SE3 raises $30M Series A to build the spatial-intelligence layer for Europe's autonomous drones

What does a drone do when the satellites go dark?

For most of the past decade the honest answer was: it stops being useful. A machine that can fly but cannot place itself in the world is a remote-controlled camera with extra steps — fine on a clear day over open ground, exposed the moment the signal is contested. Autonomy, in the military sense, has always been less about propulsion than about the map inside the machine's head.

Lukas Köstler spent years on the other side of that problem. Before co-founding SE3, he worked on Tesla's Autopilot team and at Nvidia Research — environments where the hard part was never the vehicle, but the software that turns a flood of sensor data into a usable model of the physical world. He and CTO Simon Klenk were pursuing PhDs in computer vision and robotics at the Technical University of Munich when they spun the company out of Professor Daniel Cremers' lab. Cremers, now SE3's chief science officer, holds the university's chair for computer vision and AI and is a Leibniz Prize winner.

The gap between seeing and knowing where you are

That distinction matters more than it sounds. Modern AI can describe an image, summarise a document and argue about a contract. Ask it to reason about three-dimensional space — where an obstacle sits relative to a moving vehicle, which route is exposed, how a dozen platforms should arrange themselves — and the confidence collapses. Spatial understanding is expensive to build because real-world 3D data is messy, arrives from many kinds of sensors and changes by the second.

SE3's answer is a software layer that sits between the hardware and the decision. It takes video and sensor feeds from drones and ground vehicles made by different manufacturers, fuses them into one shared three-dimensional picture, and then uses that picture to coordinate behaviour, routes and positions. Operators can issue commands by voice. The processing runs on the vehicles themselves, which is the part that matters when communications and navigation are unreliable.

A $30M bet on the layer between airframes

On 12 November, SE3 announced a $30 million Series A led by Accel, with Keen, Lakestar and Seedcamp participating. The round lands roughly five months after the company emerged from stealth with a €5.5 million seed — a fast step-up for a team whose product is mostly invisible to the end user. The company did not disclose its valuation, headcount or the value of its contracts, which is normal at this stage and also a reminder of how little of the commercial picture is public.

The money is earmarked for engineering, commercial and operations hiring, a UK office, and a deeper presence in Germany and its existing markets. Reports also point to planned offices in the Nordics and Baltics. Defence remains the immediate focus; from 2027 SE3 says it will push the same spatial layer into robotics, emergency response and other environments where machines have to cope with the unpredictable.

Why procurement wants a neutral layer

The strategic case is procurement, not gadgetry. European armed forces are moving toward modular, interoperable systems — the Modular Open Systems Approach and Germany's MOSABw programme are the clearest examples — precisely because they do not want to be locked into one supplier's stack. When every airframe maker ships its own autonomy software, mixing fleets becomes a systems-integration tax that compounds for years.

SE3 is betting that a neutral layer wins that argument. Its software does not compete with the drone; it tries to make any drone work with any other. The company says it is an AI partner to defence primes including Rheinmetall and HENSOLDT, and to autonomy scale-ups such as Auterion and ERC System. Its software runs on Auterion's Skynode computers, which gives it a route onto platforms operators already own rather than a new hardware sale.

Accel's Matt Robinson framed the bet as backing a team that can solve a problem the primes have found difficult: building spatial AI that holds up in the field. Auterion's Lorenz Meier made the interoperability case more directly — the capability runs on hardware customers already operate, across manufacturers. That is the open-architecture promise in one sentence.

What the money buys — and what it does not

The company says it has shipped thousands of systems over the past two years and holds contracts with European armed forces, including work that has been demonstrated in the field. Those claims are plausible given the partner list, but they are also company-reported: the number of systems, the contract values and the revenue attached to them are not disclosed.

The more revealing admission came from Köstler himself. The basic capability — a platform locating itself without satellite navigation — is, in his words, already robust and delivered thousands of times. The high-level fleet command AI is the younger, less mature half of the product, and he says it still needs one to two years of build-out. That is a useful corrective to the funding narrative: the round is not paying for a finished autonomy stack. It is paying for the transition from a perception layer that works to a command layer that scales.

There is also the question of what a $30 million round can actually buy in a capital-intensive sector. European defence technology has re-rated violently — Helsing reached a €12 billion valuation, Stark went from founding to €3.5 billion in two years, and Quantum Systems raised $1.2 billion in July. Against those numbers, $30 million is a Series A that buys hiring and a couple of offices, not a manufacturing base. SE3 is deliberately asset-light, which is the point, but it also means its fate depends on partners and procurement cycles it does not control.

The open question: layer or platform

Here is the judgment the round does not settle. A neutral software layer is a defensible position only while the primes prefer neutrality to control. The moment a large manufacturer decides that autonomy is too strategic to outsource — or bundles a credible equivalent into its own stack — the layer's bargaining power narrows. SE3's partnerships are impressive; they are also the exact relationships that could absorb it.

The competition is not standing still. Six Robotics raised €12 million for drone-swarm software, and the larger European players are building autonomy in-house. SE3's advantage is focus and a research pedigree that is genuinely rare. Its risk is that "neutral" becomes a nice way of saying "dependent on everyone else's roadmap."

What makes the company interesting is that it has chosen the harder, less glamorous half of the autonomy problem. Europe's drone build-out will produce plenty of airframes. Far fewer companies will own the map that lets mixed fleets act together when the signal fails. If SE3 can convert its partnerships into default status before the primes close ranks, the $30 million will look cheap. If it cannot, it will be a well-funded feature inside someone else's platform. As we wrote in our earlier look at robotics integration layers, the glue is only a business when the ecosystem agrees to be glued.

What would change this picture by 2028?

The next twelve months are the test. Watch three things: whether SE3 names a second major prime as a design partner, whether fleet-command deployments move from demonstrations to repeatable contracts, and whether the company publishes any usage numbers that go beyond "thousands of systems." Those signals will say more about the layer thesis than the size of the round.