Robotics & embodiment
Demos are cheap. Uptime is the product.
The thesis
Generalist manipulation policies are improving, but the buyer does not purchase a policy. They purchase a cost per completed task, at a reliability threshold, with a service contract behind it. Every robotics company is eventually an operations company with a hardware line item.
That means the winners look less like model labs and more like fleets: deployed hours, teleop fallback that trends to zero, and a data flywheel that only turns if the fleet is actually in production.
Unit economics
The honest model is payback period, not gross margin. Hardware cost, integration cost, downtime, and remote-operator load all sit above the line. A robot that needs a human every twenty interventions is a staffing agency with extra steps.
The bear case
Structured environments already have cheap automation, and unstructured ones punish reliability gaps brutally. The middle band where generalist robots beat both is narrower than the fundraising decks assume.
What we'd watch
- Interventions per autonomous hour, disclosed honestly
- Fleet hours in production versus pilot
- Bill of materials trajectory on the second hardware generation
- Whether customers renew on outcome pricing
Pressure-test this note with the desk — the agent pulls live sources before it answers.
Ask the desk