Intel envisions data centers in orbit to control constellations of thousands of satellites

Intel has filed a patent proposing to move part of the “brain” of large satellite constellations from ground control centers into space. The architecture envisions a two-tier network: thousands of relatively simple satellites in low Earth orbit (LEO) and a much smaller number of more powerful satellites in higher orbits, responsible for tasks such as mission planning, route calculation, link coordination, and network management. The US patent application US 2026/0230175 A1 was published on August 6, 2026, continuing a family of patents initiated by Intel several years earlier.

The key points of Intel’s orbital data centers in 20 seconds

• Intel proposes placing computing and control nodes in MEO, GEO, or high elliptical orbits.
• These nodes would manage large LEO constellations without relying continuously on ground stations.
• This could enable simpler LEO satellites and reduce certain response times.
• It is not the same concept as space-based data centers for AI training.
• The patent does not mean Intel will manufacture or launch these satellites.

This idea is particularly interesting as it arrives right when commercial constellations are expanding from a few dozen satellites to networks potentially made up of thousands of units. Services like Starlink, Amazon’s Project Kuiper, and future communication systems must constantly decide how to route traffic, which links between satellites to use, how to allocate frequencies, and how to handle failures or interference.

Much of that intelligence currently depends on terrestrial infrastructure. Intel’s proposal aims to reduce that dependency.

A small number of satellites would act as the constellation’s brain

The patent describes satellites positioned in Medium Earth Orbit (MEO), Geostationary Earth Orbit (GEO), or Highly Elliptical Orbit (HEO) that would feature vastly greater processing and storage capacities than the LEO satellites below.

These upper nodes would conceptually act as a combination of space-based data center and Network Operations Center (NOC).

Their role would not necessarily be to provide the final service to end users but to manage the overall network behavior.

Tasks described include:

• mission planning and scheduling;
• generating and updating routes;
• selecting links between satellites;
• assigning frequencies;
• monitoring telemetry;
• responding to failures;
• coordinating antennas and links;
• assessing satellite status.

The architecture would be hierarchical:

Ground control center
          │
          ▼
 Satellite NOC / Data Center
      MEO / GEO / HEO
          │
   ┌──────┼──────┐
   ▼      ▼      ▼
 LEO    LEO    LEO
   │      │      │
End users / ground stations

This concept resembles the evolution of traditional computer networks. Instead of forcing each peripheral node to make complex decisions independently, a top layer concentrates part of the intelligence.

The challenge of controlling thousands of moving objects

A LEO constellation presents a problem that doesn’t exist in conventional terrestrial networks.

Satellites constantly move at high speed relative to the surface and to each other, causing the network topology to change permanently.

A valid link now may become invalid just minutes later.

The system needs to know:

• which satellites are visible to each other;
• which ground stations are available;
• what optical or radio links are capable;
• which route offers the lowest latency;
• how much bandwidth remains available;
• which equipment has issues;
• how coverage changes over a specific area.

In traditional architectures, much of these calculations are performed on the ground, with instructions sent back to the constellation.

Intel argues that this model introduces operational latency, increased dependency on ground stations, and growing complexity as the constellation size reaches thousands of units.

Moving part of this processing into space could reduce some of these round-trip communications to ground control.

Cheaper and simpler LEO satellites

This proposal has another potentially significant consequence.

If a LEO satellite does not need to execute all the logic required for constellation management, it could incorporate less processing capability.

The model would shift from having thousands of relatively smart satellites to many simple nodes plus a few very powerful ones.

This could reduce mass, power consumption, complexity, and possibly the cost of the LEO vehicles.

In large constellations, a small cost reduction per satellite can add up to a significant difference when producing thousands of units.

However, this simplification pushes the problem elsewhere.

The higher-orbit satellites would need:

• high-performance processors;
• storage capacity;
• high-capacity inter-orbit links;
• redundancy;
• radiation tolerance;
• thermal management systems;
• high availability.

The network also must continue functioning if one of these upper nodes becomes unavailable.

Hence, this architecture introduces a new challenge: preventing orbital data centers from becoming single points of failure.

It’s not the same as putting an AI factory in orbit

The term “space-based data center” can be misleading.

Intel is not primarily describing a facility dedicated to training AI models or performing large cloud loads for terrestrial users.

Other projects explore that approach.

For example, SpaceX has proposed satellites specifically designed for AI computing, while Google has studied Project Suncatcher, exploring the use of orbital solar energy to power AI accelerators in space.

Intel’s focus is different.

In Intel’s design, space computing is primarily used to make the space infrastructure more autonomous.

This is more aligned with edge computing than with moving AWS or Azure into space.

The constellation would process its own data right where it is generated.

The patent has been developing since 2022

The August 6 publication does not mean Intel conceived the idea this summer.

The application US 2026/0230175 A1 was filed on December 29, 2025, as a continuation of a previous application filed in February 2023, which in turn claimed priority from documentation from 2022. The parent application was granted as US 12,542,604 B2 on February 3, 2026.

The inventors listed in the documentation include Stephen T. Palermo, Valerie J. Parker, and Udayan Mukherjee.

Intel has long been researching non-terrestrial networks. Other patents linked to their researchers cover mission planning for NTN, satellite links, mobile ground stations, and techniques for integrating 5G networks with satellites.

This work fits into a broader line of research on managing future space networks.

The biggest bottleneck will be inter-orbital links

Moving intelligence into space alleviates some problems but introduces others.

The most obvious is the communication between the upper tier and LEO satellites.

If thousands of vehicles depend on a limited number of orbital control nodes, links between levels must provide:

• high capacity;
• low latency;
• constant availability;
• encryption mechanisms;
• interference tolerance;
• alternative routing options.

Optical laser links could play a significant role here.

Currently, laser links are used between LEO satellites to transport traffic without immediately routing through ground stations.

An architecture like Intel’s would expand this idea: links wouldn’t only carry user traffic but also control information and constellation management data.

Greater autonomy also raises new security questions

Centralizing decision-making in space has important cybersecurity implications.

A node responsible for calculating routes and coordinating thousands of satellites would become a particularly critical target.

Protection should cover:

• control software;
• command authentication;
• remote updates;
• telemetry;
• cryptographic keys;
• inter-orbital communications.

Additionally, mechanisms would be needed to prevent software errors from propagating from the control layer through the entire network.

The advantage is that a more autonomous constellation could continue to operate temporarily even if communication with certain ground stations is lost.

This is especially relevant for military networks, critical infrastructure, or emergency communications.

A patent does not mean Intel will build satellites

This is perhaps the most important nuance.

Intel has not announced an orbital constellation or a plan to build these data centers in space.

The documentation describes a possible architecture and protects related intellectual property.

Big tech companies file thousands of patents, but only a fraction become commercial products.

In fact, the available information indicates that the patent application is still under review.

So, it’s premature to talk about an upcoming Intel constellation.

What’s noteworthy is the technological direction it reflects.

Satellite networks are growing to the point where managing them solely from Earth may no longer be the most efficient solution.

If constellations eventually comprise tens of thousands of satellites, part of the intelligence will likely need to shift into the network itself.

Intel proposes a concept similar to what happened with the internet and cloud computing: bringing processing closer to where the traffic is generated.

Only this time, the edge would be thousands of kilometers overhead.

Frequently Asked Questions

Does Intel want to build space data centers?

Intel has filed a patent describing satellites with high computing capacity intended to manage other constellations. They have not announced plans to manufacture or launch such satellites.

What would these space-based data centers do?

They would handle tasks like mission planning, route calculation, link coordination, frequency selection, and supervision of LEO satellites.

Would they be used to train artificial intelligence?

That is not the primary aim described in the patent. The architecture is mainly designed to manage satellite networks, unlike other projects exploring moving AI workloads to space.

In which orbits would they operate?

Intel envisions control nodes in Medium Earth Orbit (MEO), Geostationary Orbit (GEO), or Highly Elliptical Orbit (HEO), while the managed constellations would mainly operate in LEO.