We build the safety chip for humanoids.
Humanoids are about to live in homes and factories. We build the part that keeps them gentle around people.
We believe safety belongs in hardware.
A humanoid runs a powerful AI model, and anyone can update that model. We put the safety in a separate chip, so the part that keeps a robot gentle around people stays steady.
Mech-1 sits beside the model and holds the final say over the motors. It runs fixed, verified rules that you set. The robot moves freely inside those rules, and the rules hold.
We sell the seal. The chip is the start. The lasting business is the test, the certificate, and the mark that buyers and insurers trust, the way UL, Dolby, and USB became standards. We earn adoption, and the market chooses us.
We build fabless, the way ARM did. We own the safety core, the identity, and the rules, and we let others supply the heavy compute. The day a robot near a person needs a reason to wait, that reason is a chip, and the chip is ours.
What we hold to.
- Prove it on a bench
We ship a guarantee once we can show it on a bench.
- Open where it counts
The rules that govern a robot are open for anyone to read.
- A standard the market chooses
We earn adoption and let the market pick us.
- Own the seal
We hold the cert and the identity, and others supply the compute.
The thesis, on camera, before the silicon.
A Unitree G1 in simulation, where the safety core holds the final say over the robot in real time. It renders live on an RTX 4090 and streams to the browser. It shows the chip working while the die is still on the drawing board.
Today the G1 stands under gravity. Next comes a walking policy, and then the safety layer holding movements that would break the rules.
Talk to the team.
Partner, buyer, builder, or candidate. We want to hear what you are putting near people.