AI is leaving the screen. Early-stage investors should follow it.
For three years, the AI story has been told in software: models, copilots, agents. The capital is now telling a different story. According to Crunchbase News, venture investors put $47.4 billion into physical AI startups across 521 deals in the first half of 2026. That is more than the $41.9 billion raised by the sector in 2022, 2023 and 2024 combined, and roughly four times the $12 billion raised in the second half of 2025.
Physical AI covers robotics, autonomous vehicles, aerospace, drones, industrial automation and sensors. It is AI that has to work in the real world, with real consequences when it does not.
The headlines are late-stage. The opportunity starts earlier.
Much of that $47.4 billion sits in a handful of very large rounds for companies that are already household names in their fields. Those rounds are a signal, not an entry point. They show where large pools of capital, acquirers and public markets expect value to settle.
The companies that will fill the next generation of those rounds are being built now, at seed and Series A. They are usually led by engineers and clinicians who have spent years on one hard physical problem. They rarely look like a typical software pitch, and that is part of why they can be overlooked.
What physical AI looks like at the early stage
At Albion Trinity Capital we have long been comfortable backing companies that make things. The billion-dollar rounds get the headlines, but the same shift is visible at seed and Series A. A few examples of what we mean:
SkyfireAI builds autonomous drone systems for police, fire and EMS, reaching an incident in 60 to 90 seconds against five to seven minutes or more for ground units.
Phase3D gives manufacturers layer-by-layer, metrology-grade inspection of metal 3D-printed parts as they are built, for demanding sectors such as aerospace.
AiM Medical Robotics has developed an MRI-compatible robotic platform that helps physicians reach targets in the brain precisely during ablative treatment of brain cancer.
StimCardio is developing a non-invasive, wearable neuromodulation platform that treats atrial fibrillation, using continuous physiological sensing to personalise therapy during daily life.
What links them is the same shift the funding data describes. AI is most valuable when it is attached to something that senses, moves or acts in the physical world.
What makes physical AI different to back early
Defensibility is built in. A model can be copied or commoditised. A validated device, a regulatory clearance, a manufacturing process or years of clinical data cannot be recreated quickly.
The cost of building has fallen. As one investor quoted by Crunchbase put it, "costs to build these companies have come down", helped by more accessible AI infrastructure. Cheaper sensors, simulation and off-the-shelf compute let small teams reach a working prototype with far less capital than a decade ago.
The data is proprietary by design. Every deployed device collects information no one else has. Over time, that data compounds into a better product.
The risks are real, and knowable. Hardware companies face longer development cycles, supply-chain exposure and, in healthcare, regulatory pathways. We treat these as things to diligence carefully, not reasons to look away. Understanding them early is where specialist early-stage investors can add the most value.
The takeaway
The physical AI wave is not a single bet on robots or rockets. It is a broad re-rating of companies that combine intelligence with hardware, across healthcare, industry, security and infrastructure. The largest cheques are going to later-stage leaders today. The founders who will lead the next cycle are raising their first institutional rounds now.
We intend to keep meeting them.
Albion Trinity Capital and its partners hold investments in the companies named in this article. This article reflects the views of Albion Trinity Capital as of the date of publication and is provided for general information only. It is not investment advice or an offer or solicitation to buy or sell any security or interest in any fund. Third-party data is cited from sources believed to be reliable but has not been independently verified. Past trends do not guarantee future results.