We’ve been hearing a lot about data centers in space, a highly questionable concept that looks to be expensive, fragile and unnecessary. But satellites that think could speed up maritime intelligence by orders of magnitude: no orbital data center required. And a company that has worked with NASA’s Jet Propulsion Laboratory and the European Space Agency to deploy 30+ 30 Earth observation models and deliver hundreds of thousands of AI inferences in orbit just raised $11 million to scale even more.
The company is Ubotica, and CEO Fintan Buckley says it “has spent years pioneering Orbital AI” and applying “that knowledge to one of the hardest security challenges on Earth: protecting vast maritime zones and critical offshore infrastructure.”
Most Earth observation satellites are basically very expensive orbiting cameras with a radio station. They photograph the planet, beam raw pixels back to the ground, and wait for an analyst to make sense of them hours or days later.
Ubotica, an Irish space technology company, argues that this model is backwards.
The company’s $11 million funding round, led by Act Venture Capital and Greencode Ventures, will fund commercial rollout of Ubotica’s Live Maritime Intelligence platform, the first large-scale application of what the company calls Orbital AI: running artificial intelligence directly on board satellites so they can analyze what they see in orbit and send down the insight, not the imagery.
The timing is good: we’re hearing more about piracy, about international telecoms lines being intentionally severed and other forms of offshore vandalism. Nations are under growing pressure to protect critical maritime infrastructure like undersea communications cables, energy assets, pipelines and strategic shipping lanes
But at the same time that’s getting harder to track. Shadow fleets and dark vessels have turned the ocean into a security problem bigger than any one country can surveil by itself.
Ubotica’s pitch is that smart satellites should help.
Armed with AI foundation models trained on threats, smart satellites can continuously build a picture of where risk is rising across a maritime territory, then dynamically task themselves and other sensors to investigate … as opposed to waiting on fixed collection schedules and ground-based post-processing. In at least one case, when investigating incidents around Singapore, the company says it delivered actionable insight to operators in roughly twenty minutes while the underlying images that drove the insights showed up days later. In another case, Ubotica says its satellite constellation delivered world firsts such as being the first spacecraft to autonomously identify a target and reorient itself to capture it.
I asked Buckley for more details.
John Koetsier: How much faster is LMI than traditional satellite-based maritime monitoring?
Fintan Buckley: Today the model is: capture the image, downlink the raw data, then put an analyst on it. That is hours, often days. We flipped it. We process onboard, in orbit, and send down the insight, not the image. That lands in minutes, and it arrives already analysed, not as a pile of pixels waiting for someone to interpret. Over Singapore we had insights on the ground in around twenty minutes. The image itself turned up days later.
John Koetsier: What threats are you detecting today, and how accurately?
Fintan Buckley: Dark vessels, ship-to-ship transfers, spoofing, and vessels of interest loitering where they shouldn’t, near cables, pipelines and offshore energy. Every insight we send down carries a confidence level, and we are typically well into the high 90s. And it compounds. Over time we are building a digital signature of every vessel we see, so the longer we watch, the richer the picture gets.
John Koetsier: How much intelligence comes from Ubotica satellites versus third-party networks?
Fintan Buckley: LMI is not dependent on one satellite, or one sensor type. We’ve built partnerships with multiple constellation and sensor partners, and the platform can automatically task and fuse intelligence from hundreds of different assets.
That matters because maritime security is a multi-modal problem. Optical satellites are powerful when you have daylight and clear skies. SAR is critical at night, through cloud and in bad weather. RF satellites detect emissions from vessels. AIS, radar, drones and other sources add further context. The value of LMI is not any one sensor. It’s the fusion of all of them into a single, live intelligence picture.
John Koetsier: What is the current scale of deployment?
Fintan Buckley: We’re entering pilots with two nations, with more in the pipeline. The scale is whole EEZs, hundreds of thousands of square kilometres, often larger than the country’s own landmass.
[An EEZ is an Exclusive Economic Zone, a maritime area extending up to 200 nautical miles (about 370 km) from a country’s coastline, established under the UN Convention on the Law of the Sea.]
That’s exactly where traditional approaches struggle. The ocean is too large, the threats too dynamic, and human-led monitoring does not scale. LMI is built for that problem: using AI to prioritise where risk is rising, task the right sensors, and surface the intelligence that matters. Over each EEZ we build a pattern-of-life model, what normal looks like, so the unusual stands out, and every observation we take, threat or benign, makes that model sharper.
John Koetsier: The advanced AI systems that predict where risk is emerging — those are yours?
Fintan Buckley: Yes, they’re ours. They learn what normal looks like for an area, predict where risk is rising, and decide what to observe next. The risk map people see is the output of that model, not the model itself.
John Koetsier: What is the biggest technical challenge in scaling Orbital AI to large commercial constellations?
Fintan Buckley: There’s the physics. You’re operating in a brutal environment, high radiation, swinging temperatures, inside a tiny power budget, moving at around seven and a half kilometres a second, hundreds of kilometres up, with seconds to turn raw data into an insight.
But an equally hard challenge is the intelligence onboard. The future isn’t necessarily one model for one task. It’s AI agents and multimodal models capable of understanding complex scenes, identifying multiple phenomena at once, and adapting as missions evolve.
In a single observation, Orbital AI may be detecting vessels, recognising patterns of behaviour such as ship-to-ship transfers, and analysing other events simultaneously, all within the limited power and compute available on orbit.
John Koetsier: Beyond maritime, what are the next applications for Orbital AI?
Fintan Buckley: Maritime is the first application, not the limit. What we’ve really built is a way to make Earth observation cognitive and software-programmable, so the same platform takes on a new problem just by changing the models. Swap maritime models for wildfire models and Live Maritime Intelligence becomes Live Wildfire Intelligence. Same loop: predict the risk, catch the anomaly early, fuse the sensors into a picture of what’s coming. Defence, disaster response, border security, environmental monitoring, they all fit. The shift is simple. We now have satellites that can think and act in orbit. That changes what space is for.
John Koetsier: Thank you for your time.
