FANUC opens its robots to Google's Intrinsic and Gemini and deepens NVIDIA ties in a two-day physical AI push

FANUC Corp., Japan's largest industrial robot maker by robot revenue, spent the middle of May tying its hardware to the two largest American AI platforms in robotics. On May 13 it announced a collaboration with Google covering Gemini, Google Cloud, Google DeepMind's robot models and Intrinsic, Google's AI robotics company. On May 15 it said it was strengthening its partnership with NVIDIA in simulation, robot learning and edge computing. Both sets of technology were scheduled for live demonstrations at FANUC's new product show in May.

The most consequential line in the Google release concerns software control. FANUC said every one of its robots will be fully supported on Intrinsic's enterprise software platform, which is designed to build, test and deploy AI robot applications, and that it will fully support robot control compatible with Flowstate, Intrinsic's development environment. Intrinsic, according to FANUC, plays a leading role in the development of ROS, the open robot operating system. For a company long known for its closed controller ecosystem, that is a notable opening.

FANUC also said it has joined Google DeepMind's Gemini Robotics Trusted Tester Program to pursue research on foundation models for robots. Gemini Robotics is the DeepMind model family that turns vision and language input into robot actions. Separately, FANUC said it is building AI agents on Gemini Enterprise, Google's secure generative AI platform for companies, that can understand spoken or written instructions from workers and recognise objects, and will connect them to its robot range, which spans payloads from 3 kilograms to 2.3 tonnes.

FANUC put a number on early demand. Since the International Robot Exhibition in Tokyo in December 2025, where it first showed its open physical AI platform, the company said it has already shipped more than 1,000 FANUC robots for physical AI related uses. It did not break down customers or applications, and its April earnings call summary noted that it does not count physical AI units separately in its financial reporting.

The NVIDIA announcement on May 15 covered three areas. The first is simulation. FANUC has deepened the link between its ROBOGUIDE offline programming software and NVIDIA Isaac Sim in two configurations. In one, Isaac Sim leads and ROBOGUIDE runs underneath, so engineers can move robots in a virtual cell with a connected teach pendant as if they were handling the real machine, and use GPU acceleration for process design and sensor simulation. In the other, ROBOGUIDE leads and calls on NVIDIA's PhysX physics engine, which FANUC said makes it possible to simulate bin picking, a task that was previously hard to model in its virtual workspace.

The second is robot learning. FANUC showed a dual-arm setup built from its CRX collaborative robots that learned to fold T-shirts by imitation learning with NVIDIA's Isaac GR00T N foundation model. Operators demonstrate the task and the system then folds autonomously, adjusting its movements in real time from camera input. FANUC presented this as a way around the limits of conventional teach-and-playback programming for soft, deformable objects.

The third is edge compute. An AI collision-avoidance robot system that FANUC exhibited last year has been moved from NVIDIA's Jetson AGX Orin to the Jetson Thor T5000 module, which FANUC said raises AI computing performance by 7.5 times or more.

The context is FANUC's strongest year on record. In the year to March 2026 the robot segment grew 14.9% to ¥378.6 billion, or 44.1% of group sales, and the company guided for operating profit of ¥212.2 billion this year. Its April presentation said inquiries for physical AI had increased in Japan and overseas after the December show.

The partnerships raise the stakes for rivals. ABB Robotics and KUKA compete with FANUC for the same automotive and general industry customers, while Yaskawa Electric is pushing its own AI robot line, MOTOMAN NEXT. Among Japan's large makers, FANUC is the most profitable and has the largest robot business, as the table shows.

The open questions are commercial. FANUC did not say how Intrinsic or Gemini software will be licensed or priced, whether customers will pay Google or FANUC, or when the T-shirt folding capability will move beyond a demonstration. Opening its controllers to third-party software also creates a risk that value shifts from FANUC's own software, such as ROBOGUIDE, to the platform providers.