Kawasaki Has a Robot Surgeon. It Cannot Leave Japan.
Kawasaki Has a Robot Surgeon. It Cannot Leave Japan.
Kawasaki Heavy Industries built a surgical robot called Hinotori. It received Japanese regulatory approval in 2020 Kawasaki Robotics press release. Five years later, it has not been cleared by the FDA, has no CE marking, and has not been deployed in a single US or European hospital. The robot works. The paperwork does not exist.
That is the gap Kawasaki is now trying to close with a Physical AI Center in San Jose, California, announced May 22, 2026. The center's stated first priority is healthcare and elder care. Jensen Huang, whose company Nvidia is one of the four named partners, delivered a video message at the opening ceremony in which he named Hinotori as one of the systems that would benefit from the partnership.
The center's purpose, as Kawasaki describes it, is to accelerate deployment of physical AI by combining Nvidia's simulation platform with Microsoft Azure's cloud infrastructure, Analog Devices' sensing hardware, and Fujitsu's integration services. The Nvidia stack — Omniverse for digital twin simulation, Isaac Lab for training environments, and Jetson for onboard edge AI — is presented as the core infrastructure.
The regulatory problem that infrastructure is meant to solve is not small. Medical robots approved in Japan must run separate, full clinical trial programs to reach US or European markets. Hinotori has cleared Japan, Singapore, and Malaysia. It has not cleared the United States or Europe. Kawasaki has not filed an FDA 510(k) submission for Hinotori as of May 2026, according to Kawasaki Robotics' own press materials and the product page for Medicaroid, the Kawasaki-Sysmex joint venture that manufactures Hinotori. If a submission exists, Kawasaki has not announced it.
The bet the center represents is that simulation — specifically the Nvidia Omniverse-Isaac Lab pipeline — can compress the evidence-building phase of regulatory clearance. Simulated surgical outcomes, in this framing, are not a substitute for clinical data but a complement to it: faster to generate, repeatable, and designed to satisfy regulators that the system's failure modes are understood before a single patient is touched. Whether the FDA or EU notified bodies will accept simulation-heavy submissions for a novel surgical robot is an open question. No clearance pathway for Hinotori was described in the announcement.
The geographic structure of Kawasaki's medical robotics push is worth noting. Medicaroid opened a training center at IRCAD France in 2025. Kawasaki opened its Physical AI Center Strasbourg in March 2026, at the same IRCAD campus. Jacques Marescaux, who founded IRCAD and pioneered computer-assisted surgery, delivered a video message at the San Jose opening. Kawasaki's chief executive, Yasuhiko Hashimoto, founded Medicaroid and served as its president from 2013 to 2023. This is not a coincidence: Kawasaki is threading a medical robotics corridor from Kobe to Strasbourg to San Jose, using each site to serve a different regulatory jurisdiction.
What the announcement does not contain is equally revealing. There is no committed capital figure for the center, no disclosed engineering headcount, no timeline for an FDA submission, and no named hospital or clinical trial partner in the United States. The partners are named; the depth of their integration is not described. Kawasaki is describing an ambition, not a budget.
The second-order question is whether this matters beyond Kawasaki. If simulation-based validation works for Hinotori — if Nvidia's stack genuinely allows a surgical robot to accumulate the safety evidence that regulators require without years of physical clinical trials — it becomes a template for every hardware-first Japanese medical device company stuck behind the same regulatory wall. Japanese industrial robotics expertise is deep. Japanese medical device export has been constrained by a regulatory pathway that rewards companies with the patience and capital to run multi-year trials in target markets. A shortcut changes that calculus.
For now, the robot works in Kobe. What happens in San Jose determines whether it ever works anywhere else.