Why missions need to get smarter
Datasets are getting larger
Satellite cameras keep getting sharper. Commercial satellites now capture details about 30 centimetres across, and every halving of pixel size quadruples the data needed to cover the same area.
Europe’s Copernicus programme alone generates up to 40 terabytes of Earth observation data a day, roughly the size of ten million smartphone photos. The volume of data satellites collect is growing faster than the capacity to send it to Earth.
Missions are going further
Missions are extending from Earth orbit to the Moon and beyond. Distance adds delay: a radio signal needs about 1.3 seconds to reach the Moon and between 3 and 22 minutes to reach Mars, depending on the planets’ positions.
A spacecraft that waits for instructions from Earth loses at least twice that time on every decision.
One-way signal time from Earth
Distance on a logarithmic scale; signals travel at the speed of light.
2 ms
0.12 s
1.3 s
5 s
3–22 min
Both trends point to one requirement: a mission needs to analyse data where it collects it, and act on the result.
Smart space missions
A smart space mission uses AI and on-board data processing to analyse data in orbit and act on the results.
- 1
Capture
Instruments record images, spectra and telemetry.
- 2
Analyse on board
A data processing unit runs AI models next to the sensor.
- 3
Decide
The spacecraft selects, prioritises or responds to what it finds.
- 4
Send results
The downlink carries results and selected data.
Real-time processing
Data is analysed as it is captured. About two-thirds of the Earth is under cloud at any moment, so a satellite that recognises cloudy images can skip sending them.
Reconfigurable in orbit
Algorithms can be updated after launch, so a mission can take on new tasks.
Autonomous operations
On-board AI detects anomalies and events and responds without waiting for a ground station.
On-board intelligence is a design decision taken at the start of a mission. It shapes the choice of computer, software and algorithms.
Building the standard for smart space missions
KP Labs, founded in Gliwice, Poland, in 2016, builds the three layers a smart mission needs: data processing units, flight software and AI algorithms. They are designed to work as one system, and mission teams can test them remotely through KP Labs’ Smart Mission Lab before choosing hardware.
- HardwareLeopard and Antelope data processing units for small satellites, both operated in orbit.
- SoftwareFlight software for command handling, payload management and in-orbit processing.
- AlgorithmsAI for Earth observation, telemetry analysis and anomaly detection.
- 10
- space missions with KP Labs computing systems or algorithms
- 35+
- projects for ESA, NASA and private clients
- ~100
- specialists, from microelectronics to AI
Why we invest
AI on Earth scaled once it had dedicated processors and a software platform developers could build on. Space AI needs the same foundation, built for radiation, tight power budgets and short contact windows with the ground.
We expect smart missions to converge on a standard stack of hardware, software and algorithms. KP Labs’ strategic goal is to become the European leader in complete hardware and software solutions for on-orbit data processing.
Visit KP Labs websiteA growing space economy, and computing as its next layer
Space is a large and growing economy. Computing in orbit is becoming a market inside it, and it reads best as three layers at different stages of maturity, each with its own buyers and its own open questions.
The space economy
| Measure | Figure | Source |
|---|---|---|
| Global space economy, 2025 | $686 billion, up 12% on 2024; commercial activity accounts for 79% | Space Foundation, 2026 |
| ESA Ministerial Council, November 2025 | €22.1 billion committed by member states, up from €16.9 billion in 2022 | European Space Agency |
| Active satellites in orbit, 2026 | About 14,000, most of them in constellations launched since 2019 | Industry estimates, 2026 |
| Luxembourg space sector, 2024 | 80 space-related companies and about 1,650 professionals; SpaceResources.lu since 2016, Luxembourg Space Agency since 2018 | Luxembourg Ministry of the Economy, 2025 |
Computing in orbit: three layers
1. On-board processing
A satellite analyses its own data. The first on-board AI experiment on a European Earth observation satellite flew in 2020, and every new constellation adds demand. Buyers are satellite manufacturers, space agencies and defence users. This is the layer KP Labs supplies today.
2. Shared computing for other spacecraft
Nodes in orbit process data for third parties and pass it on over optical links. A major cloud provider first ran its software on a satellite in 2022, and the first commercial computing nodes serving other spacecraft began operating in 2026, with optical links of a few gigabits per second.
3. Orbital data centres for terrestrial AI
Large solar-powered constellations to run AI workloads for customers on Earth. Filings with the US regulator in 2026 cover constellations from about 50,000 to one million satellites. The first prototype satellites are planned for 2027.
What has to be true
- Launch costPublished analyses put the break-even for orbital data centres at launch prices below about $200 per kilogram, a level the industry expects in the mid-2030s.
- HeatIn vacuum, heat leaves a spacecraft only by radiation, so radiator area sets the limit on how much computing power each satellite can carry.
- Radiation and reliabilityHardware runs for years without repair and has to keep working through radiation that corrupts memory and degrades electronics. Qualifying a computer for those conditions takes years of testing and flight experience.
- LinksOptical links carry the traffic. Links in service today run at a few gigabits per second; laboratory tests reach hundreds.
- Regulation and debrisNone of the 2026 filings is approved. Space agencies have questioned the debris mitigation plans, and applicants have petitioned against each other’s filings.
- DemandThe first layer sells to missions that exist today. The third sells to AI companies that currently buy terrestrial capacity, and its revenue depends on the four conditions above being met.
Where KP Labs fits
KP Labs works in the first layer and is moving into the second. Its processing units have operated in orbit, its software runs the missions around them, and its AI models are built for the radiation, power and link limits that apply in every layer. KP Labs presents the next step on its website as space data centres: shared computing resources in orbit.
Our reading of the three layers: the first has customers today and grows with every constellation; the third depends on launch prices that do not yet exist. Flight-proven compute, software and AI for orbit hold their value in both cases. If orbital data centres arrive on schedule, their operators need these building blocks. If they are delayed, satellites keep needing to process more of their own data.