Physical AI Is Already Working, Not Just Doing Backflips
Physical AI and robotics have existed for years. Many companies in different forms already do jobs at a speed, rate, and quality that beats human performance. What is new is humanoid robots starting to compete in the same spaces people work in. That gives you hands that can do many tasks, and those hands do not need to sit on a humanoid body. They can work on a conveyor belt in normal manufacturing, or hand off a package between point A and point B for self-running logistics. Each of these is a new unlock stacked on the last, driven largely by progress in AI.
The market history: a big growth spurt in logistics deployments in 2021, then a decline in the overall robotics market for several years as PMI slowed. Now an "ultra bull" wave has arrived, and a bear case against the sector is getting harder to find.
Where the Revenue Comes First
Humanoids get the most attention because they are fun to watch and people imagine owning one at home. But the money spreads across warehouses, factories, logistics, health care, and agriculture - all of the above.
Robotics wins on two sides at once. First, capabilities keep expanding. Second, robots build out the data centers themselves. Chips are not etched by humans with magnifying glasses; these are highly automated systems. Servers get built, then packaged into end products deployed in servers and warehouses, plus electrification work around it.
Of the robotics companies tracked in the Robo Index, 92% are already profitable. One example is Advantech, a Taiwanese company that makes industrial PCs. It supports other technologies you hear about - edge sensors, computer vision, and more - by partnering with companies that deploy those systems rather than building the whole stack alone. Advantech's embedded and edge intelligence division is showing large growth. It is expected to be 30% of the business, was 50 at the end of this year, was 20 or 15% at the end of last year, and earlier was under 10%. That is a real win happening now, before the bigger unlock - humanoids in homes and other new use cases - has even started. The backflips and demos are not full-scale deployment yet, but that is coming.
Who Else Benefits
Nvidia (NVDA) supplies the hardware, software, and "brains" for many robots, even when the shell is made by Unitree. Intel (INTC) plays a similar supplier role. Nvidia's Jensen Huang has spent years in South Korea talking to SK Hynix about robotics.
The enabling technology and infrastructure breaks into subsectors. A single robot is a very different story from deployment at scale, and all the parts have to work together:
- Sensors
- Systems on chips that act as controllers for the brains
- Software
- Controllers and motors, meaning precision energy and electricity
The core question now is how precise and useful the extracted energy can be, whatever the source - oil, gas, solar, or nuclear - and how much energy is lost in transit. This is "physics maxing": the companies making the most money are the ones pushing the limits of physics itself. That means moving data at higher bandwidth and lower cost, inside a data center or through the cloud, and securing machine identities. Soon there will be more agents and more robots than humans, and securing those machine identities is a big market.
A whole ecosystem takes an individual robot up to fleet scale in systems operations. This will run across every industry and environment - homes, and out beyond warehouses and manufacturing plants. It is bots making bots. Many of these companies are already profitable, already blue-chip leaders in their divisions, and watching their addressable market expand daily. Robotics is a now-and-later winner, and the physical AI revolution is here.
