The same people who taught cars to see are now teaching your office building to think.
The Summary
- Coram AI raised $35M Series B, bringing total funding to $66M, to turn existing security cameras into predictive AI systems that spot problems before they happen
- Founded by former Lyft and Zoox autonomy engineers who are applying self-driving car vision tech to physical security across 1,500+ sites
- Physical AI VC investment exploded from $4B in 2019 to $26B in 2025, with $23B raised in just the first half of 2026
The Signal
Security cameras have been dumb witnesses for decades. They record everything and see nothing. You only check them after the theft, the slip-and-fall, the unauthorized entry. Coram AI is betting that $66 million that the same computer vision that taught autonomous vehicles to navigate highways can teach cameras to actually watch.
The founding story here matters. CEO Ashesh Jain led autonomy at Lyft's self-driving division. Cofounder Peter Ondruska ran AI research at Lyft and Toyota's Woven unit. These aren't security people trying to bolt on AI. These are autonomy engineers who realized the tech stack they built for robotaxis works better when the environment stays still.
"Our earlier mission was to protect every public road, and now it is to protect every physical space."
The business model is elegant: Coram doesn't sell cameras. It plugs into the ones you already have, plus badge readers, visitor logs, emergency systems. The AI layer sits on top, turning passive surveillance into active monitoring. Hershey's Ice Cream and multiple school districts are already running it across 1,500+ sites. The value prop isn't "rip and replace," it's "make what you bought actually work."
But here's the real play: Coram is infrastructure for the physical AI layer that's coming. Jain told Business Insider the platform could eventually integrate with robot dogs and humanoid robots. That's not product roadmap fluff. That's the whole thesis. Security cameras are just training wheels.
Key data points:
- Global robotics and physical AI VC funding: $4B in 2019 → $26B in 2025
- $23B raised in physical AI space in first half of 2026 alone
- Coram deployed at 1,500+ sites in less than four years
Physical AI is having its infrastructure moment right now. Everyone's building robots and agents that need to see and act in meat space. Those systems need eyes everywhere. Coram is building the nervous system. The $35M Series B from Ansa Capital and Battery Ventures says institutional money sees this as the picks-and-shovels play while everyone else fights over which humanoid is prettiest.
The autonomy-to-security pipeline is also worth watching. We're seeing talent migration from self-driving (crowded, capital-intensive, regulatory hell) into physical AI applications with clearer ROI and faster deployment cycles. Coram's traction suggests the pattern recognition and sensor fusion techniques from AV work translate directly. That brain drain is going to accelerate.
The Implication
If you're running physical spaces—warehouses, factories, retail, schools—the question isn't whether AI vision gets added to your security stack. It's whether you control that layer or rent it. Coram's bet is that most organizations will rent, because building this in-house requires autonomy engineers you can't hire.
Watch for the robotics integration. When humanoids and robot dogs start doing rounds in your building, they'll need to share a visual understanding of that space with the static cameras. Whoever builds that common operating layer wins the physical AI platform game. Coram is positioning for that.