals
Project Suncatcher is a multi-year research effort, not a near-term product. For researchers and technical professionals, the mission signals a serious investment in understanding whether space-based compute infrastructure can move beyond theory. The data collected from this launch - on hardware degradation, thermal performance, and radiation tolerance - will shape engineering decisions for years. Organizations working on remote sensing, climate modeling, or disaster response may eventually benefit from orbital AI processing that reduces data transmission bottlenecks. For now, the work centers on a single question: can the hardware function reliably outside Earth's atmosphere. The answer starts arriving next week.
Your membership also unlocks:
