The Robotics Watchlist: How a Tweet Became a Screen, Not a Trade
Years ago, a tweet from Leandro made me curious about semiconductors. It was not a stock tip, and it did not immediately turn into an investment. It simply pushed me into a rabbit hole. I started reading about the semiconductor value chain, trying to understand who mattered, where the bottlenecks were, and which companies could quietly benefit from a trend much bigger than any single product cycle.
That work eventually led me to ASML. Now my second biggest position.
The important word is “eventually”.
Good investing ideas often do not arrive fully formed. They usually begin as curiosity, then become a map, then become a watchlist, and only sometimes become an investment. The map is often more valuable than the original idea.
Recently, something similar happened again. Leandro, who writes at Best Anchor Stocks and posts on X as @Invesquotes, published a deep dive on the humanoid robot supply chain.
The post framed the humanoid robot like a human body: brains, senses, nervous system, muscles and bones, training ground, and immune system. That framing immediately made sense to me because it moved the question away from “which robot company will win?” and toward a better question: “which companies sell the critical parts that many winners will need?”.
That is usually where I prefer to look. In a gold rush, I am often more interested in the boring supplier than the miner. Not because suppliers always win, but because they can sometimes participate in the growth of a whole ecosystem without needing to perfectly predict the final winner.
With semiconductors, that way of thinking led me toward ASML. So the natural question was: who are the ASMLs of robotics?
The honest answer is probably: nobody. At least not in the same way. ASML is a very unusual company. EUV lithography is an extreme bottleneck with decades of accumulated engineering, supplier depth, customer dependence, and near-monopoly economics.
Robotics looks more distributed. It has many bottlenecks: AI models, simulation, sensing, power, batteries, actuators, safety, testing, cybersecurity, manufacturing, and real-world deployment.
But even if there is no single ASML of robotics, the question is still useful. It forces you to look below the surface of the story.
My first instinct was that the best layer was not the robot maker. It was the senses and nervous system. Robots need to perceive the world and then act on it. They need cameras, sensors, analog chips, signal processing, power management, edge processors, secure connectivity, safety systems, and test equipment. They need to convert light, heat, pressure, torque, motion, distance, current, and position into usable information.
That sounded more interesting to me than trying to guess which humanoid robot brand becomes famous.
So I made a list of companies to study: Keyence, Analog Devices, Texas Instruments, Infineon, NXP Semiconductors, and Teradyne.
Each one had a different role in the robotics stack. Keyence looked like the highest-quality sensing and machine vision company. Analog Devices looked like one of the cleanest “nervous system” companies because of its analog, mixed-signal, sensing and power exposure. Texas Instruments looked like the boring but durable analog and embedded processing tollbooth. Infineon looked like the European contrarian candidate in power semiconductors, automotive, industrial chips, sensors, microcontrollers, and security. NXP looked like a physical-AI edge company through automotive processors, radar, secure connectivity, industrial IoT, and embedded intelligence. Teradyne looked like the most direct robotics name because it combines semiconductor test with Universal Robots and MiR.
On paper, all of this was exciting.
In the spreadsheet, it was less exciting.
And that is exactly why I use the spreadsheet.
My screening system is not there to make decisions for me. It is there to slow me down. It looks at market cap, EBITDA, potential valuation, cap/EBITDA, ROIC or ROCE, the latest year and five-year average, a qualitative hypothesis score, the distance from highs and lows, and finally an action band: Strong Buy, Add Small, Fair Price, Too Expensive / Wait, or Too Expensive / Trim.
This system is not perfect. One of the notes in my own spreadsheet says: “More fiction has been created using Excel than Word.” I believe that. A spreadsheet can make a bad idea look precise. But it can also protect you from your own enthusiasm, which is one of the most useful things a system can do.
After adding the robotics companies to my CCG Stocks screen, the result was clear: none of them was ready for entry.
Keyence was probably the best “senses” quality candidate, with excellent exposure to factory automation, sensors, measurement and machine vision. But it was expensive, and for me there is added friction: Japan, yen, language, dividends, market access, and the fact that the Japanese market has already rerated strongly.
Analog Devices had one of the strongest theses. It sits exactly where I think robots will need help: converting the physical world into electrical and digital signals. But the valuation was rich, and the ROIC numbers I found were weaker than I expected.
Texas Instruments was almost too boring, which I usually like. Analog chips, embedded processors, long product lives, industrial and automotive exposure. The ROIC numbers were better than ADI, but the stock was still not cheap enough.
Infineon was the European contrarian in the group: power semiconductors, AI data-center power, automotive, industrial, sensors, MCUs and security. But the ROIC was weak compared with the others, and the stock had already moved a lot.
NXP was probably the closest to fair value among the six. It has good exposure to software-defined vehicles, radar, secure connectivity, industrial IoT and physical AI at the edge. But close to fair value is not the same as attractive. There was not enough margin of safety.
Teradyne was the most interesting and probably the most dangerous. It had excellent ROIC, strong AI semiconductor-test exposure, and direct robotics optionality through Universal Robots and MiR. But the stock had already exploded. The valuation looked like the story had arrived before me.
So the result of the whole exercise was not a purchase.
It was a watchlist.
That may sound disappointing, but I think it is the opposite. The job of an investor is not to find exciting stories. There are plenty of exciting stories. The job is to find situations where the story, the numbers, the quality of the business, and the price all line up well enough to take risk.
Robotics may become one of the defining themes of the next decade. Physical AI may be real. Humanoid robots may eventually move from demos to factories, warehouses, hospitals, roads, farms and homes. The world may need many more sensors, analog chips, power semiconductors, edge processors, safety systems and testing infrastructure.
All of that can be true, and the stocks can still be too expensive.
This is the part investors forget when a theme becomes exciting. The internet was real in 1999, but many internet stocks were terrible investments. AI is real, but some AI stocks will still disappoint investors. Robotics may be real, but I do not want to pay any price for the story.
For me, this is the quiet edge. It is not having the loudest opinion, the biggest prediction, or the most impressive stock pitch. It is doing the work, building the map, putting the names into a system, and then being willing to do nothing.
Doing nothing is underrated, especially when the theme is exciting and everyone else seems to be making money.
I may be late to robotics. Probably very late to the first wave. But being late to the first wave does not mean the game is over. EBITDA will change. ROIC will change. Management execution will change. Stock prices will change. Narratives will overheat and cool down. Some companies will disappoint. Some will compound quietly. Some may fall 40% and suddenly become interesting. Some may never become cheap enough.
That is fine. The point is that now they are in the system.
I know where Keyence, ADI, TXN, Infineon, NXP and Teradyne sit in the robotics stack. I know what I like about each one. I know what worries me. I know what valuation would start to interest me. That is useful, even if I never buy any of them.
With ASML, the semiconductor rabbit hole eventually led to an investment. With robotics, the rabbit hole has so far led to patience.
That is not failure. That is the process working.
The conclusion for now is simple: I may be late to robotics, so I built a watchlist instead.
And sometimes the best trade is not buying.
July 2026