Decision making at the speed of light.

Silicon shuttles charge across a chip trillions of times. Light resolves it in a single pass.

QBoltz tackles NP-hard combinatorial optimization — the class of problems quantum computers are being built to solve. We do it classically, at room temperature, in a package small enough to deploy anywhere. That is the math under real-time autonomy: a drone swarm re-planning mid-flight, a robot fleet re-tasking the moment conditions change, a delivery network rerouting around a closure.

What we do

Modern machines already perceive the world well.

Deciding what to do next, which asset covers which task, which route, which schedule, which allocation, is a different kind of computation, and it does not fit on the processors machines already carry.

We compute using light interference, not electrons. Each optical pass evaluates an entire candidate solution at once, doing in a single operation what conventional hardware needs millions of operations to approximate. No cryogenics. No data center. No round trip.

The machine.

Architecture
Free-space optical computer for optimization problems.
Problem class
NP-hard combinatorial optimization, expressed as QUBO or Ising.
Scale
Millions of variables, fully connected.
Time to solution
<10 ms.
Environment
Room temperature. No cooling, no vacuum, no shielding.
Power class
Tens of watts. Battery-compatible.
Availability
Sampling in 2027. Have a problem you want to try? Write to us.

Where it applies.

Built for

  • Task assignment and resource allocation
  • Routing and scheduling
  • Network and graph optimization
  • Real-time decisions on-platform

Not built for

  • Neural network training
  • LLM inference
  • General-purpose compute
  • Replacing your GPU
Reach us