Nvidia Rallies Fujitsu, Fanuc, Yaskawa and Kawasaki Around Physical AI

Nvidia said Wednesday that Fujitsu is exploring a shared control platform for physical AI, built alongside Fanuc, Yaskawa Electric and Kawasaki Heavy Industries. The announcement folds four of Japan’s biggest industrial names into a coalition that now runs to at least ten companies.

The wording is softer than the headlines suggest. It also lands weeks after two of those same manufacturers quietly kept working with Nvidia’s rivals too.

Fujitsu Steps Into the Middle of Nvidia’s Japan Play

The announcement, dated from Tokyo on July 15, describes an initiative led by Fujitsu that aims to build a collaborative control platform integrating Nvidia’s physical AI stack to bridge digital and physical operations across all industrial sectors. The platform is built with Cosmos world foundation models, the open Isaac robotics development platform and the Newton physics engine, supporting AI model development, digital twins, robot learning, simulation-to-real workflows and pre-deployment validation.

Alongside it, Japan’s physical AI ecosystem leaders AIRoA, Fanuc, Fujitsu, Hitachi, Kawasaki Heavy Industries, Kubota, NEC, SoftBank Corp., Sony Group Corporation and Yaskawa Electric intend to join the Nvidia Cosmos Coalition to help build open frontier physical AI models. That is ten named Japanese companies inside a single announcement, and the roster reaches into agriculture and construction too, not just factory floors.

The new model doing the heavy lifting is Cosmos 3 Edge. Nvidia says it is lightweight enough to run on Jetson, including new T2000 and T3000 modules, as well as RTX GPUs and DGX systems, enabling local, real-time reasoning and robot action prediction on edge computers. Fanuc’s own product materials describe the mechanics on the shop floor: Fanuc is collaborating with Nvidia to advance physical AI in industrial robots, and Fanuc robots will realize a digital twin in a photorealistic virtual factory built on Nvidia Isaac Sim, an open-source robotic simulation framework built on Omniverse.

Beyond the four headline names, the coalition includes smaller, sharper bets:

• AIRoA – Japan’s AI Robot Association, established in 2025 to build foundation models for AI-powered robots, sits behind much of the coordination work.
• Hitachi – is developing physical AI that connects information technology and operational technology systems inside factories.
• SoftBank Corp. – is building its own physical AI development platform on top of Cosmos, Omniverse and Isaac Sim.
• Telexistence – is applying Isaac tools and testing Cosmos for retail automation.
• GROOVE X – is building its Jetson-powered LOVOT companion robots on the same stack.
• Enactic – is fine-tuning the Isaac GR00T model for semi-humanoid elder care robots.

Each of those is a small, specific bet. None of them is a finished factory floor yet.

Why Nvidia Picked Now to Widen the Net

The timing traces back almost a year. Fujitsu announced an expanded strategic collaboration with Nvidia to create full-stack AI infrastructure that integrates AI agents in an October 2025 release covering healthcare, manufacturing and robotics. That collaboration paired an AI agent platform with computing infrastructure integrating the Fujitsu-Monaka CPU series and Nvidia GPUs through NVLink Fusion. Yaskawa’s involvement was floated as a possibility even then.

Japan’s demographics explain the urgency. An older report picked up on Nvidia’s own blog cited Japan Times reporting that the nation is expected to face an 11 million shortage of workers by 2040. Fujitsu’s expansion into robotics was framed around exactly that gap, and the same pressure now shows up in Kawasaki’s pitch for its new Silicon Valley center.

The wider industry backdrop is consolidating fast, too. ABB Group announced plans to spin off its ABB Robotics division as a separate entity sold to SoftBank for nearly $5.4 billion, around the same month Fujitsu and Nvidia first went public with their plans. Fujitsu’s ambitions do not stop at robots, either: it also flagged an expansion into quantum computing, a corner of the market that just watched a $1.68 billion Nasdaq debut open hot and close flat.

The next frontier of AI is in the physical world, and this is a once-in-a-generation opportunity for Japan.

Jensen Huang, founder and chief executive of Nvidia, said that in the July 15 announcement. It is a bigger claim than anything Fujitsu, Fanuc, Yaskawa or Kawasaki have committed to on paper so far.

Four Companies, Four Different Bets

Strip away the joint press language and the four Japanese partners are not doing the same thing. Each is running its own project under the same Nvidia umbrella.

Kawasaki’s piece of this is the most concrete so far. Under its collaboration agreements, Nvidia works with Kawasaki on AI and robotics integration, Analog Devices contributes sensing and voice-recognition technology, Microsoft supports cloud and AI infrastructure, and Fujitsu focuses on integrating AI with healthcare business and robotic systems, according to a Kawasaki newsroom announcement of its new San Jose center. The company plans to combine physical AI with existing systems including the Nyokkey service robot, the Forro delivery robot, the hinotori surgical robot and the Corleo multi-legged vehicle. Yasuhiko Hashimoto, Kawasaki’s president and chief executive, put it plainly at the center’s opening: “This speed is exactly why Silicon Valley matters to us.”

Are Fanuc and Yaskawa Also Betting on Google?

Yes. Fanuc struck a separate deal with Google weeks before the Cosmos Coalition news, covering more than a million of its own installed robots, and that announcement sent Fanuc’s shares to a record on its own. Nvidia does not have Japan’s largest robot maker exclusively.

Fanuc partnered with Google to deliver smarter, more adaptive robots for manufacturers, the companies said May 19, integrating Gemini Enterprise and the Intrinsic robotics platform into its 1.1 million installed industrial robots, sending Fanuc shares to a record. Days later, when Kawasaki unveiled its Nvidia-linked Silicon Valley center, shares of the 130-year-old industrial group rose as much as 12 percent, their biggest move since February 9, while Fanuc gained 8 percent and Yaskawa Electric added 5.9 percent on the read-through.

Two chip-and-software ecosystems are now both leaning on the same handful of Japanese hardware makers. Nothing here is exclusive, and the robot builders appear happy to keep both options open until one platform clearly pulls ahead.

The Market Can’t Agree on How Big This Is

Ask three research firms how large the physical AI market actually is and you get three very different answers, all published within the same two-month window.

• MarketsAndMarkets – puts the market at $1.50 billion in 2026, reaching $15.24 billion by 2032 at a 47.2 percent annual growth rate.
• Mordor Intelligence – starts from a higher base, $7.11 billion in 2026, climbing to $34.89 billion by 2031.
• Grand View Research – uses the broadest scope of the three, projecting the market to reach $960.4 billion by 2033.

The gap comes down to definitions. One widely cited estimate has the physical AI industry reaching $15.24 billion by 2032 from $1.50 billion in 2026, according to MarketsAndMarkets research on Nvidia’s robotics push. Mordor Intelligence instead puts the market at $7.11 billion in 2026, growing to $34.89 billion by 2031, in its physical AI market report. Grand View Research folds in autonomous vehicles, wearables and smart infrastructure, which explains why its forecast reaches $960.4 billion by 2033, with North America holding a 30.7 percent share as of 2025, per its industry analysis of the physical AI sector.

The Gap Between Coalition Logos and Shop-Floor Robots

Set the announcement against actual shipped hardware and the picture gets smaller fast. Fanuc is the world’s leading maker of industrial robots, with shipments surpassing 1 million units. Against that base, Fanuc has shipped more than 1,000 robots for physical AI-related applications since releasing its systems at the International Robot Exhibition in Tokyo in December. That is a rounding error next to the installed fleet, not a platform shift, at least not yet.

Regulation has not caught up either. Safety compliance is a major restraint because autonomous systems must fit within standards not designed for machines that perceive, reason and act in changing environments, with ANSI/RIA R15.06, IEC 61508 and ISO 10218 forming an overlapping, fragmented compliance path rather than one framework. Liability is just as unsettled, since legal exposure can sit with the robot maker, the model provider, the integrator and the operator simultaneously, and that uncertainty tends to favor larger firms that can absorb insurance and legal overhead. Nvidia’s own answer, a safety framework called Halos for Robotics, only arrived in June.

Even inside Nvidia, the skepticism has surfaced publicly. Jim Fan, head of Nvidia’s robotics business and co-head of its GEAR lab, posted on social platform X criticizing the current state of the robotics industry, according to Digitimes. A Tokyo-based research note from the Institute of Geoeconomics reached a similar conclusion from the outside: in standardized manufacturing environments, high-speed, high-precision repetition remains essential, so Japanese manufacturers may prioritize simpler wins like natural-language robot programming rather than fully autonomous systems. And competition for the compute layer itself is intensifying, since companies including AMD, Intel, Qualcomm, Tesla and Arm are aggressively investing in rival robotics AI platforms.

Fujitsu’s platform is, in Nvidia’s own phrasing, still something the company is exploring. The next scheduled proof point is GR00T N2, a next-generation robot foundation model slated to be available by the end of the year. A more immediate one arrived already: Kawasaki Robotics and Dexterity expanded their collaboration at Automate 2026, pairing Kawasaki’s RL030N, described as the industry’s first eight-degree-of-freedom robot arm built specifically for physical AI, with Dexterity’s Mech humanoid for warehouse logistics.

Frequently Asked Questions

What does Nvidia mean by physical AI?

Physical AI refers to systems that operate outside a screen, in a factory, hospital or warehouse. Kawasaki defines physical AI as artificial intelligence capable of autonomously perceiving, reasoning and acting within real-world environments through machines and robotic systems, which is the working definition behind Wednesday’s announcement too.

Has Fujitsu’s control platform actually launched?

No. Nvidia’s own release describes Fujitsu as exploring the platform’s development with Fanuc, Yaskawa Electric and Kawasaki Heavy Industries, language that stops short of a committed product or shipping date. No launch timeline has been given publicly.

How many companies are in Nvidia’s Cosmos Coalition?

At least ten Japanese organizations intend to join, spanning far beyond robotics. The coalition is meant to help advance open frontier physical AI models across factories, logistics, agriculture, construction, healthcare and smart spaces, not just industrial robot arms.

Why is Kawasaki starting with hospitals instead of factories?

Labor shortages in elder care are more politically urgent in Japan right now than factory automation. Hashimoto said the company will first focus on healthcare and elder care, where aging societies and labor shortages are global challenges, before the center expands into semiconductors, mobility and manufacturing.

What comes next for the Nvidia-Japan alliance?

Two dated milestones stand out. Nvidia’s GR00T N2 model is due by year end, and Kawasaki’s RL030N arm, paired with Dexterity’s Mech humanoid, is already being tested for warehouse logistics work under the Automate 2026 expansion.