Intel orbital data centers are, for now, a patent filing rather than a spacecraft program. Intel Corporation’s application, US 2026/0230175 A1, published on Aug. 6 after a Dec. 29, 2025 filing, describes using a small group of capable satellites in higher orbits to run parts of a far larger low-Earth-orbit satellite network. Patent-tracking site Patentlyze lists the case as docketed for examination, and Tom’s Hardware reported no indication that Intel is building the proposed satellites.
The design divides a constellation into two layers. The lower layer consists of relatively simple LEO satellites, which circle close to Earth and move rapidly relative to each other and users on the ground. Above them, Intel proposes satellites in medium Earth orbit, geosynchronous orbit or highly elliptical orbit, equipped with more computing and storage capacity.
Those higher nodes would handle compute processing, network operations and coordination among satellites, according to Patentlyze’s record of the application. Tom’s Hardware reported that Intel’s filing also identifies routing, mission planning and scheduling as potential tasks. In network terms, the upper tier is a control layer: it makes decisions about how the lower tier operates rather than serving as a general-purpose cloud service for Earth users.
How would Intel orbital data centers manage LEO satellites?
Higher-orbit satellites could maintain a wider, more persistent view of a LEO fleet than any individual LEO spacecraft, Intel argues in the filing described by Tom’s Hardware. That view could let them coordinate links and traffic without repeatedly returning every control task to computers on the ground.
Satellite constellations still need to decide where traffic goes, which spacecraft should communicate and how to respond when conditions change. Today, much of that planning is handled by terrestrial network operations systems, then sent up to the satellites. Intel’s proposed arrangement moves part of that control processing into orbit. Ground stations remain part of the system; the application does not propose cutting them out.
The potential payoff, as Intel frames it, is faster handling of time-sensitive network decisions and less sophisticated hardware on each LEO satellite. Offloading heavier management work to a smaller number of supervisory spacecraft could allow simpler, cheaper vehicles in the lower layer. Those are design claims, not demonstrated results. A patent application does not establish technical feasibility, a customer, a deployment timetable or a commercial commitment.
The proposal also differs from orbital-computing concepts aimed at running large AI workloads in space. Intel’s described system is primarily infrastructure for operating a satellite network itself, closer to edge computing for a constellation than a cloud data center transplanted into orbit.
Intel’s application is a continuation of an earlier filing, according to Tom’s Hardware and Patentlyze. The new publication documents an architecture and 21 claims. It does not announce a launch.
This story draws on original reporting from Tom's Hardware.
