Portfolio
We build 20 core positions across the enabling layers of Physical AI — diversified by constraint and end market, unified by cross-system leverage.
Companies
Our first portfolio positions, each funded in 2025. Profiles are illustrative until each round is publicly announced.
Bind more. Leak less.
Seed · Feb 2025View profile →Code it. Collect it.
Seed · Mar 2025View profile →Every alert, investigated.
Seed · Jan 2025View profile →Always audit-ready.
Pre-seed · Apr 2025View profile →Ship AI you can trust.
Seed · Feb 2025View profile →Legal at the speed of business.
Seed · May 2025View profile →Your supply chain, self-driving.
Seed · Mar 2025View profile →Intelligence on every line.
Seed · Jun 2025View profile →Ask your company anything.
Seed · Apr 2025View profile →Close the books on autopilot.
Seed · May 2025View profile →Care for every plant.
Seed · Jun 2025View profile →A brain for every robot.
Seed · Jul 2025View profile →Simulate reality in seconds.
Seed · Jul 2025View profile →Discover at machine speed.
Pre-seed · Aug 2025View profile →Inspect everything. Miss nothing.
Seed · Aug 2025View profile →Eyes on everything. Threats stopped.
Seed · Sep 2025View profile →A self-driving grid.
Seed · Sep 2025View profile →Build autonomously.
Seed · Oct 2025View profile →Every scan, read instantly.
Seed · Oct 2025View profile →Autonomy for the hardest ground.
Seed · Nov 2025View profile →Every cut, perfectly made.
Seed · Feb 2025View profile →Every burn, perfectly made.
Seed · Mar 2025View profile →Every dose, perfectly sterile.
Seed · Jan 2025View profile →Every surface, perfectly figured.
Seed · Apr 2025View profile →Every batch, perfectly cultured.
Seed · May 2025View profile →Every board, perfectly built.
Seed · Jun 2025View profile →Every firing, perfectly made.
Seed · Jul 2025View profile →Every module, peak power.
Seed · Aug 2025View profile →Every box, perfectly made.
Seed · Sep 2025View profile →Every ply, perfectly placed.
Seed · Oct 2025View profile →The constraint map
Capture, curation, provenance, and feedback loops that make robot learning cheap and reliable.
Synthetic data, digital twins, and validation that closes the gap between demo and deployment.
Observability, regression, and incident systems that turn "it demoed fine" into "we know it works."
Inference, compute optimization, secure updates, and device-to-fleet orchestration at the edge.
Dexterous primitives, tactile sensing, force control, and specialized actuation.
Teleoperation fallback, maintenance intelligence, and heterogeneous coordination.
Sensors, compute, power, and hardware–software co-design — only where the asset is reusable.
End markets span logistics, health, agriculture, manufacturing, mobility, energy, construction, and inspection.
Selection
Criticality · Leverage · Defensibility · Productization · Capital fit · Capture. A company we back should score well on all six.
What qualifies. Enabling companies that own a performance-critical integration point, generate proprietary learning or benchmark data, and stay relevant to two or more credible systems over time.
What we skip. Undifferentiated robot hardware, thin wrappers around one vendor's model, consumer novelty robotics, and decks that say "multi-vertical" because they can't name a single urgent customer.
At investment we expect. A falsifiable technical result, a defined performance metric, design-partner pathways, and an 18–24 month milestone that attracts institutional follow-on.
What we expect, by stage
A defined performance metric, an early technical result, and a credible path to design partners.
Evidence the constraint you own matters to two or more credible systems — not one bespoke deployment.
A milestone that attracts a Series A: repeatable product, proprietary data, and pricing power.
The anti-portfolio
Commodity hardware. A robot or sensor with no reusable enabling asset behind it.
Single-vendor wrappers. Products that live and die with one model provider's roadmap.
Demo-stage novelty. Impressive footage, no metric, and no customer losing money without it.
Passing is most of the job. We write down why, and we stay reachable when the evidence changes.
What you get from us
The tests that turn a demo into evidence enterprises and investors trust.
Compute, hardware-test environments, and design-partner introductions across robot types.
Technical hiring support and warm introductions into the next institutional round.
Robot data, simulation, evaluation, edge inference, manipulation, fleet ops — this fund exists for you.