October 1, 2026 marked a major milestone for Google’s Orbital Data Center (ODC) aspirations. On a quiet stretch of California coastline just outside Santa Barbara, the tech giant launched its Project Suncatcher prototype into Low Earth Orbit (LEO) from Vandenberg Space Force Base. While Google is only testing four in-house Tensor Processing Units (TPUs) on a satellite payload, important lessons will be learned to help scale Artificial Intelligence (AI) compute in space.
ABI Research forecasts effective orbital compute power to reach 1.5 Gigawatts (GW) by 2035. Google does not want to miss out on this growing demand for non-terrestrial cloud solutions, which is largely induced by grid constraints.
Being one of the most influential tech companies globally, it’s no surprise that Google’s moonshot project is being covered by numerous news outlets and industry publications. In this Q&A, I’d like to share my own perspective as a Space Tech Analyst. What does Project Suncatcher mean for the ODC market? How does Google compare with competitors? What should you watch for over the next few years?
1. What could Google learn from this first Project Suncatcher mission that would cause it to significantly change its proposed architecture rather than simply scale it up?
The original paper Google published on orbital AI infrastructure has several assumptions this mission is stress-testing that could force redesign rather than scaling. The first is formation flying, as Google proposes keeping the satellites within ~1 Kilometer (km) of each other for efficient interconnect.
If this premise collapses, it will need to design for looser constellations with longer-range Inter-Satellite Links (ISLs), shifting the architecture from a single distributed supercomputer to many independent compute nodes—a fundamentally different workload mapping.
Radiation tolerance and thermal management performance are also critical factors, as these will indicate how Google’s TPUs perform in real space environments. Poor results would mean reworking the assumptions on radiation shielding requirements and the per-satellite power density required, which would cascade into constellation size and formation dynamics.
2. Does Google bring any unique advantages to Orbital Data Centers (ODCs) that traditional space companies do not? And where will it still need partners from the space industry?
The TPUs themselves. Google designs and mass-produces these TPUs, and no traditional space company builds cutting-edge Machine Learning (ML) accelerators at data center-scale. Google also has one of the largest fleets of ML workloads on the planet to test these TPUs on.
All of this is backed by significant capital, talent, and patience. However, Google is not a space company (yet), and it still needs partners in the space industry for launch, satellite bus and payload integration, space-qualified optical terminals, space-adapted thermal management, and operating assets in orbit. Case in point, Project Suncatcher used the SpaceX Falcon 9 rocket to launch the satellite into space.
In short, Google is bringing the brain (chips, workloads, software, capital) but needs the space industry for the body (launch, bus, optics, thermal, operations).
3. What would have to happen for the economics of Google space data centers to become competitive with terrestrial data centers?
This remains strongly debated, but cost inputs for space systems need to come down across the board, particularly for launch. For terrestrial data centers, electricity is only a small slice of overall costs, so relocating to orbit exchanges this relatively small cost input with more expensive launch, radiation, insurance, redundancy, and manufacturing inputs.
The economics of ODCs won’t close without other inputs falling alongside the launch costs. This means Starship will need to reach a significant number of launches before amortization helps drive down the economics of launch. In the paper, Google points to launch costs of around US$200/Kilograms (kg) by the mid-2030s, broadly in line with where ABI Research estimates launch economics will make ODCs viable.
4. If Project Suncatcher succeeds, which use cases will Google likely target?
With Google testing TPUs for AI workloads, the near-term opportunity will likely be space-native use cases: processing Earth Observation (EO) data on orbit before downlinking, fusing data from on-orbit sensors, and serving compute to other assets on orbit. Longer term, Google may pursue terrestrial use cases such as serving compute to autonomous machines, infrastructure, and enterprise mobility applications, and ultimately offloading large-scale training and AI inference, the stated ambition behind Suncatcher. Whether Google will be an ODC operator or remain a merchant TPU vendor is still an open question. I think it will do both.
5. What should we watch over the next 2 to 3 years to determine whether Project Suncatcher is becoming the foundation of a real Google space data center business rather than remaining an experiment?
Over the next 2 to 3 years, I’d watch three signals.
First, whether Google publishes prototype results that validate its TPU performance, radiation tolerance, and optical-link throughput in orbit.
Second, a follow-on mission scaling beyond the prototypes toward a true cluster, as described in the paper.
Third, a commercial deal to deploy TPUs at scale, whether through Google’s own network or as a supplier to other ODC operators. Many ODC operators already see Google TPUs as a potential workhorse for their constellations, so a deal beyond the current Planet prototype partnership would signal that Suncatcher is becoming a business, not just Research and Development (R&D).
Let’s Continue the Conversation
Google and its partner Planet’s 1-year test will be pivotal to building AI compute hardware that is resilient to cosmic radiation, can handle intense heat, and is distributed optimally. Whether Google needs to completely rehaul its ODC roadmap remains to be seen. ABI Research’s Space Technologies & Innovation team will be watching closely over the next 12 months. If your organization needs timely analyst insight into the Orbital Data Center (ODC) landscape, subscribe today.
If you’d like to discuss this topic more in-depth, shoot me a message on LinkedIn.
Dimitris Mavrakis