Google is done waiting for Earth to catch up. In a partnership that reads like science fiction, Google and SpaceX are teaming up on a project called "Project Suncatcher" — a plan to launch TPU-equipped, solar-powered satellites into orbit that will beam AI compute power down to any point on the planet. If it works, it could make terrestrial data centres obsolete for certain AI workloads. And it changes everything about how we think about the cloud.
What Is Google Project Suncatcher?
Project Suncatcher is Google's plan to place AI-capable satellites — powered by solar energy and equipped with Google's custom Tensor Processing Units (TPUs) — into low Earth orbit. These satellites would process AI inference workloads directly in space and transmit results back to users globally, bypassing the need for traditional land-based data centres entirely.
The partnership with SpaceX provides the launch infrastructure. SpaceX's Starship and Falcon 9 rockets would carry the satellites into orbit, while SpaceX's existing Starlink constellation could potentially serve as a communication backbone. Google brings the AI hardware and software stack; SpaceX brings the rockets and the orbital expertise.
Why Would You Put AI in Space?
The answer is twofold: power and latency. Traditional data centres face two massive constraints in 2026. First, they consume enormous amounts of electricity — AI training and inference are already straining power grids globally. A 2026 Gallup survey found 71% of U.S. adults oppose having a data centre in their local area, largely due to power and water usage concerns. Second, terrestrial data centres are constrained by geography — some regions of the world simply lack the fibre infrastructure for low-latency AI access.
Solar-powered orbital compute solves both. Satellites in low Earth orbit receive uninterrupted sunlight (no night cycle at the right altitude), meaning near-infinite clean energy. And a constellation of orbital compute nodes could deliver AI inference to any point on Earth with millisecond latency, regardless of ground infrastructure.
What Are the Challenges?
The engineering hurdles are immense. TPUs generate significant heat — a solved problem on the ground with liquid cooling, but a serious challenge in vacuum. Radiation in low Earth orbit degrades semiconductor performance over time. And the cost per FLOP in orbit is currently far higher than on the ground. Google and SpaceX will need to solve heat dissipation, radiation hardening, and launch economics before Project Suncatcher becomes commercially viable.
There are also regulatory dimensions. Orbital AI infrastructure raises questions about jurisdiction, data sovereignty, and international spectrum rights that no existing legal framework fully addresses.
What Does This Mean for the Future of AI?
If Project Suncatcher succeeds, it democratises AI access in a way nothing else can. A farmer in rural Sub-Saharan Africa, a doctor in a remote Pacific island, a startup in a country with unreliable internet — all would have access to the same low-latency, high-power AI compute as a Fortune 500 company in New York. That is genuinely transformative. Google knows this. The company that "organises the world's information" now wants to be the company that powers the world's intelligence — from space.
Bottom line: Project Suncatcher is ambitious, technically challenging, and potentially the most consequential infrastructure project in AI history. Watch this space — literally.
