A satellite the size of a refrigerator can now spot a ship running dark off the coast of Portugal, process that image in orbit, and alert a naval operator within minutes without a single downlink to a ground station. That capability went from research project to funded product in June 2026, when Ubotica Technologies closed an $11 million round to scale its Live Maritime Intelligence platform.
Nations face a basic asymmetry. Their exclusive economic zones cover millions of square kilometers, often larger than their land territory. A single frigate or patrol aircraft covers a sliver at enormous cost. Satellites solve the coverage problem, but the conventional model, capture then downlink then process, introduces a delay measured in hours or even days. By then, a dark vessel has moved, transferred its cargo, or disappeared.
Ubotica's approach cuts out the middle step. Instead of sending raw imagery to Earth, the satellites transmit only the coordinates, classification, and intelligence product. The data volume drops by orders of magnitude. The latency drops from hours to minutes.
Dark Vessel Detection AI Market
North America held 38.6% of global revenue in 2025, driven by over 60 active detection programs across the U.S. Coast Guard, DHS, and the Office of Naval Research. Europe is accelerating after the 2025–2026 undersea cable incidents. · Dataintelo, April 2026
Growing: orbital AI as a service
Live Maritime Intelligence is not a future concept. The company has run hundreds of thousands of AI inferences in orbit across 30+ Earth observation models on multiple satellite missions, achieving a 100% mission success rate. Its CogniSAT-6 mission, flown with Open Cosmos, marked the world's first commercial deployment of real-time AI-based Earth observation processing on orbit. In a 2025 demonstration with NASA's Jet Propulsion Laboratory, one of its satellites became the first spacecraft to autonomously identify a target ahead along its orbital path and reorient itself to capture it.
The LMI platform closes a continuous loop: it identifies areas of elevated maritime risk from space-based data, tasks the most suitable satellite and sensor modality (optical, SAR, or RF), processes the resulting data at the orbital edge, and delivers an intelligence product to the operator. The tasking is dynamic. If a satellite passes over a high-risk zone where a shadow fleet transfer is suspected, the system can redirect a second satellite from a different orbital plane to investigate, all without a human in the loop.
The May 2026 partnership with NOVI Space, an orbital edge computing firm, gives it access to standardized compute nodes on satellite buses. That means deploying models across a wider constellation without bespoke integration on each platform. It is the difference between a custom prototype and a scalable service. The company now needs to convert that architectural advantage into recurring government contracts.
New: processing at the sensor
The shift from ground-based analysis to on-orbit inference is still early, and the infrastructure to support it is being built in real time. In February 2026, ESA officially launched Φ-lab Ireland in Mullingar, a national hub for space-enabled AI and advanced manufacturing. Ubotica and MBRYONICS are the first supported companies. ESA's Φ-lab is also researching "frugal AI" architectures: neuromorphic networks running under 5W per module, designed to push inference onto smaller, cheaper satellites.
Founded in 2016, the Dublin-based company has raised a cumulative $15.21 million across multiple rounds, with Act Venture Capital and Greencode Ventures co-leading the latest $11 million tranche alongside existing backer Atlantic Bridge. The capital is earmarked for commercial deployment of LMI across government customers, not R&D. CEO Fintan Buckley told Payload it has "the proof points now, on over ten different missions. We understand what it takes to deploy edge applications on board a satellite." That distinction between demonstrated capability and funded scale-up separates Ubotica from dozens of space-AI startups that raised seed rounds on slide decks alone. Its technology already flies on NASA and ESA missions; the question is how fast it can convert those credentials into recurring government contracts.
Act Venture Capital and Greencode Ventures are both European deep-tech funds with defense-adjacent portfolios. Greencode's Terhi Vapola explicitly tied the investment to European resilience: "Ubotica's edge AI technology enables satellite constellations to detect threats and anomalies early." The framing puts strategic security first and commercial opportunity second. That mirrors how the European defense technology market has evolved since 2022. Government customers in this space buy on capability and trust, not price. Its 100% mission success rate across 30+ model deployments and its partnerships with NASA JPL and ESA provide the trust credential.
Parallel to this, Axelera AI and ESA partnered in March 2026 to standardize space-based inference hardware, focusing on radiation hardening by design and sovereign European AI accelerators as an alternative to U.S.-proprietary chips. The TerraMind foundation model, released jointly by ESA and IBM Research Europe, provides the software backbone: a cross-sensor data fusion model that can combine SAR and optical data directly in orbit.
All of these are bets that processing power in space will follow the same cost curve it followed on Earth. If they're right, the bottleneck shifts from bandwidth to inference capacity, and the companies that own the on-orbit compute layer will hold a structural advantage over those that sell raw imagery and expect the customer to process it.
Comparison: the competitive field
The field of companies targeting maritime domain awareness from orbit is growing. Ubotica is the only one that combines on-orbit AI inference with a dynamic tasking loop. The distinction matters.
KSAT — Vake Platform
Kongsberg Satellite Services launched Vake in January 2026, integrating optical, RF, and SAR data from 15 satellite providers into a single interface. KSAT claims a 15-minute collection-to-dissemination timeline. But Vake is a ground-based fusion platform — it integrates multiple satellite feeds after downlink, rather than processing at the source. KSAT's target of sub-5-minute delivery by 2030 suggests the current latency floor is structural, limited by downlink constraints its approach bypasses entirely.
Planet — Vessel Detection
Planet's Vessel Detection uses its PlanetScope to deliver automated monitoring over vast ocean areas. Its strength is scale — near-daily revisit over 10 million km² — but the processing is ground-based. Planet is a data provider with an analytics layer. The two models are complementary: Planet's broad-area imaging provides the raw input; on-orbit processing filters it into actionable intelligence at the source.
GSTS — OCIANA Platform
Global Spatial Technology Solutions operates OCIANA, a data fusion platform specialized for Canada's Arctic and SAR operations. Like Vake, it is a ground-side aggregator of satellite and sensor data, not an orbital processing play. Its strength is multi-source fusion for harsh environments — a capability its architecture addresses through the other end of the pipeline — processing earlier rather than fusing more.
Its edge is architectural. The others optimize how much you can process once the data reaches Earth. It minimizes how much data needs to reach Earth at all.
Signals
Three developments worth tracking:
Demand-pull from undersea infrastructure. The 2025–2026 incidents involving severed undersea cables in the Baltic and North Seas created an immediate procurement driver. Governments previously content with reactive surveillance are now funding persistent monitoring of cable routes and offshore energy assets. The LMI platform was designed for this use case. The $11 million round is early proof that institutional capital sees the same gap.
ESA's orbital compute stack. The combination of Φ-lab Ireland, the Axelera AI partnership, and the TerraMind foundation model represents a coordinated European effort to build a sovereign orbital AI layer. Ubotica sits at the center of this as ESA's preferred inference provider. If the stack matures as planned, on-orbit AI shifts from a specialized capability to a standard satellite bus component, and its model deployment experience (30+ models, 100% mission success) becomes the benchmark every competitor must match.
Tier-1 defense budgets entering the market. The funding round included Act Venture Capital and Greencode Ventures, but the customer pipeline is predominantly government. CEO Fintan Buckley frames LMI as a response to "shadow fleets, dark vessels, sanctions evasion and potential sabotage." This is not commercial shipping logistics. It is a defense-intelligence procurement category that typically operates on 5- to 10-year contract cycles and carries high switching costs. First-mover advantage matters more in this market than in commercial SaaS, not less.
