The startup that decided the answer to a cheap drone isn't an expensive missile - it's a beam of light that thinks. Meet Laser Armor.
Somewhere over a power plant, a $500 quadcopter hums where it shouldn't. The old playbook fires a missile that costs more than the building it defends. Thor Dynamics thinks that math is backwards.
The threat arrives quietly. A drone - or forty of them - drifting toward a substation, an airport fence, a forward base. It is small, disposable, and it does not care how much your defenses cost. For a decade the standard reply was to spend thousands, sometimes millions, to swat something that came off a hobby-shop shelf. That asymmetry is the whole game, and until recently the drone was winning it.
Thor Dynamics, a defense-tech company headquartered in Palo Alto, was built to flip the equation. Its product, Laser Armor, is a directed-energy system: a high-power laser that detects an incoming drone, locks on, and burns it out of the sky. No magazine to reload. No interceptor to buy. The marginal cost of a shot is measured in electricity, not ordnance. The beam travels at, well, the speed of light - which turns out to be a useful property when the target is fast and the decision window is a heartbeat.
But a laser is only as good as its aim, and a drone swarm is a moving, dodging, multiplying nuisance. So the interesting part of Thor Dynamics isn't the photons. It's the brain steering them.
Laser Armor runs a simple-sounding loop. The complexity hides inside how fast it closes.
Reinforcement learning is the family of AI best known for mastering games - the software that learned to beat humans by playing itself millions of times. Thor Dynamics took that idea off the game board and pointed it at the sky. Laser Armor's targeting improves in real time, optimizing how it tracks and engages fast, maneuvering, swarming drones instead of a single obliging target.
To run that intelligence where it matters - at the edge, on the platform, with no time to phone home - the company partnered with NVIDIA. Laser Armor integrates NVIDIA's Jetson and RTX platforms to process sensor streams, run neural models, and make split-second targeting decisions. The partnership, announced in early 2025 and later deepened, is what lets a 20-person startup talk credibly about counter-swarm defense.
An AI-powered directed-energy counter-drone system that detects, tracks, and eliminates aerial threats in seconds, with up to 20kW of output and minimal collateral damage.
Reinforcement-learning algorithms that optimize engagement in real time - the difference between hitting one drone and holding off a swarm.
Core directed-energy subsystems engineered for accuracy, reliability, and scalable deployment - including non-ITAR configurations built for export.
Multi-sensor tracking, target radar, and range-finding beacons feeding the Monitor-Detect-Track-Eliminate loop.
It's an unusual founding table: a serial AI entrepreneur, a former U.S. Secretary of the Navy, and an operator to run it all.
Return to that quadcopter humming where it shouldn't. In the old story, it wins - either it gets through, or it forces someone to spend a fortune stopping it. In the story Thor Dynamics is trying to write, the sky is already being watched. The system sees the drone, the reinforcement-learning model does the arithmetic of the intercept, and a beam that costs pennies makes the problem disappear before the operator finishes their coffee. Then it goes back to watching, because the next one - or the next forty - could arrive at any moment.
Thor Dynamics is still small, still early, still a seed-stage company betting that light plus intelligence beats metal plus money. Whether it becomes a fixture of critical-infrastructure defense or a footnote in the counter-drone race is unwritten. But the substation, the airport fence, the forward base - they no longer have to flinch at every hobby drone on the horizon. That is the change. Everything else is aim.