The AI trade is starting to move one layer down the stack. Chips and data-center power are still the headline winners, but traders are increasingly looking for the second-derivative names that can turn software ambition into physical throughput. That is why the newest signals from Korea, Japan, Europe, and the United States matter together rather than separately.
The immediate catalyst came from South Korea. On June 26, 2026, Reuters reported via Yahoo Finance that Samsung is preparing a reported 1,000 trillion won, or roughly $648 billion, long-cycle investment plan spanning AI data centers, semiconductors, batteries, displays, and robotics. Even if the final spend path changes, the market message is clear: Korea’s national champions are framing AI as a full industrial rebuild, not just a memory-chip boom.
Japan matters because it sits at the intersection of robotics know-how and factory deployment. FANUC said on May 13 that it is collaborating with Google on physical AI so a Google-powered AI agent can operate robots, and on March 24 it said it would invest $90 million in a new factory and distribution center in the United States with production-ready capacity for robot manufacturing. That combination is important. It says the robotics trade is no longer just about lab demos. It is about capacity, localization, and how fast robot makers can convert AI headlines into installed systems.
Europe’s signal is more technical but still market-relevant. ABB said in June that it is working with PSYONIC to use human-generated data to improve robotic grasping and dexterity, which goes straight to one of the hardest bottlenecks in physical AI. In other words, the next stage is not merely building more robots. It is making them useful in less scripted environments, where the productivity promise becomes real enough to support valuation.
The U.S. angle broadens the trade beyond robot arms themselves. Teradyne said on June 8 that it is introducing an integrated test solution for AI and data-center devices in collaboration with Tokyo Electron. That matters because once AI hardware cycles become more complex, the metrology and test layers often gain pricing power before the end-market volume fully shows up. Traders who only screen semiconductors can miss that equipment and validation names are often where the margin quality hides.
My view is that physical AI is becoming a better market frame than narrow “AI infrastructure” labels. The opportunity set is widening from chips and turbines into robots, machine vision, industrial software, and test systems. But this is also where hype can outrun orders. For now, the most credible names are the ones pairing AI language with visible factory spend, deployment partnerships, or hard engineering milestones.
The risk is that this theme gets overextended before revenues catch up. National-investment headlines can arrive faster than purchase orders, dexterity breakthroughs can stay stuck in pilot programs, and automation customers can delay upgrades if industrial demand softens. If that happens, physical-AI beneficiaries may trade like concept stocks rather than durable cash-flow stories.
Sources: Reuters via Yahoo Finance on Samsung’s reported AI investment push | FANUC on its physical AI collaboration with Google | FANUC on its new U.S. robot factory and distribution center | ABB on robotic dexterity work with PSYONIC | Teradyne on its AI-device test collaboration with Tokyo Electron
Risk notice: This article is market commentary for informational purposes only. It is not personalized investment advice or a recommendation to buy or sell any asset.
原创文章，作者：financial transaction，如若转载，请注明出处：https://www.fanbi.net/archives/468
