Google’s chief of data‑center energy has warned that moving AI compute to orbit may be far more expensive than building on the ground. Speaking at The Information’s WTF Summit, Amanda Peterson Corio said space‑based data centers could cost up to a hundred times as much as terrestrial facilities, a stark reminder that the sky‑high vision of orbital compute still faces massive economic hurdles.
Key takeaways
- Google estimates space data centers could be 100× pricier than Earth‑based ones.
- The cost concern is about scaling; a single 1 kW satellite prototype is already operational.
- Google continues to invest heavily in terrestrial data‑center projects, including a $17 billion Wyoming campus.
- Industry experts view permitting and power logistics as bigger bottlenecks than raw energy costs.
Why the cost difference matters
Building a data center in space involves launch expenses, custom‑engineered hardware, and the need to operate in a harsh environment with limited maintenance opportunities. Corio’s 100× estimate reflects not just the price of a launch but also the ongoing challenges of powering and cooling hardware millions of miles from Earth. If the cost cannot be justified, the business case for orbital AI compute collapses.
Google’s current space‑compute experiments
Google recently launched the first prototype of its Project Suncatcher on SpaceX’s Transporter‑18 rideshare mission (October 1). The satellite, roughly 1 kW in power, has already run Gemma 4, a TPU‑based AI model, that generated a short haiku in orbit. While this demonstrates that AI chips can function in space, the experiment is limited to a single chip and does not address the massive scale needed for commercial AI services.
Terrestrial hurdles versus orbital ambitions
Corio emphasized that the real bottleneck on Earth is permitting, not power. Securing land, grid connections, and regulatory approval can delay data‑center builds for years. Space promises to sidestep those hurdles, but the cost premium may outweigh the benefits. Her advice was simple: "Let's explore all the terrestrial options before we go extraterrestrial."
Google’s parallel investment in ground‑based capacity
While debating orbital compute, Google is expanding its Earth‑based footprint. The company purchased an unfinished Wyoming data‑center project from Crusade, a Denver‑based developer, and announced a $17 billion investment to develop the campus. The move underscores Google's confidence in scaling traditional facilities despite permitting challenges.
Comparing costs: Space vs. Earth (illustrative)
| Aspect | Space‑based | Terrestrial |
|---|---|---|
| Launch & hardware | Very high (one‑off and custom) | Standard procurement |
| Power & cooling | Limited, requires radiation‑hard solutions | Grid‑connected, scalable cooling |
| Maintenance | Remote, costly servicing | On‑site, routine |
| Overall cost factor | ~100× per unit of compute | Baseline |
What this means for you
If you are a developer or enterprise evaluating AI compute options, the takeaway is clear: orbital data centers are not yet a cost‑effective alternative. Focus on providers that are expanding terrestrial capacity, such as Google’s newly acquired Wyoming campus, which promises massive power availability at a more predictable price. Keep an eye on permitting reforms and renewable‑energy initiatives that could lower ground‑based costs further.
What happens next?
Google is likely to continue its modest orbital tests while pouring billions into land‑based infrastructure. Corio’s comments suggest the company will prioritize solving permitting and power‑supply challenges on Earth before committing to large‑scale space deployments. Industry observers will watch for any updates from The Information’s summit and future announcements from Google’s Project Suncatcher program.
Conclusion
Space‑based AI compute remains an intriguing research avenue, but the economic reality is stark: building a data center in orbit could cost up to a hundred times more than a comparable Earth facility. For now, the practical path for most businesses is to leverage the rapid expansion of terrestrial data centers, where costs are falling and capacity is scaling to meet growing AI demand.
Frequently asked questions
How much more expensive could a space data center be?
Google’s data‑center energy head estimates up to 100 times the cost of an Earth‑based facility.
What is Project Suncatcher?
It is Google’s prototype satellite that runs AI chips in orbit, currently delivering about 1 kW of power.
Is Google abandoning space data centers?
No, Google continues testing orbital hardware but says scaling would be prohibitively costly at present.
What terrestrial projects is Google pursuing?
Google is investing $17 billion in a Wyoming data‑center campus after acquiring an unfinished project from Crusoe.
Should I consider orbital compute for my AI workloads?
Given the current cost estimates, focusing on terrestrial providers is the more economical choice.
