Is FANUC a Humanoid Robot Stock? Its Order Surge Says No

When a humanoid robot demo goes viral, FANUC trades like a humanoid stock. I don't think it is one. FANUC's order surge this year was led by CNC machine-tool controls, not robots, and part of that CNC demand is machining parts for other companies' humanoids. The same goes for most of the ten names the robotics trade bundles with Tesla's Optimus: each one already makes money from something specific, and it isn't a walking robot.

FANUC's first-quarter orders (the three months to June) rose 36.9% to ¥281.9 billion. FA orders, the CNC and servo business, grew fastest, up 63.8%. Robot orders rose 8.1% and slipped from the previous quarter. On the July 31 call FANUC said the FA surge came from chip equipment for data centers and AI servers, electronics, electric vehicles and machined components for humanoid robots. So the humanoid shows up in FANUC's report as someone else's parts list, cut on a FANUC controller.

Yaskawa is the best test of whether FANUC's quarter was a robot cycle. Its Motoman arm is one of the four global industrial-robot standards. If the humanoid trade were right, Yaskawa's robot line should have jumped. It didn't: Robotics sales rose 2.0% and Robotics operating profit fell 82.3%, hit by a new ERP system that shut production lines and by restructuring in Europe. What jumped was servo motors, with orders up 65%, mostly into chip equipment.

ABB is even clearer. It agreed in October 2025 to sell its robotics division to SoftBank for an enterprise value of $5.375 billion, and already reports it as a discontinued business. What's left is a grid, drives and process-automation company whose Q2 orders were led by electrification. A fund that holds ABB for robots is holding a business ABB is selling.

The industrial robot that ships today isn't a general-purpose worker. It's a device qualified into a specific cell. Aerospace machining standards name FANUC CNC models. A welding cell is approved for a model and a controller. A hospital trains its surgeons on a da Vinci console. A pharma line is validated to a Rockwell Logix version, and a Keyence sensor is written into its inspection steps. Replacing any of those means qualifying the whole cell again, which is why the humanoid pitch is really a promise to skip the cell altogether. That's a different bet, and it shouldn't be priced into companies whose money comes from the cell.

The other side isn't a cartoon. A humanoid that walks onto a line without anyone rewriting its programs would make FANUC's 40 years of machinist habits a stranded asset. And the threat doesn't need a humanoid at all: cheaper Chinese arms can win on price, and FANUC's own Q&A said some China customers bought a competitor's robot when FANUC's lead times stretched. Yaskawa is where that would show first, since its arms compete in cells that never had FANUC's CNC lock. If Yaskawa's robot sales pick up while FANUC's robot orders stay flat, FANUC is being walked around by a cheaper arm, not by Optimus.

Harmonic Drive is the honest way to own the humanoid upside. A humanoid needs dozens of precision gears where an industrial arm needs a handful, and the company already reports orders from customers moving toward mass production. But its own filing still calls the AI-robot market nascent, and most of its sales are still industrial robots and chip equipment. It can be right about being in the joint of whoever wins and still be early.

Two more names could prove the label partly right. If Teradyne's Universal Robots grows from $100 million a quarter into a real share of the company, then human-scale cobots were the robot market all along. And Tesla's Construction status isn't a refusal: it has taken out car lines to make room for Optimus. The test is revenue from outside customers, not a factory tour.

My conclusion is narrower than "avoid robotics." Own the companies whose products are already qualified into a factory, an operating room, a plant's control system or a grid, and judge them on reports about that business. Don't pay a humanoid price for a machine-tool recovery, a servo cycle, a chip-test company or a robot unit that's being sold.