The race for robotics is not between robot companies. It is between AI software platforms racing to become the invisible brain inside every robot on earth. The hardware is almost irrelevant. The OS is everything.
In 1980, IBM needed an operating system for its new personal computer. Microsoft licensed one, adapted it, and sold it to IBM. Then — and this is the move that changed everything — Microsoft retained the right to license that same operating system to other manufacturers. IBM thought they were buying exclusivity. They were not.
Within five years, Compaq, Dell, HP, Toshiba, Acer, and dozens of others were all shipping IBM-compatible PCs running Microsoft software. Microsoft had not built a single computer. They had built the layer that made every computer possible. The hardware became a commodity price war. The software captured the value — for thirty years.
That is precisely what is about to happen with robots. One platform — NVIDIA Isaac, Google DeepMind, Figure's Heuro OS, Physical Intelligence, or a dark horse nobody is watching yet — becomes the Windows of robotics. The moment that happens, hardware becomes a commodity almost immediately. And the platform that wins owns the next twenty years.
"Microsoft didn't manufacture a single PC. They wrote the layer that made a thousand manufacturers possible. That is precisely what is about to happen with robots."
The traditional S-curve assumption for robotics is wrong — and wrong for the same reason people underestimated Android in 2009. The iPhone scaled slowly because Apple controlled every component. One manufacturer, one supply chain, one price point, one cadence. The curve was orderly and predictable.
Android exploded differently. The moment Google released Android openly, fifty manufacturers shipped simultaneously. Samsung, LG, Huawei, Xiaomi, Oppo, Vivo, Realme, Tecno — from $3,000 flagships to $49 entry-level devices. Every price band, every geography, every carrier, simultaneously. The market did not grow on a curve. It detonated.
Robotics will follow the Android model, not the iPhone model. Three structural reasons: margins compress before scale when fifty-plus manufacturers compete globally from day one. China alone has the manufacturing infrastructure to move from 10,000 units to 10 million units in 18 months once the software stack is ready. And a robot OS moment opens floodgates to every contract manufacturer on earth simultaneously. The curve does not bend gradually. It breaks vertically.
NVIDIA is not building robots. They are building the compute layer, the simulation environment (Isaac Sim), and the foundation models (GR00T) that every robot company builds on. This is the Microsoft playbook exactly. NVIDIA's advantage: they already supply the chips. Owning the silicon and the OS is an almost unassailable position. The risk: they are currently a platform for developers, not yet a product for manufacturers.
DeepMind's robotics research — RT-2, SayCan, and the latest multimodal models — represents the deepest scientific foundation in the field. If the robot OS requires genuine general intelligence rather than narrow task performance, DeepMind may be the only organisation with the models to deliver it. The risk: Google has a long history of world-class research and losing the commercialisation race.
Pi was founded by former Google, OpenAI, and academic researchers specifically to build a general-purpose robot brain. They have raised significant capital, published compelling research on policy learning, and have no legacy business to protect. Startups have won platform races before — Android was built by a small team acquired by Google. Pi could be acquired or could win outright.
Figure is attempting the Apple model — controlling hardware and software simultaneously. This could be the strongest position if achieved, or the most fragile if a pure-software competitor commoditises their hardware. The BMW partnership signals enterprise buyers are ready. The question is whether Heuro OS can be licensed or remains proprietary to Figure's own robots.
Foxconn manufactures for Apple, Tesla, NVIDIA, and 200+ companies. They have the tooling, the workforce, every supply chain relationship on earth, and the capital. The moment a licensable robot OS exists, Foxconn can ship a credible humanoid at $4,000 within 18 months. They are already building Tesla Optimus. They need to license one thing — not invent anything.
Atlas and Spot represent fifteen years of hardware development — balance, locomotion, manipulation — that no startup can replicate quickly. Hyundai paid $1.1 billion for Boston Dynamics and has been patient. If the OS race produces a winner, Boston Dynamics hardware is the most ready body to run it.
The moment a licensable OS exists, Indian and Vietnamese manufacturers will enter the race targeting the price point that makes robots accessible to small businesses globally. India's electronics manufacturing is scaling rapidly — the same trajectory as mobile phones five years ago. This is a pattern repeating, not a prediction.
Saudi Arabia's Vision 2030, the UAE's AI strategy, and India's robotics mission are all funding national robot initiatives with state-level capital and urgency. These are not research projects. They are industrial policy. Sovereign buyers at scale change the economics of any platform — the way government procurement built the early internet.
"The race is not between robot companies. It is between AI software platforms racing to become the invisible brain inside every robot on earth. Find that platform before the market names it. That is where the quadrillion-dollar compounding begins."
The $1Q thesis identifies AI and automation as the fastest-moving of the 16 forces compounding toward a quadrillion-dollar global economy by 2040. Robotics is where AI leaves the screen and enters the physical world — warehouses, factories, hospitals, construction sites, farms. Every sector that employs human labour at scale is addressable.
The platform that wins the robot OS race captures a share of every economic activity that robots displace or augment. This is not a sector play. It is a horizontal infrastructure play — the same way Microsoft captured a share of every business that ran on a PC, and Google captured a share of every business that needed to be found online.
The investor question is not which robot company to back. It is which software platform becomes the Windows of robotics. That company does not need to manufacture a single robot. It needs to be the layer that every robot runs on. Find that company before the market names it. That is where the compounding begins.
When Microsoft created Windows, the software itself was not the economic event. The economic event was everything that grew because of Windows. Intel built chips to run it. Dell built computers to ship it. Oracle built databases on it. SAP built enterprise software on top of it. Adobe built creative tools for it. Cisco built networks to connect it. A million IT services companies built careers maintaining it. Every office on earth hired people to operate it.
The PC became the platform for the entire knowledge economy. Windows did not create one industry. It created the conditions for fifty industries to emerge simultaneously — most of which nobody predicted in 1985.
Now multiply that by physical reality. A robot OS does not just create robot companies. It creates an entirely new physical-digital ecosystem — industries that do not exist today and will be worth trillions by 2035.
"The Windows moment for robotics is not just a technology story. It is a youth story. A billion young people in emerging markets are arriving at the exact moment the platform is being laid. The $1Q is not a forecast. It is an underestimate."
Each of these ancillary industries is a multi-billion dollar opportunity that does not meaningfully exist today. The robot OS is not the destination. It is the starting gun. What follows — the sensors, the actuators, the maintenance networks, the insurance markets, the robot-native software, the physical infrastructure — is an economic explosion that compounds on itself for decades.
The $1Q thesis envisions a quadrillion-dollar global economy by 2040. The robotics ecosystem — the OS and everything it makes possible — may be the single largest contributor to that number. Not because robots replace humans. But because they expand what is economically possible. The ceiling on human economic output has always been the number of human hours available. Robots remove that ceiling entirely.
Part of an ongoing journal — observations recorded when something in the world economy is worth saying. No schedule. No noise. The next entry when there is something worth adding.