FANUC and Google’s AI welding robot reported by The Nikkei

The Nikkei reported on September 11, 2026, on an AI welding robot from FANUC and Google. The headline says it “reads paper drawings,” making the mechanism for reading paper drawings a key focus.

An AI welding robot that reads drawings could reduce reliance on conventional teaching and affect setup processes on factory floors. Because the information comes from a Nikkei report, the facts need to be verified. Still, the involvement of FANUC and Google points to a connection between robot control and AI infrastructure.

What is new in this report is the combination of FANUC and Google and the introduction of drawing-reading capabilities to a welding robot. However, the report’s headline in The Nikkei offers only limited insight into what has been automated and which processes are covered. For now, it suggests a direction in which a robot builds its tasks from drawings, but the scope of implementation needs to be confirmed in the original report. Related coverage by this newspaper shows that FANUC announced the launch of its “5-Axis Integrated Initiative” to support productivity improvements in 5-axis machining on August 26, 2026, as part of efforts to improve productivity in existing manufacturing operations. By contrast, the AI welding robot appears to focus on generating tasks based on drawing interpretation rather than optimizing machining conditions. In addition, in FANUC publishes a basic guide to industrial robots on August 5, 2025, the company identified welding as a representative application for industrial robots. The latest report suggests a possible step beyond describing that application in general terms, toward linking design drawings to actual work. Another interpretation is that a software-focused competitive axis—combining drawings, recognition and control—is coming to the fore, rather than competition focused solely on robot performance. At the industry level, welding is a representative example of work that is difficult to standardize. If drawing interpretation becomes practical, it could lower barriers to robot adoption in high-mix, low-volume operations as well. On the other hand, it is unclear which drawing formats the system will support, how safety will be ensured in the event of misinterpretation, and how much of the existing teaching process it can replace. For FANUC, the focus will be whether this is a step toward integrated offerings that include generative AI and recognition infrastructure, beyond hardware-centered competition. Google, too, will face questions about how to connect cloud services and AI models to manufacturing sites. Points to watch include: 1) whether drawing recognition expands to formats other than paper, 2) whether the technology is applied to processes beyond welding, and 3) how safety design and the allocation of responsibility will be defined in cases of misrecognition. The original source should be consulted for details.

Japan has room to expand the use of welding robots, but drawing interpretation and adaptation to site conditions can be bottlenecks. If the structure indicated by this report is accurate, the roles of industrial robot makers, control software providers and domestic companies with expertise in handling drawing data may come under scrutiny.