01
Constraints govern outcomes
Remove one severe bottleneck and the whole system moves. Most of what surrounds it is downstream of that single unlock, and priced as though it were not.
Pre-seed & seed · Physical AI
Fund the constraint. Unlock the system.
Paretia backs the technical bottlenecks Physical AI cannot scale without — robot data, simulation, evaluation, edge intelligence and manipulation. We fund the enabling layer, not the robot brand.
$40M
Fund I target
7
Constraint families
0
Investments to date
48h
Response time
The belief
Value is not evenly distributed across the stack.
A small number of performance-critical constraints decide whether an entire class of physical systems works. We would rather own one of those than a share of everything downstream.
02
The visible winners may not be the returns
Robot brands attract the attention and the mega-rounds. The durable seed-stage returns are more likely to sit in the enabling layer those brands all depend on.
03
Capability is outrunning evaluation
Platform capability is rising faster than the tooling needed to evaluate, adapt, secure and operate it. That gap is where a pre-seed cheque still changes an outcome.
What we back
Seven families, one test.
Each family is a place where the field repeatedly fails in the same way. We underwrite the constraint first and the company second — and we publish the map so it can be argued with.
- Robot data engines Data
- Simulation & sim-to-real Data
- Evaluation, safety & observability Evaluation
- Edge inference & orchestration Runtime
- Manipulation primitives Runtime
- Fleet & intervention infrastructure Runtime
- Selective enabling hardware Hardware
Think you own one of these constraints?
Name the constraint you control, the metric that proves it, and the systems that depend on it. That is the whole first pitch.