Israeli Physical AI — what we know, what we don't

The Israel Innovation Authority published its AI strategy report in April 2026. Buried inside is one of the most useful datasets I've found: a mapping of the Israeli Physical AI ecosystem by the Technology Research Division.

The headline number is 123 companies. The IIA calls this a conservative estimate. Total venture capital raised across the ecosystem exceeds $12 billion. In computer vision and perception alone, the report counts 61 companies, approximately $4 billion in investment, and 2,347 registered patents. Edge computing contributes another 50+ companies across hardware processors, sensors, and software. Defense AI is described as extensive but uncounted.

The names are familiar to anyone in the industry. Mobileye, whose computer vision technology for autonomous driving was acquired by Intel for $15.3 billion. Innoviz, developing solid-state LiDAR. Arbe, pioneering 4D imaging radar. Vayyar, building radar-on-chip sensing technology. Hailo, shipping edge AI processors. Deci, whose neural architecture search technology was acquired by NVIDIA. Israel's perception layer is arguably the strongest technical export outside the US and China.

But the report has gaps that matter. There is no count for agri-robotics companies. No count for underwater robotics. No count for space robotics or academic spin-offs still in stealth. No count for defense-to-commercial crossovers — companies that developed autonomous systems for military use and are now adapting them for civilian markets. The real ecosystem is almost certainly larger than 123.

The gap that concerns me most isn't in the count — it's in safety readiness. Of the 123+ companies identified, fewer than 30 show public evidence of an active functional safety certification effort. Not a completed certification. An effort. A mention of IEC 61508 on a product page. A job posting for a functional safety engineer. Anything.

The Israeli Physical AI ecosystem builds intelligence at world-class scale. Almost nobody is building the independent safety infrastructure layer underneath. That gap is the one I'll keep tracking.

Three predictions for the next 12 months. At least two Physical AI exits from Israeli companies. A major safety incident involving an Israeli-designed autonomous system somewhere in the world. And the Innovation Authority will fund at least one dedicated Physical AI safety program. I'll score myself in June 2027.