Hitachi and FANUC pair edge AI chips with industrial robots, aiming for a global physical AI rollout from fiscal 2027
Hitachi and FANUC signed a strategic partnership on Sept. 30 to commercialize physical AI jointly, combining Hitachi's HMAX Industry software and in-house edge AI semiconductor with FANUC robots. Trials start in Hitachi's Ibaraki-area plants as “Customer Zero,” with global sales planned from fiscal 2027. No financial terms were disclosed.
Hitachi Ltd. and FANUC Corp. have signed a strategic partnership to build and sell physical AI systems together, pairing Hitachi's AI software and an edge AI semiconductor it developed in house with FANUC's industrial robots, the two companies said in a joint release on Sept. 30, 2026. They plan to prove the technology inside Hitachi's own plants in the Ibaraki area, which they have designated as “Customer Zero,” and to begin offering it to customers worldwide in fiscal 2027, the year that starts in April 2027.
The release did not disclose an investment amount, an equity component or a revenue target. It did set out the scale of the two partners. Hitachi reported revenue of ¥10.59 trillion for the fiscal year ended March 2026, with 606 consolidated subsidiaries and about 290,000 employees, while FANUC reported ¥857.8 billion in revenue and 10,040 employees, according to the company profiles attached to the announcement.
What was agreed
Hitachi contributes its HMAX Industry AI portfolio, the edge AI semiconductor, AI that can be retrained in a short cycle, and the operational technology know-how it has built running factories for more than 110 years, according to the release. FANUC brings its robots, its high-precision control technology, a large installed base and what it describes as an open platform for advanced robot control. The trials in Ibaraki will cover picking parts that vary in shape and handling production changeovers, and will measure recognition accuracy, motion control, cycle time and quality. The partners will also verify Hitachi's edge AI chip running with FANUC robots.
The timetable has two steps. The companies aim to establish practical physical AI technology through the joint verification by the end of fiscal 2027, and to start global deployment to both companies' customers from fiscal 2027. The release lists semiconductors, pharmaceuticals, healthcare, advanced materials, automotive, logistics, food, shipbuilding and agriculture as target industries.
Hitachi President and CEO Toshiaki Tokunaga said in the release that the tie-up joins FANUC's robot technology with Hitachi's manufacturing know-how and digital capabilities, and that Hitachi would support customers from installation through operation to tackle labour shortages and productivity. FANUC President and CEO Kenji Yamaguchi said AI now allows machines to recognise and control real-world environments, and that using Hitachi's factories as Customer Zero gives the partners a practical development model for making whole manufacturing processes autonomous.
Two companies with momentum
FANUC reported consolidated net sales of ¥231,035 million for the April to June 2026 quarter, up 17.7% from a year earlier, and operating income of ¥53,492 million, up 26.1%, according to its quarterly earnings filing dated July 31. Robot division sales rose 18.7% to ¥96,103 million. The filing said domestic automotive demand was weak while the Americas and China were strong, with Chinese demand driven by electric vehicle projects. FANUC raised its full-year forecast to net sales of ¥948,100 million and operating income of ¥218,000 million. Hitachi posted first-quarter revenue of ¥2,709,586 million, up 20.0%, and adjusted EBITA of ¥323,529 million, up 36.2%, and forecasts ¥11.7 trillion in revenue for the full year, according to its July 29 filing.
The Hitachi deal is the latest in a run of FANUC AI alliances. In May, FANUC announced a collaboration with Google covering Gemini Enterprise and Google's Intrinsic robotics software platform, and said it had shipped more than 1,000 robots for physical AI applications since unveiling its physical AI system at the International Robot Exhibition in December 2025. On July 16 it began business talks with Fujitsu on physical AI built with NVIDIA technology, in parallel with Yaskawa Electric and Kawasaki Heavy Industries. On July 24 it announced an investment in Noetra, a startup aiming to develop a Japanese multimodal foundation model, and on Sept. 11 it unveiled an AI Welding Agent running on Google's Gemini Enterprise. Hitachi, for its part, expanded its HMAX lineup on Sept. 3 with offerings including HMAX AI Operations, a system for monitoring and operating physical AI as it is updated.
Rivals and the numbers
Domestic rivals are moving on parallel tracks. Yaskawa Electric said on July 15 it had built an agentic robot system that combines its MOTOMAN NEXT robot with Google DeepMind's Gemini Robotics ER 1.6 model, and on July 13 reported a demonstration with SoftBank in which a vision-language-action model handled wire harnesses. Kawasaki Heavy Industries opened the Kawasaki Physical AI Center San Jose in May 2026, it said in its July 16 release. The table compares the most recent reported quarter for the main Japanese players in these alliances.
- FANUC
- Apr to Jun 2026
- 231,035
- 17.7
- 96,103
- 41.6
- 948,100
- Hitachi
- Apr to Jun 2026
- 2,709,586
- 20
- No data
- No data
- 11,700,000
- Yaskawa Electric
- Mar to May 2026
- 138,982
- 10.6
- 56,728
- 40.8
- 580,000
- Kawasaki Heavy Industries
- Apr to Jun 2026
- 543,576
- 11.3
- 69,594
- 12.8
- 2,560,000
- Fujitsu
- Apr to Jun 2026
- 779,400
- 3.9
- No data
- No data
- 3,510,000
Figures as disclosed in each company's quarterly earnings filing. Robot segment = FANUC robot division, Yaskawa robotics segment, Kawasaki precision machinery and robot segment (external sales). Robot share = robot segment revenue / consolidated revenue. Hitachi and Fujitsu report no robot segment. Yaskawa's fiscal year ends in February.
As of Oct 1, 2026
Physical AI, as the companies use the term, describes systems that perceive their surroundings, judge what to do and act, rather than repeat a path a technician has taught. Running inference on an edge chip next to the robot keeps response times short and keeps production data inside the plant. The short-cycle learning Hitachi describes is aimed at exactly the jobs named in the release: picking mixed parts and switching a line to a new product are tasks where setups change too often for conventional teaching to pay off.
For the global market the deal shows FANUC assembling several AI suppliers at once rather than tying itself to one. Hitachi offers something Google and NVIDIA do not, namely its own factories as a test site and a long list of industrial customers who already buy its operational technology. For overseas robot makers, the combination of a robot supplier and a conglomerate that is also its first customer is a template that is hard to copy without a comparable in-house manufacturing base.
The release leaves important questions open. It does not name the edge AI chip or give its performance, nor does it say how the companies will split revenue or who owns the resulting models and data. Verification runs to the end of fiscal 2027, so the joint offering is unlikely to contribute materially to either company's results before then. FANUC's overlapping alliances with Google, Fujitsu and NVIDIA, and now Hitachi, also raise the question of which software stack its customers will be steered toward.
The next markers are FANUC's results for the July to September quarter, the scheduled start of AI Welding Agent shipments at the end of December 2026, any published data from the Ibaraki trials, and the start of fiscal 2027 on April 1, 2027, when the companies say global deployment begins.
ROBOTNESS analysis
FANUC is positioning itself as the neutral robot layer for Japanese physical AI, and Hitachi's factories give this alliance a proving ground that software partners alone cannot provide.
The evidence is the pace of FANUC's deal-making. Between May and September 2026 it announced the Google collaboration, talks with Fujitsu on NVIDIA-based physical AI, an investment in Noetra, the Gemini-based welding agent and now the Hitachi partnership. Its robot sales grew 18.7% in the April to June quarter and it says more than 1,000 robots have shipped for physical AI applications, so the alliances sit on top of real demand rather than replace it.
The strongest counter-argument is that none of these agreements carries disclosed money, and a partnership without capital or revenue commitments is easy to sign and easy to let fade. Japanese industry has a long record of consortia that produced demonstrations rather than products.
Bull case: the Ibaraki trials show measurable cycle time and changeover gains, Hitachi bundles FANUC robots into its HMAX deals across its customer base, and FANUC gains a second sales channel into pharmaceuticals, semiconductors and food, where it has been less dominant than in automotive. That would lift robot division growth above the current high-teens rate.
Bear case: the edge chip proves less capable than GPU-based systems from NVIDIA partners, customers hesitate between FANUC's competing AI stacks, and the joint offering stays a showcase until after fiscal 2027. In that scenario the deal adds marketing value but little revenue.
Signals to watch:
- End of December 2026: whether FANUC starts AI Welding Agent shipments on schedule, a test of how fast it converts AI alliances into products.
- April 1, 2027: start of fiscal 2027, when Hitachi and FANUC say global deployment begins; look for named customers.
- March 31, 2028: end of fiscal 2027, the deadline the partners set for establishing practical physical AI technology through the Ibaraki verification.
- Hitachi Ltd. and FANUC Corp.
- Sept. 30, 2026 (joint release)
- Hitachi factories in the Ibaraki area as “Customer Zero”
- From fiscal 2027 (starting April 2027)
- Picking parts of varying shapes, production changeovers
- Not disclosed
- Semiconductors, pharma, healthcare, materials, auto, logistics, food, shipbuilding, agriculture
- ¥96,103 million, up 18.7% (Apr to Jun 2026)
- HMAX Industry, in-house edge AI semiconductor, short-cycle learning AI
Why it matters
The partnership pairs Japan's largest industrial robot maker with a conglomerate that is both a technology supplier and a large factory operator. That combination matters because the bottleneck in physical AI is no longer model research but repeatable deployment on real lines with real cycle-time targets. Hitachi can supply the plants, the process data and the operational technology layer; FANUC supplies the hardware that already sits in those plants.
It also signals that FANUC, historically protective of its closed controller ecosystem, is now building outward. Its May statement that robots support ROS through open-source drivers and the Intrinsic platform, combined with the Hitachi deal, shows a company trying to stay the default arm regardless of which AI layer wins.
Rival analysis
Yaskawa Electric has chosen a similar multi-partner route, with Google DeepMind's Gemini Robotics ER 1.6 integrated into MOTOMAN NEXT and a VLA demonstration with SoftBank. Its robotics segment, however, grew only 2.0% in the March to May quarter and its segment operating profit fell 82.3% to ¥888 million, which leaves less financial room than FANUC has. Kawasaki Heavy Industries is investing through its San Jose physical AI centre and posted 22.2% growth in precision machinery and robot revenue, helped by semiconductor robots.
Outside Japan, NVIDIA is the common platform provider to several of these makers through the Fujitsu initiative, which reduces differentiation at the AI layer and pushes competition back toward integration skill and installed base, where FANUC is strongest.
Valuation context
No price was attached to the deal, so valuation effects run through earnings expectations. FANUC's revised forecast implies full-year net sales of ¥948,100 million, 10.5% above the ¥857.8 billion stated for the prior year, and operating income of ¥218,000 million. Its robot division already accounts for 41.6% of first-quarter net sales.
For Hitachi, the robot partnership is small relative to ¥11.7 trillion of forecast revenue. Its value lies in giving HMAX Industry a hardware anchor that can be sold into existing customer accounts.
Supply-chain implications
The edge AI semiconductor is the least transparent element. Hitachi has not named the device, its process node or its foundry, so it is not yet possible to judge cost or supply risk. If the chip ships inside FANUC cells, Hitachi becomes a component supplier to FANUC, a reversal of the usual flow in which robot makers buy compute from NVIDIA or Intel.
On the robot side, the target tasks of mixed-part picking and changeovers rely on vision, force sensing and grippers rather than new arm designs, so near-term demand effects fall on sensors and end-effectors more than on reducers and servo motors.
Signals to watch
First, whether Hitachi or FANUC publishes quantified results from Ibaraki, such as changeover time or pick success rates, before the end of fiscal 2026. Second, whether the edge AI chip receives a product name and specifications. Third, whether named external customers appear at the start of fiscal 2027.
Also watch whether FANUC clarifies how the Hitachi stack coexists with its Google and NVIDIA-based offerings, since overlapping alliances can confuse buyers.
Analyst view
Thesis: the deal strengthens FANUC's position as the default robot platform in Japanese physical AI, but its financial impact will not be visible before fiscal 2027. Confidence: medium.
Reasons for medium rather than high confidence: the partners disclosed scope, sites and timelines, which is more concrete than many memoranda, and both companies are growing. Against that, no money, chip specifications or customer names have been disclosed, and the verification window is long.
Questions you should be asking
What are the specifications of Hitachi's edge AI semiconductor and who manufactures it? How will revenue and intellectual property from jointly developed models be divided? Will Hitachi buy FANUC robots for its own plants at scale beyond the trials?
How will FANUC position the Hitachi stack against its Google Gemini and NVIDIA-based offerings when a customer asks which to choose?