SE3 Raises $30M Series A for Military Drone 'Spatial Intelligence'

SE3, a Munich-based defense AI startup, has raised $30 million in a Series A round led by Accel — less than five months after emerging from stealth with €5.5 million ($6.1 million) in seed funding. Keen, Lakestar, and Seedcamp also participated.

The company, which recently dropped "Labs" from its name, builds software that lets drones and other unmanned military systems understand the three-dimensional world around them. CEO Lukas Köstler frames the core product simply: the software fuses video and sensor feeds from multiple drones and ground vehicles into a single 3D picture of the battlefield — "like a mini map, how you would know it from a strategy game," he told Fortune.

On top of that map sits a labeling layer that tags objects and their relationships, so soldiers can query it in plain language: How do I reach a target without being seen? What's a good route for a heavy vehicle?

How does the system actually work?

Commands work the same way. Instead of piloting a drone with a joystick or tapping on a tablet, a soldier can say: "I need reconnaissance two kilometers in the north over the forest," Köstler said. A suitable drone then takes off, flies the mission, and reports back.

Beyond the map-level view, SE3 builds the AI inside the drones themselves. That onboard stack handles:

• Maneuvering and navigation, even when GPS is jammed
• Object detection
• "Swarm" coordination with other drones

The most basic capability — drones locating themselves without GPS — has been delivered thousands of times and is "very robust," said Köstler, who previously worked on Tesla's Autopilot team and at Nvidia Research. The high-level fleet command AI is less mature: SE3 holds contracts with Germany's armed forces, the Bundeswehr, to develop it, and has contracts in Ukraine, where the technology is already deployed. But "there's still much that needs to be built out over the next one or two years," he said.

Why not just use a large language model?

SE3 deliberately avoids feeding raw data to general-purpose large language models (AI systems trained to parse and generate text) and expecting them to understand physics or terrain. That "wouldn't work at all," Köstler said. Instead, the company uses classical computer vision algorithms (traditional image-analysis software) to distill sensor data into a database a language model can query, and equips the model with a harness of conventional software tools it can call. If the model needs a flight path, a standard planning algorithm calculates one — avoiding mountains or enemy lines of sight.

Safety and flexibility are built in. Drone takeoffs can be set to require a human operator's approval. The system runs entirely offline, and customers can swap in their own preferred frontier AI model as the "brains." In plain terms: SE3 keeps the language model on a short leash and leans on proven algorithms for anything physical.

Who's behind it, and who's watching?

SE3 spun out of the computer vision lab of Daniel Cremers at the Technical University of Munich (TUM), where Köstler and CTO Simon Klenk were PhD students; Cremers serves as chief science officer. Military AI work has long made some researchers uneasy — a tension sharpened when the Pentagon designated Anthropic a supply-chain risk in March after the company refused to let its models power fully autonomous weapons or mass domestic surveillance. As a German company, SE3 is bound by strict export laws. "It's really the elected parliament that decides in the end what we will do," Köstler said. SE3 serves European NATO members and their allies.

What's the business bet?

SE3 is wagering that militaries want AI software untethered from any single hardware maker. If each drone maker controls its own software stack, customers struggle to make different vendors' systems work together and, within two or three years, face costly lock-in, Köstler argued. Partners include sensor maker Hensoldt, drone software company Auterion, and drone developer ERC. Accel partner Matt Robinson said SE3's team was "uniquely qualified to take on the challenge of building spatial AI for autonomous systems — something defense primes have found so difficult to do."

The company plans to grow from roughly 40 employees to about 120 and open offices in London, Berlin, the Nordics and Baltics, and Poland. Defense will drive revenue for at least the next 12 months, Köstler said, though a small team will begin exploring civilian robotics. In short: a $30 million bet that hardware-agnostic AI becomes the defense sector's default layer.