Silicon edge, sensors, optics, robotics, memory, connectivity – the entire map of actual AI trading
Listen to the audio version of this article (generated by artificial intelligence).
For the better part of three years, the story of physical AI — the idea of AI moving from the cloud to the physical world, powering robots, wearables, autonomous vehicles, and smart devices — has been like every great technology story: loud, early, and mostly theoretical.
Elon Musk stood on stage and told us that Optimus robots would soon be doing our laundry. Venture capitalists competed to fund the most human-looking object they could find. CNBC ran breathless segments about the robot revolution. The stock market assigned multibillion-dollar valuations to companies whose most impressive product was a press release and a demo reel.
That was then. This is now.
Proof points accumulate
Consider what happened in one quarter:
- Microsoft (MSFT) AI computers with on-device inference chips shipped from Qualcomm (QCom) – Real products, real volumes, real revenues.
- Genesis AI It has launched an industrial robot that can think adaptively, more than just execute programmed sequences.
- Anthem It targets $500 million in sales of wearable AI devices this year.
- Applied materials (huge) in partnership with EssilorLuxottica To manufacture smart optical systems for augmented reality glasses.
- apple (Apple) has confirmed the presence of cameras in the 2027 AirPods, suggesting that physical AI is now a key component of the product roadmap for the world’s most valuable company.
- Mobile i (Wet) announced a concrete deployment of robotaxis in the United States with a plan to expand to 17,000 vehicles.
Six different companies. Six different products. One of the fundamental shifts in what AI needs to operate.
What physical AI actually means – and why the architecture is so different from cloud AI
What makes this course different from the AI wave we are riding is not ambition. It’s architecture.
Cloud-based AI is about scaling – put the computation in the model, let it learn, and provide answers via an application programming interface (API). Physical AI is on the verge efficiency — Get the correct answer, in milliseconds, on a device with a 40W thermal budget, offline.
It’s the AI inside the headphones that filters out background noise before you even notice it…
The vision system on the warehouse robot that decides which bin to pick next…
Self-driving vehicle perception suite that identifies pedestrians at 60 mph.
The requirements are very different, and this difference extends throughout the supply chain.
The six pillars of the supply chain of physical AI
Think of physical AI not as a single industry but as several different classes of devices that all need to scale simultaneously.
1. AI silicon edge
This is the basis. Every physical AI device needs a chip that can run inference locally, fast, cool, and cheap. Qualcomm’s Snapdragon
arm(arm) The infrastructure supports almost every mobile AI chip on the planet. Nvidia (NVDA) is pushing towards embedded inference through its Jetson platform. AMD (AMD) and Intel (Intech) are fighting for their share of the AI personal computer market. The cutting-edge silicon wars have only just begun, and the winners here are getting paid for every device shipped.
Main names: QCOM, ARM, NVDA, AMD, INTEC
2. Sensors and machine vision
Image sensors, depth cameras, radar, lidar, microphones – these are the eyes and ears of every robot, wearable, and autonomous vehicle.
The AMAT-EssilorLuxottica partnership to develop intelligent optical systems for augmented reality glasses tells you all: The optics industry is employing artificial intelligence in the supply chain at the component level. Apple’s AI-powered AirPods with built-in cameras will drive a new demand cycle for miniature sensor modules.
Main names: Ambarella (Amba), On semiconductors (on), ST Microelectronics (STM), Sony (Sony), Cognex (CGNX)
3. Advanced optics
Augmented reality glasses and AI glasses are not just a consumer curiosity anymore, but rather a category of devices. What is the bottleneck? optics.
Waveguides, optical displays, specialized glass, and laser projection systems are what separates a pair of glasses from a head-up display. Corning (GLW) and coherent (COHR) are two of the most underrated physics AI games on the market for precisely this reason. The pivot to applied materials in smart optics manufacturing indicates how seriously the semiconductor equipment industry takes this category.
Main names: amat,glo, Lumentum (Light),COHR
4. Robotics and industrial automation
Genesis AI’s Eno robot isn’t interesting because it’s humanoid — it’s interesting because it has reasons. This is the leap from Artificial Automation 1.0 (programmed movement) to Physical Artificial Intelligence 1.0 (adaptive intelligence).
Companies like symbolic (SYM), Teradine (Third), Rockwell Automation (year), and Honeywell (presence) is already deploying AI-driven automation in factories and warehouses on a large scale. Tesla(TSLA) Optimus is the flashy version. The boring but profitable version is already running in distribution centers across America.
Main names: SIM, TR, ROK, HOON, TESLA
5. Memory, storage and energy
On-device AI needs more local memory than anyone planned. This means low-power dual data rate 6 (LPDDR6) RAM, expanded NAND storage, power management integrated circuits (PMICs) that can handle sequential inference workloads, and analog semiconductors for signal processing.
Micron (in) actually wins here with LPCAMMs for AI computers. Storage plays – Seagate (STX), Western Digital (WDC), SanDisk (Your support) – Get a backend request as each edge device needs to store a local model.
Main names: Mo, stix, dak, sank, Homogeneous power (MPWR), Analog devices (Addy), Texas Instruments (Texan).
6. Communication and infrastructure
Even AI at the edge needs the cloud. Local inference handles latency-sensitive tasks; Cloud AI handles the heavy lifting — model updates, data synchronization, fleet coordination for robo-taxis, and telemetry from billions of wearable devices.
This means that the optical networking and communication layer is the direct beneficiary of scaling physical AI. Sync Robotaxis with the cloud. Augmented reality glasses streaming map data. Industrial robots communicate with the home through diagnostic telemetry Broadcom (Afgo), Marvel (MRVL), Arista (Anita), respect (hundred), doctrine (CRDO), and Corning are all toll roads on this data highway.
Main names: AVGO, MRVL, ANET, CRDO, CIEN, GLW
Investor’s Guide: Own the picks and scoops of the biggest hardware cycle since the smartphone
No one made more money in the California Gold Rush by panning for gold. The real riches went to the people selling the equipment.
Physical AI follows the same logic, with one important difference.
In Gold Rush, you can only sell one frying pan at a time. In physical AI, every device that ships—every robot, wearable computer, AI PC, and self-driving vehicle—needs chips, sensors, optics, memory, power management, and connectivity. Suppliers do not need to choose the winning application. They get paid per unit, across every category, regardless of the corporate robot that ends up in your warehouse or the augmented reality glasses that end up on your face.
The move from cloud AI to physical AI is the biggest single hardware cycle since the smartphone. And like smartphones, the companies that win are not just the device manufacturers, but the entire supply chain that follows them.
The hype was right. It only took hardware a few years to catch up.
The names in this article — peripheral silicon suppliers, sensor makers, optics companies, memory and connectivity companies — are the general market expression of this thesis. But the smartest money isn’t just jumping into the obvious deals.
Take Peter Thiel’s latest 13F: zero shares in Nvidia, Apple, Microsoft, or Tesla. Not trimmed – completely filtered. Meanwhile, his own fund has been quietly building positions in energy infrastructure, nuclear, chip manufacturing, and natural resources — the physical backbone of everything described in this article.
He cannot buy most of those positions publicly.
But seven of them have a back door…
We think it’s among the most compelling AI plays hiding in plain sight.
