This month, Google announced Project Suncatcher, their plan to build scalable machine learning compute systems in space. The vision: constellations of solar-powered satellites equipped with Google's tensor processing units (TPUs), operating as a distributed data center in orbit.
Why space? Nearly constant sunlight for power and the ability to eliminate many earth-bound infrastructure constraints. This sounds like science fiction, but Google is serious. They're planning prototype satellites for early 2027 in partnership with Planet.
What Google's moonshot tells us extends beyond the technical challenges of orbital computing. It reveals something about AI infrastructure constraints and raises a question worth considering: What moonshot is your association taking?
What Project Suncatcher Actually Is
Project Suncatcher is a research initiative to move computation into orbit. Satellites equipped with TPUs would be powered by solar energy, creating a distributed data center that operates in space rather than on the ground.
The math behind this is compelling. The sun emits more power than 100 trillion times humanity's total electricity production. That's not a typo. The energy available in space dwarfs anything we can generate on Earth.
Google CEO Sundar Pichai described Project Suncatcher as part of their history of moonshots, joining quantum computing and autonomous driving. Like any moonshot, it requires solving complex engineering challenges. This is experimental work without guaranteed success.
The Engineering Challenges Are Massive
Google has already started testing. They ran their latest generation TPUs in a particle accelerator to simulate low-Earth orbit radiation levels. The results: TPUs survived without damage for up to five years.
But major challenges remain:
- Thermal management: How do you cool processors in a vacuum with no air convection?
- Radiation shielding: Five-year survival is promising, but longer timescales need validation
- System reliability: On-orbit repairs aren't like calling a technician
- Ground communication: Efficient data transfer between satellites and Earth-based systems
The partnership with Planet aims to launch two prototype satellites by early 2027. Sundar Pichai was clear about the experimental nature of this work. More testing and breakthroughs will be needed. The path forward isn't certain.
What This Tells Us About AI's Infrastructure Constraints
Project Suncatcher represents a response to one of AI's biggest bottlenecks: energy consumption and data center capacity. The scale of compute needed for frontier AI models is growing exponentially. Terrestrial data centers face real constraints around power availability, cooling requirements, and physical space.
Moving compute to space leverages an abundant, constant energy source while eliminating many infrastructure limits we deal with on Earth. The demand for AI compute has become so massive that companies are seriously considering solutions that would have seemed absurd five years ago.
This isn't just Google. Multiple companies are exploring unconventional infrastructure solutions. The fact that space-based computing has moved from theoretical to actively pursued tells you something about where we are with AI infrastructure demands.
The Broader Context: Stacking Exponential Improvements
We're seeing exponential growth in AI demand while simultaneously seeing major leaps in efficiency. Models like Claude Haiku 4.5 deliver frontier performance at a fraction of previous costs. Diffusion models promise 10x speed improvements. These efficiency gains matter enormously.
But we still need massive infrastructure investments. The improvements in efficiency we're seeing don't exist yet at the scale needed to meet demand. We need all of it: radical efficiency gains AND new infrastructure approaches.
Why both matter:
Orders of magnitude growth in demand requires orders of magnitude growth in supply. Space-based compute is one of many approaches being explored. SpaceX's Starlink already demonstrates viable low-Earth orbit satellite infrastructure. Next generation Starlink devices will include compute capability. Multiple companies pursuing this direction signals it's becoming viable, not just theoretical.
What Is a Moonshot (And What It Isn't)
A moonshot is a goal you don't know how to achieve. Classic timeframe: 10-20 years. If you know how to do it, if there's a clear path from here to there, then it's not a moonshot.
The goal must be something that most people would perceive as unattainable. You must be inspired by it while also fundamentally not knowing the path to get there. Google doesn't know how to make Project Suncatcher work. That's the entire point.
Some will fail. Google has had plenty of failed moonshots over the years. But the pursuit pays off over time if you're relentless about trying and smart about when to kill projects that aren't working.
In AI timescales, moonshots are compressing. What seemed like a 20-year moonshot might be achievable in 5-10 years. The acceleration of technological capability means the impossible becomes possible faster than our intuition suggests.
Why Associations Need Moonshots
Moonshots force you to think beyond incremental improvements. They inspire teams and stakeholders. They create permission for radical experimentation.
Acknowledging that the path isn't clear can be liberating. Being willing to say "we don't know how to do this yet" is powerful. It removes the pressure to have all the answers before you start.
What associations can learn from Google:
- Pursue ambitious goals without guaranteed paths
- Be willing to fail on some of them
- Get smart about when to pivot or kill projects
- Keep the long-term vision even when near-term obstacles appear
The constraints associations have faced are loosening rapidly. What seemed impossible last year becomes attainable this year. AI enables capabilities that were previously only available to massive corporations with enormous budgets.
Your moonshots won't literally be putting something in orbit (probably). But they can involve total transformation of your engagement model or doing things at a scale you currently think is unattainable.
The Permission to Fail
Google knows Project Suncatcher might fail. That's built into the concept of a moonshot. The engineering challenges might prove insurmountable. The economics might not work out. Unexpected obstacles might emerge that make the whole approach impractical.
Associations often don't give themselves permission to pursue goals that might fail. Fear of wasting resources, disappointing stakeholders, or looking foolish creates pressure to only pursue guaranteed wins.
But measured moonshots with clear evaluation criteria can be incredibly valuable. The learning from "failed" moonshots often leads to unexpected successes. You need to be smart about when to kill projects that aren't working, but you also need to be willing to start them in the first place.
Practical Steps Toward Moonshot Thinking
Start by asking: What would we do if we had Fortune 500 resources? Then ask: Which of those things are actually possible with AI?
Building your moonshot:
- Identify one goal that seems unattainable right now
- Make sure you're genuinely inspired by it
- Acknowledge openly that you don't know how to achieve it
- Set a 5-10 year timeframe
- Build in evaluation points to assess progress
- Create permission for experimentation without guaranteed success
- Communicate the moonshot nature of the goal to stakeholders
The last point matters. If you tell your board or membership that you're pursuing a moonshot, you're setting different expectations than if you present it as a standard strategic initiative. Moonshots come with built-in permission to not have all the answers.
Moving Forward
Google's Project Suncatcher represents the scale of ambition needed to meet AI infrastructure demands. Putting data centers in space sounds absurd until you understand the constraints they're trying to solve. Then it becomes one possible solution among many being explored.
Associations can adopt moonshot thinking at their scale. Your moonshot won't be orbital satellites, but it can be equally transformative for your community. The tools that enable moonshot achievements are becoming more accessible. AI levels the playing field between associations and Fortune 500 companies in ways that weren't possible before.
What will your association accomplish that seems impossible today? The time to set moonshot goals is now, while the constraints are still loosening and the capabilities are still expanding. Give yourself permission to pursue something you don't yet know how to achieve. That's where the interesting work happens.
.svg)
