Google has secured its first dedicated heavy-lift launch for Project Suncatcher, its long-term effort to move AI infrastructure into orbit and power it with near-constant sunlight. Under a launch services agreement announced on November 19, 2026, Blue Origin’s New Glenn rocket will carry a set of prototype data center satellites to space in the first quarter of 2029.
- Launch vehicle: Blue Origin’s heavy-lift New Glenn, targeted for Q1 2029.
- Payload: Satellites equipped with Google’s Tensor Processing Units (TPUs), each with roughly the computing capacity of a data center rack.
- Orbit: Sun-synchronous orbit, where the satellites can rely on solar panels for power almost continuously.
- Context: The deal follows an October prototype mission that ran Google’s Gemma 4 model on a TPU in orbit and transmitted an AI-generated haiku back to Earth.
What Google plans to fly — and why orbit makes sense
Project Suncatcher is Google’s research moonshot exploring whether a network of solar-powered satellites carrying its custom AI chips could scale machine learning beyond the limits of terrestrial power grids. According to Travis Beals, Google’s senior director for the project, the dedicated New Glenn mission represents a crucial milestone on the company’s roadmap.
“Project Suncatcher is a long-term moonshot to scale AI infrastructure in space by tapping into the abundant solar power in orbit. This dedicated launch with Blue Origin represents a crucial milestone on our roadmap as we expand the boundaries of what’s possible for the future of compute.” — Travis Beals, Google
Google argues that solar panels in orbit can generate up to eight times more energy than on Earth, offering relief for electricity networks increasingly strained by AI data centers. The satellites will fly in a sun-synchronous orbit, where they receive sunlight for nearly the entire day and draw power from onboard solar arrays. Google has not disclosed the number of satellites, nor exact figures for compute, power or mass.
The October prototype, launched as part of a SpaceX rideshare mission that also carried a Planet satellite fitted with a Google TPU, successfully booted its chips and used the onboard Gemma 4 model to transmit an AI-generated haiku to Earth. That experiment operated in low-Earth orbit and could only run for about 15 minutes at a time, underscoring how far the project has to go before orbital data centers become practical.
Engineering hurdles remain steep
Cooling is the most immediate challenge. Standard air-based heat dissipation does not work in the vacuum of space, so heat must be released through pipes and radiator panels rather than conventional cooling systems. Scaling those thermal systems to handle rack-level compute loads is an open engineering problem.
Radiation poses another risk. High-energy particles can cause bit flips, corrupting binary data in ways that would be unacceptable for production AI workloads. Google’s satellites will also need to maintain a tight formation so they can communicate via laser links, a requirement that adds precision-station-keeping demands to the mission.
The launch vehicle itself carries uncertainty. New Glenn has been grounded since an explosion during a test in May, and its return to flight has slipped to January. Google plans to link two satellites in 2027 as a precursor to the larger Blue Origin launch.
A competitive race with real costs
Google is not alone in pursuing orbital compute. Starcloud raised $170 million (around €155 million) in March to build space-based data centers, while SpaceX — Blue Origin’s rival — warned in its IPO documentation that its own orbital data center ambitions might never work. The economics remain daunting: a single Falcon 9 launch costs roughly $74 million, and payload limits barely cover the hardware for a small server room, let alone a hyperscale facility.
Brandon Lucia, a professor of electrical and computer engineering at Carnegie Mellon University, told The New York Times that expanding from one satellite to a vast network operating like a giant data center will take years and enormous funds, with additional engineering problems at large scale. Scientists have also long warned about Kessler syndrome, a theoretical cascade of orbital debris collisions; the number of tracked objects in low-Earth orbit has risen from just under 2,000 to more than 47,000.
Project Suncatcher is both a striking technical bet and an admission of how energy-hungry AI has become. The physics of sun-synchronous orbit are attractive, but cooling without air, flying in millimetric laser formation and depending on a rocket that has spent months on the ground after an explosion are not trivial obstacles. If the 2029 mission succeeds, it will mark the moment AI infrastructure began to outgrow the planet — a milestone that says as much about terrestrial energy constraints as it does about orbital ambition.
Sources & References
- We’re partnering with Blue Origin on our first heavy-lift Project Suncatcher launch.
- Blue Origin will launch Google’s AI server satellites in 2029
- Google signs deal with Blue Origin to send satellite data centers into orbit – Engadget
- Google Orders Heavy-Lift Launch for Project Suncatcher from Blue Origin