high machines
Impossible. But doable.
Orbital data centres for AI inference. European sovereign. Architecture-lock phase.
Closer to the data
Why orbital compute wins: round-trip latency scales with altitude. LEO is a hundred times faster than GEO. The Moon is 2,500× slower. The Sun–Earth Lagrange points are minutes away.
The Three Whys
The full story lives on /story. Here's the 30-second version.
Why this
The orbital compute thesis.
AI inference is bottlenecked by terrestrial power, cooling, and grid constraints. Orbital data centres eliminate all three: continuous solar, radiative cooling into deep space, no permitting on the ground.
Why now
The 18-month architecture window.
Launch costs crossed the LEO economic threshold in 2024. AI inference demand scaled 10× in 2025. Whoever locks the architecture by mid-2027 sets the standards the primes have to comply with. We are racing for the standard, not the compute market.
Why us
6 AI4CE papers + a PhD in mission-architecture design.
JP (the founder) has spent 4 years building the AI4CE methodology — agent-based deep-RL for CubeSat system generation. 6 peer-reviewed papers, 1 ESA CDF collaboration, the only open-stack agent framework for concurrent engineering in space.
Read the full Three Whys on /story →
Who are you?
The site has a different pitch for each audience. Pick the one that fits.
Partners
Satellite builders, ground-station operators, integrators. Reciprocal, network-effects.