// INVESTMENT THESIS
Physical intelligence is the next platform shift.
Software ate the world. Now AI is learning to move through it. We believe the next decade of value creation happens at the intersection of machine learning and the physical world — and that the companies building that infrastructure will be generational.
// WHY NOW
Four forces converging.
01
Foundation models hit the physical world.
Large-scale pre-training is moving from language to perception and action. The same architectural shifts that produced GPT-4 are now producing generalist robot policies. The window to build the infrastructure layer is open.
02
Hardware costs collapsed.
Compute, sensors, and actuators that cost six figures in 2015 cost four figures today. A two-person team can now prototype a robot system that would have required a lab of 20. The barrier is no longer hardware — it's the stack above it.
03
Industrial demand is real and urgent.
Labour shortages, supply chain fragility, and reshoring pressure are forcing OEMs and Tier-1s to automate faster than their internal R&D can move. They need external innovation. The market is pulling, not being pushed.
04
The funding gap is structural.
Most VCs are optimised for software cycles. Hardware startups need longer runways, different diligence, and investors who understand TRL progression. That gap is where 1206 VC operates.
// DOMAIN 01
Physical AI
The dominant paradigm in AI has been disembodied: models that process text, images, and audio but have no grounding in the physical world. That is changing. Embodied AI — systems that perceive, reason, and act in real environments — is the next frontier.
We invest in the full stack: from the foundation models that learn generalised policies, to the perception systems that ground them in reality, to the edge inference hardware that runs them at the speed of physical action.
Embodied Intelligence
Agents that perceive, reason, and act in unstructured physical environments. Not simulation — real-world deployment.
Foundation Models for Robotics
Pre-trained action models, world models, and generalisation across tasks and hardware platforms.
Sim-to-Real Transfer
Closing the reality gap: domain randomisation, physics-accurate simulation, and transfer learning pipelines.
Perception & Scene Understanding
3D perception, semantic mapping, object detection, and spatial reasoning at the edge.
Edge AI & Inference
Running inference on constrained hardware: model compression, quantisation, and custom silicon for real-time control.
// DOMAIN 02
Robotics
Most robotics investment goes to the application layer. We go deeper. The companies that will define the next generation of robotics are building the enabling technologies: novel actuators, smarter drives, better sensors, and the control software that ties them together.
We understand the full stack — from motor winding geometry to real-time OS scheduling. That lets us diligence hardware companies properly and add value that generalist investors cannot.
Full-Stack Systems
Complete robotic systems from mechanical design through software stack — not just components.
Actuators & Motor Drives
Novel actuator architectures, high-torque-density motors, and intelligent drive electronics.
Sensors & Perception Hardware
Custom sensor fusion, proprioception, tactile sensing, and low-cost LiDAR alternatives.
Control & Motion Planning
Model predictive control, whole-body control, and real-time motion planning for dynamic environments.
Embedded Software & Safety
Real-time operating systems, functional safety (IEC 61508, ISO 26262), and hardware-software co-design.
// WHAT WE LOOK FOR
The founders we back.
Deep technical credibility
At least one founder with hands-on hardware or ML experience. We can tell the difference between a pitch deck and a working prototype.
Clear TRL progression
A credible path from current TRL to a fundable milestone. We help map it — but the founder needs to understand the journey.
Industrial insight
Understanding of who buys, why they buy, and what the procurement cycle looks like. Hardware sells to organisations, not consumers.
Coachable and collaborative
We co-build. That requires founders who want a genuine partner, not just a cheque.
// WHAT WE DON'T INVEST IN
Honest about our edges.
Pure software plays with no hardware differentiation
Consumer robotics without a clear path to industrial deployment
Single-sensor or single-component plays without system-level ambition
Teams without at least one founder with hands-on hardware experience
If you're on the edge, reach out anyway. The best founders often don't fit neat categories.
// ARCHITECTURE NOTE
These two domains are our launch focus. The platform is designed to add verticals as the market evolves — climate hardware, space systems, biomedical robotics. If you're building at the frontier of a new physical domain, we want to hear from you.