Three Industries Just Collapsed Into One Battlefield
Three Industries Just Collapsed Into One Battlefield
Jean-Marc Chéry has been in the semiconductor business for 42 years. At TSMC's European Symposium in Amsterdam last week, the STMicroelectronics CEO said the next five years will be the most disruptive he's ever seen — and he wasn't talking about a product cycle. He was talking about a competitive landscape collapse.
The mechanism is architectural. Electric vehicles, industrial robots, and humanoid machines have converged on the same computing architecture: central processing with zonal control, tying sensors, drivers, switchers, and controllers into a single unified system. Until recently, these were separate industries with separate supplier ecosystems, separate engineering cultures, and separate moats. They're not separate anymore.
"Architectures for EVs, industrial robots, and humanoid robots have a reasonable amount of crossover," Chéry said. "All use central computing with zonal control." What this means is that a chipmaker who can solve the system-level engineering problem for one of these verticals can, for the first time, sell the same underlying platform into all three. The moats weren't lowered — they vanished.
ST has already reorganized around this reality. The company has moved from a traditional automotive power, analog, and MCU vendor to segment marketing organized by vertical — automotive, industrial, and robotics — with a systems-engineering perspective baked in. Instead of quoting individual component specs, ST is now telling customers which products they need to hit a specific robot grip strength or power band. That's a different sales model. That's a different company.
Chéry called it "a major transformation" and "the most passionate period I ever faced in my 42-year career in Silicon Valley." He also said plainly: "What will happen in the next five years is clearly a competitive landscape displacement."
He's not alone in recognizing the shift. Cisco president Jeetu Patel, also speaking at the symposium, quantified what agentic AI means for infrastructure: AI agents consume 450% more bandwidth than a human doing the same task via chatbot, and they run 24/7 at machine scale. "No matter how much infrastructure you feel that will be needed in the world," Patel said, "you are underestimating the capacity that's going to be needed." The infrastructure buildout to support physical AI is not a future problem. It's a present constraint.
The numbers behind the shift are large. TSMC expects the semiconductor market to cross $1 trillion in 2026 and expand to $1.5 trillion by 2030, with HPC and AI accounting for roughly 55% of that. The company is investing $165 billion in US expansion to support the buildout. McKinsey data shows semiconductor economic profit ranking among 31 industries has risen from 12th in the early 2000s to 3rd in the 2020–2024 window — a roughly 40-fold increase in EP since 2000.
ST's own moves confirm the thesis. In March, ST announced a Physical AI integration with NVIDIA, embedding its sensors, microcontrollers, and motor control solutions into the NVIDIA Holoscan robotics ecosystem. The goal: let developers design, train, and deploy humanoid and industrial robots using ST components as part of a validated reference stack rather than a component picking exercise.
The competitive implication cuts both ways. Chipmakers that solve the systems engineering problem become platform players across three of the largest industrial markets in the world. Those that stay component vendors — selling individual MCUs, analog chips, or power discretes without system context — get squeezed between the platform layer above them and the fab layer below. The companies that built moats around automotive-grade reliability or industrial temperature ranges are discovering that the walls between those categories are now porous.
The 5-year timeline Chéry named is a prediction, not a certainty. Infineon, NXP, and Renesas have not publicly announced equivalent reorganizations, and it's possible that specialized requirements for each vertical reassert themselves as the engineering gets harder. But the direction of travel is clear from the architecture, and the first movers in this convergence are already drawing the map.
The question for anyone building in EVs, robotics, or industrial automation is straightforward: are you buying your system platform from a chipmaker who's reorganized to sell outcomes, or from a component vendor who's still quoting datasheets? The answer matters more now than it did last quarter.