Exein hits $1.7bn valuation with $270m for physical AI
The Rome-based runtime security company says its round, led by Headline, will fund US and Asia Pacific expansion as attacks on connected machines accelerate.
Rome-based Exein has become Italy’s newest unicorn, closing a $270 million round at a $1.7 billion valuation to build a security layer for robots, drones, vehicles and the other machines that use AI to act in the physical world.
Who is backing it
The round was led by Headline, with Sofina, Goldman Sachs, the EIB Group through the European Tech Champions Initiative, KfW Capital and T.Capital taking part, alongside existing investors including Balderton, HV, Intrepid Growth Partners, 33N Ventures, Lakestar, Supernova Invest, Blue Cloud Ventures and Geodesic Capital. Exein says the raise closed significantly oversubscribed, and the money is earmarked for expansion in the United States and Asia Pacific.
Attacks at machine speed
Exein’s argument is that connected machines need to defend themselves whether or not the network around them can be trusted. Its Photon runtime architecture, released earlier this year, works at kernel level to stop malicious code executing before it runs. The company says it sees around 5,000 new, non-repetitive attacks a week across a network of two billion devices — five times last year’s volume — as AI makes it cheaper to target hardware that moves, runs grids or operates medical equipment. Exein is also training a foundation model on two years of machine telemetry, with its agentic security architecture shipping by the end of 2026 and the first models expected in the first quarter of 2027.
Our opinion
Physical AI is where the cloud’s security debt gets real: a compromised server leaks data, but a compromised robot takes an action. Working at kernel level is a coherent answer, and Europe’s Cyber Resilience Act hands Exein a regulatory tailwind its American rivals cannot entirely ignore. The open question is whether a two-year head start on machine telemetry survives the arrival of better frontier models — a defence built on understanding how machines behave is only as strong as the machines it has watched.