Google and Orbital Compute Develop AI Data Centers in Space

Google’s Project Suncatcher

Google’s Project Suncatcher, unveiled by CEO Sundar Pichai in a November 2025 tweet, marks a bold step toward placing AI processors in space. The initiative, developed over 18 months, aims to harness solar energy by situating specialized machine learning chips near the Sun, addressing the growing energy crisis in AI. Pichai wrote that their TPUs are headed to space and that Project Suncatcher, inspired by their history of moonshots ranging from quantum computing to autonomous driving, is exploring how they could one day build scalable AI compute systems in space. The first satellites, set for 2027, will test TPU resilience in space, with Planet, a U.S. microsatellite company, collaborating on the launch. However, technical hurdles like thermal management and radiation resistance remain, though Google claims initial tests in particle accelerators have been successful, as noted in Elpais.

Orbital Compute, a five-month-old startup, has filed a proposal with the Federal Communications Commission (FCC) to deploy up to satellites for AI processing, aiming to challenge SpaceX and Blue Origin. Founded by Euwyn Poon, a former Spin CEO, the project seeks to leverage solar power and space’s cold environment to cool servers without water. Each satellite, equipped with a 100-kilowatt solar panel, would operate in sun-synchronous orbits, using radiative cooling to manage heat.

NVIDIA Vera Rubin platform

NVIDIA is reengineering data centers with its Vera Rubin platform, which uses liquid cooling to tackle AI’s energy demands. Unlike traditional air-cooled systems, Rubin circulates a 75% water, 25% propylene glycol mixture directly over GPUs and processors, reducing energy use significantly. The key is to transfer heat more efficiently, NVIDIA explains, noting that Rubin could add 10% more racks to existing facilities.

Thales Alenia Space, part of a European consortium, has validated the viability of orbital data centers through its ASCEND project. The project, aiming for 1 gigawatt of capacity by 2050, envisions modular, robotically assembled infrastructure. Christophe Valorge, Thales Alenia’s technical director, called it a “paradigm shift” for digital sovereignty and green energy. The EROSS IOD demo, scheduled for 2026, will test key technologies, with a return on investment projected at hundreds of billions by 2050, according to techno-science.net.

The race to space-based computing intensifies as companies like Nebius, backed by Nvidia and Meta, expand into Europe. Nebius, originally a Yandex spinoff, now operates in Spain via Merlin Properties, leasing 18 megawatts in Getafe. Orbital Compute’s plan to launch a large number of satellites also raises concerns about space debris, with critics warning of environmental risks.

Thales Alenia Space ASCEND project

Thales Alenia’s ASCEND project, part of a European consortium including Carbone 4, VITO, Orange Business, CloudFerro, Hewlett Packard Enterprise, ArianeGroup, DLR, Airbus Defence and Space, and Thales Alenia Space, validated the environmental and technological feasibility of orbital data centers. The study, which compared impacts of space and terrestrial data centers, found that a 10x less emission launcher would be required to minimize CO2. The project also highlighted the absence of water usage for cooling, a critical advantage in drought-prone regions. The development of such a launcher was validated through ArianeGroup’s contributions and the ESA’s PROTEIN study, as outlined in Thales Alenia Space/Master Image Programmes.

Nebius and Merlin Properties

Nebius, which expanded to Spain after operations in Iceland, Finland, the UK, and France, leases 18 megawatts from Merlin-Edged in Getafe. Merlin Properties, also a partner with CoreWeave, positions Spain as the fifth European market for Nebius, which was originally part of Yandex. Arkady Volozh, a Yandex co-founder who fled Russia after the Ukraine war, repositioned the company under its current name in Amsterdam, as noted in EL MUNDO.

The interplay between space-based and terrestrial solutions will define the next era of AI infrastructure. While Google and Thales Alenia push boundaries, NVIDIA and Nebius focus on optimizing Earth-based systems. The coming years will reveal whether the stars hold the key to sustainable AI or if innovation remains grounded in traditional data centers.