NVIDIA, SpaceX and Google Advance Announced Orbital Computing Initiatives

NVIDIA, SpaceX and Google are pursuing separate orbital computing initiatives, signalling increased activity around space-based AI and data-center infrastructure. The reported developments encompass an NVIDIA space-computing platform, SpaceX’s A/1 AI data-center satellite and STARMIND concept, and Google’s planned Project Suncatcher demonstration.

NVIDIA presented computing infrastructure for space data-center applications and its Space Computing direction in March 2026. In June 2026, SpaceX disclosed information on A/1, described as a first AI data-center satellite intended to operate in orbit. SpaceX is also developing STARMIND, a concept for a space-based AI data center comprising up to 1 million AI satellites. This is a conceptual maximum, not deployed computing capacity.

Google’s Project Suncatcher is planned as a constellation of small satellites equipped with TPU AI chips and solar panels. A space demonstration is expected in early 2027. The source also reported plans to launch two Planet Labs prototype satellites for the project; the launch was described as planned and is not confirmed as completed.

The developments point to an effort to shift some terrestrial data-center constraints—power availability, heat management and site limitations—into an orbital infrastructure model. That shift would replace them with different constraints, including launch cost, satellite performance and reliability, failure servicing, and the economics and security of space infrastructure.

One unnamed US research organization, cited in the source, forecasts that up to 18,600 data centers could be operating in space by 2035, with 1.5 GW of effective orbital computing capacity. The 1.5 GW figure refers to projected IT computing capacity, rather than confirmed facility power or deployed capacity. The forecast should be treated cautiously because neither its methodology nor the underlying organization was identified.

For infrastructure suppliers, orbital data centers could represent a longer-term source of demand for radiation-hardened accelerators, satellite platforms, solar generation, inter-satellite connectivity and autonomous compute operations. However, the initiatives described remain concepts, announcements or planned demonstrations. They are not evidence of commercially operating orbital data centers, and the source provides no CAPEX, confirmed capacity, launch schedule or commercial-readiness data for specific projects.

The article also notes European work by ASCEND on assessing whether locating data centers in space could reduce CO2 emissions. Its broader claim that solar generation in space can be more than ten times as efficient as on Earth is presented without a methodology, comparison conditions or primary source and should not be read as a project-specific performance metric.