Satellites + Data Centers

Today, there are roughly 13,000 active satellites in orbit, of which ~80% are in Low Earth Orbit (LEO). Here on Earth, investment in artificial intelligence (data centers, compute, foundational models, application layers, and more) is accelerating. This technological leap forward is also impacting space, and specifically satellites (which have both defense and commercial use cases).

We wrote about the satellite industry back in August with our Satellites & Digital Markets (+$157bn / year) piece, and today we are spending more time on how AI is addressing problems around routing, traffic, anomaly detection, and connectivity speeds. Additionally, we think mesh networks and data centers in orbit are coming quickly and will unlock new use cases.

In this piece, we give a quick overview of the debris belt that is building in space, explain a few use cases for how AI is helping satellites navigate orbit, and dive into why data centers and mesh networks of satellite fleets are coming to the skies very soon.

Debris in Space: Traveling at 17k mph

Since the dawn of the space age in 1957, humanity has launched roughly 16,000 satellites. Of those, about 11,700 remain in orbit, and roughly 7,500 to 8,000 are still operational.​ That leaves several thousand inactive satellites circling Earth at extraordinary speeds, alongside used rocket propulsion systems and millions of smaller debris fragments.

The European Space Agency (ESA) released a great video 6 months ago, “Space Debris: Is it a Crisis?”, which visualizes how congested orbit is becoming. For reference, the orbital velocity at about 250 miles (400 kilometers), the altitude of the International Space Station (ISS), is close to 17,100 miles per hour (27,500 km/h).

At that velocity, even a small fragment of debris can cause significant damage to any spacecraft in orbit. ESA’s 2025 Space Environment Report estimates that Earth orbit now contains about 40,000 tracked debris objects larger than 10 centimeters, roughly 1.2 million between 1 and 10 centimeters, and an astonishing 330 million particles larger than 1 millimeter but smaller than 1 centimeter.​

Avoiding these objects is becoming an increasingly pressing issue, and the ISS alone has had to conduct more than 30 collision-avoidance maneuvers since 1999, several just in the past few years.

AI & Satellites: Intelligent Orbits, Smarter Skies

AI systems are already transforming satellites from reactive controlled vehicles into proactive and dynamic systems. These systems analyze orbital traffic (object avoidance), make autonomous decisions, and optimize their network performance; often without human oversight.

In an increasingly crowded orbit (especially LEO), it is becoming paramount that these expensive satellites leverage AI to optimize their performance (and not run into objects, specifically debris and floating fragments).

Note: in this piece we are largely referring to classical machine learning and AI techniques like neural networks and deep learning that satellite systems are increasingly leveraging. We are not referring to LLMs, like ChatGPT or similar conversational AI technologies.

Here are a few ways that satellites are currently leveraging autonomous systems, largely powered by AI technology on Earth:

1) Inter-satellite communications – Satellites can now handle a significant rise in inter-satellite data traffic, thanks to AI-enabled routing technologies. Projects like Starlink, Kuiper, and OneWeb are deploying reinforcement learning models to autonomously reroute network traffic without needing ground intervention (reducing latency).

2) Maintenance monitoring - AI models are now monitoring the satellite’s systems for temperature fluctuations, radiation exposure, and mechanical wear in real time. This is a jump in performance from the legacy SCADA (Supervisory Control and Data Acquisition) system that was originally developed in the early 1960s that most satellites have been running on. While reliable, SCADA still required human oversight and could not adapt dynamically; AI systems today can. This can extend the lifespan of satellites and save the owners and operators millions over time.

3) Coverage recovery – AI’s ability to dynamically allocate connectivity and coverage back to earth is invaluable during natural disasters when cell towers and power lines on earth get destroyed. AI-driven satellite systems (like those from Eutelsat and JSAT) are restoring connectivity much faster, enabling people on Earth to regain access to the internet, communication, and emergency coordination tools much sooner after a disaster. This is crucial for saving lives, coordinating aid, and stabilizing affected infrastructure.

Data Centers in Low Earth Orbit (LEO)

The next chapter of satellite intelligence is extending beyond Earth-to-space communication to computation in space. As launch costs continue to fall (down 70% since 2015) and onboard compute capacity rises, satellites may soon function like edge servers, extending cloud capabilities beyond Earth’s atmosphere.

Here are a few predictions of how we think it will evolve over the next decade:

1) Orbital Data Centers – the satellite data center market is already in motion. Companies like Lonestar are looking to put a data center on the moon, and Starcloud is trying to put one in orbit. Both U.S.-based, they are experimenting with small data centers on satellites. Combined with solar power and vacuum-based cooling, orbital edge computing could reduce space-to-ground latency. For example, a satellite constellation in Low Earth Orbit (LEO) can achieve ~20 ms round-trip latency from Earth-to-Space and a large coverage area.

Having data centers in space is also worth considering because 1) it’s already cold and 2) the availability of un-obstructed solar energy in space for these data centers provides a reliable source of energy. One of the hurdles will be the maintenance of these data centers, so that will be a cost headwind.

Data centers in space are attracting global interest as groups like Madari Space, based in Abu Dhabi, are launching early versions of computational capacity into space. In May of this year (2025), China launched 12 satellites for a space-based computing constellation — the first of a proposed 2,800-satellite fleet to process data in space.

2) Fully Autonomous Constellations - as satellites and fleets of satellites further adopt AI technology, the coordination amongst thousands of satellites is going to be profound. They will all be able to talk, learn, share information, and reconfigure their real-time networks to better serve Earth.

Future constellations will deploy on-device machine learning, allowing satellites to collaboratively train AI models across nodes rather than centralizing data. In this distributed approach, each satellite (or node) uses its locally collected data to train a portion of the model and periodically shares updated parameters, rather than raw data, with others or a central aggregator.

This approach improves privacy, reduces downlink load (the volume of data sent from a satellite to Earth during a given time), and enables real-time decision making independent of ground infrastructure. This will be a step change to the chaos of Earth-led satellite management today. As we approach 50,000-100,000 satellites in orbit by 2035, it is imperative that AI technology help us manage this space fleet.

3) Quantum Secure Communications - satellite companies are experimenting with new technology to detect data interception during communication. For example, Quantum Key Distribution (QKD)  transmits quantum-encrypted keys between space and ground stations using photons. South Africa and China just set a record of the longest quantum-secure communication link between satellites. NASA’s breaks this technology down in an extensive 95 page report.

As this is further implemented, this technology makes eavesdropping immediately detectable and allows satellites to establish unbreakable encryption links over vast distances, overcoming the range limits of terrestrial fiber systems.​

To strengthen this, satellites can also use Post-Quantum Cryptography (PQC)—mathematically resilient algorithms that protect data from future quantum computers. Together, QKD and PQC will enable globally secure communication networks, ensuring cryptographic safety for governments, enterprises, and financial systems even in a quantum-capable world (all enabled by satellites).

Takeaway: Satellites in space will unlock incredible progress for humanity on Earth and beyond. It is very early for the space industry, and the chaos of debris, traffic, and collisions even this early on in the commercial space race is evidence of the need for better coordination and technologies. We believe that AI and quantum technologies will increasingly bring robust advancements to communications, autonomous satellite fleet control, cost savings, and benefits to Earth. Additionally, the rise of compute in space will lead to incredibly efficient data centers operating in orbit with low cost cooling and unobstructed solar power. The next decade of space, especially in LEO, will be incredibly exciting.