Orbital Secures a16z Funding to Launch AI Data Centers in Space
Andreessen Horowitzs a16z Speedrun fund has backed Orbital Inc., a startup pursuing one of the most ambitious infrastructure plays in the AI era: building data centers in low-Earth orbit. The funding amount was not disclosed, but it will finance the companys first test mission, Orbital-1, targeting launch in April 2027.
The startup is tackling what has become the AI industrys most pressing constraint: energy. Terrestrial data centers housing thousands of GPUs are consuming electricity faster than power grids can supply it, prompting major tech firms including Google, Microsoft, and Amazon to recently meet with President Trump at the White House to address the crisis. Orbitals solution is radical: move compute infrastructure 1,200 miles up, where solar power is continuous and cooling is free.
The Orbital Model
Orbital plans to deploy a constellation of satellites in sun-synchronous orbits, allowing them to harvest solar energy 24/7 without interruption from Earths day-night cycle. Each satellite will house clusters of Nvidia-powered servers, functioning as decentralized processing nodes for AI inference workloads. Heat generated by the chips dissipates directly into the vacuum of space via radiative cooling, eliminating the massive energy costs of terrestrial data center cooling systems.
In orbit, solar power is continuous and cooling is fundamentally different, said Euwyn Poon, Orbitals co-founder and CEO. Orbital is building compute infrastructure that removes the energy ceiling and scales with AIs potential.
Latency and Technical Challenges
Critics have raised concerns about latency from space-based compute. Orbital counters that at 500-600 kilometers altitude, round-trip latency is 20-40 millisecondscomparable to a fiber connection between Los Angeles and Denver. Starlink already demonstrates this commercially with sub-50-millisecond latency to homes and businesses, Poon noted.
The company is targeting inference workloads specifically, which can be distributed across clusters and are not affected by the stateless nature of individual requests. This makes them ideal for orbital deployment, where radiation-induced bit flips could corrupt longer-running training jobs.
Radiation hardening remains the primary technical challenge. The Orbital-1 test mission will validate the companys protection systems in real orbital conditions. Rather than attempting repairs, each satellite is designed for replacement at end-of-life, with controlled deorbit and complete burn-up on reentrya approach Poon argues is more environmentally friendly than terrestrial e-waste streams.
Market Implications
If successful, Orbital could decouple AI progress from terrestrial power constraints, enabling a future where cloud computing operates independently of ground-based energy grids. The startup is opening a dedicated RD facility, Factory-1, in Los Angeles to begin manufacturing its first specialized compute satellites.
Elon Musk has recently begun discussing orbital data centers publicly, but Orbital has been developing the concept for some time and aims to launch its first satellite 12 months from now on a SpaceX Falcon 9 rocket.
The harder the problem, the better, said a16z General Partner Andrew Chen. Orbital is taking on AIs biggest constraint with a bold and radical idea.
The success of Orbital-1 could validate a new category of infrastructure investment, positioning space-based compute as a critical enabler for the next generation of AI applications.
Sources: SiliconANGLE, a16z Speedrun
