Robotics & Physical AI Fund

Backing deep tech founders where hardware meets intelligence.

We back the full physical AI stack, from inference hardware to working robots. Pre-seed to Series A, primarily in North America.

01

The GPT-3 moment for AI dissolved moats that once belonged in the world of bits.

02

In the same vein, AI has unlocked new moats only found in the world of atoms.

03

What once were dumb mechanical machines are now being trained into intelligent thinking machines.

04

The GPT-3 moment for physical AI is on the horizon.

05

Aexodus invests where hardware meets intelligence.

Scroll to explore

In physical AI, durable value sits with whoever runs machines in the field long enough to gather real operating data, ahead of whoever ships the underlying model.

// Why now

The investment question is which machines will become commercially useful within the next decade, and which companies will capture the resulting value.

Three forces lined up at once. Here is how ↓
SCAN
01
02
03
00%
01 · Labor
There aren't enough humans to do the work

Aging populations, shifting immigration policy and reshoring, all at once. Robots are not replacing workers. They are filling roles nobody wants, or can safely do.

Sources: WEF, Future of Jobs 2025 · Deloitte, 2026 Manufacturing Outlook ↗

02 · Compute
0%

In 2023, an AI model running on a laptop could answer 23% of real-world queries as well as the big cloud models. By 2025, it was 71%. Intelligence no longer has to live in a data center. It can run on the machine itself.

Source: Intelligence per Watt · arXiv, Nov 2025 ↗

03 · Capital
0%

China installed 295,000 factory robots in 2024, almost 9x the 34,200 installed in the USA. Washington has raised its advanced manufacturing investment credit from 25% to 35%.

Sources: IFR, World Robotics 2025 · 26 U.S.C. § 48D ↗

A robotic hand, monochrome studio study, reaching to the right.
// Portfolio

We've backed the full physical AI stack. From inference hardware to working robots.

// Where the portfolio sits Selected companies · 13 by sector
InfrastructureCompute, power, hardware and orbit5
Embodied AIRobots that work alongside humans5
Frontier DeploymentAgriculture, aerial and space3
AI & Robotics Infrastructure05
Positron Positron AI inference hardware. Memory-optimized chips and servers for large models.» Rendezvous Rendezvous Robotics Space infrastructure. Self-assembling structures in orbit.» Valar Valar Atomics Factory-built microreactors. Power for data centers and industry.» Standard Bots Standard Bots Six-axis robot arms. Factory automation built in the US.» Vaxon Space Vaxon Space Air-breathing satellites. Spacecraft built to fly in very low Earth orbit.»
Embodied AI05
Apptronik Apptronik General-purpose humanoid. Logistics and manufacturing.» Persona AI Persona AI Industrial humanoid. Welding for shipyards and steel.» Agility Agility Robotics Bipedal humanoid. Material handling in warehouses and factories.» Feather Feather Bimanual mobile robot. Hardware for physical AI developers.»
Stealth Physical AI for manufacturing. Robots that assemble wire harnesses.
Frontier Deployment03
Tensorfield Tensorfield Agricultural field robotics. Precision thermal weeding.» Vantage Vantage Robotics Reconnaissance drones. Defense and public safety.» Barn Owl Barn Owl Precision Agriculture Autonomous nano tractors. Weeding and planting for smaller farms.»
// The opportunity

We think the early-stage window for physical AI is right now.

If you see what we see, drop us a line.

or
// Insights

Market outlook

All insights →
// In the media
// The companies