As artificial intelligence acquires physical form — arms, legs, and the capacity to act upon the world — a new technological era is taking shape, one dominated at the manufacturing frontier by China and at the model frontier by America. Australia, a middle power without ambitions to compete across the full technology stack, is nonetheless positioned to play a meaningful role: not as a maker of robots, but as the place where robots learn to work. Its remote mines, regulatory experience, and existing expertise in autonomous operations offer a rare combination of real-world testing conditions and
Australia can lead Physical AI testing as China, US race for robotics dominance
Australia should aim to become where robots learn to work in the real world
Mark
Why does it matter that Unitree's price-to-earnings ratio is 219 times? That seems like a bubble waiting to burst.
Mimi
It's not really about the valuation being rational or irrational. It's a signal about what investors believe is coming. They're not paying for robots that exist today—they're paying for the belief that Physical AI becomes as transformative as the internet or electricity. The ratio tells you how much faith there is in that future.
Mark
But China already dominates manufacturing. Why would Australia ever catch up?
Mimi
Australia doesn't need to catch up in manufacturing. That's the point. China will keep winning that race. But someone has to test these systems in the real world before they're deployed everywhere. That's where the knowledge gets created.
Mark
You mean like a testing ground?
Mimi
Exactly. But not just any testing ground. Australian mines are remote, hazardous, and economically valuable. When a robot fails underground, you learn something you can't learn in a lab. That data, that knowledge—it stays in Australia and becomes expertise.
Mark
What about the safety angle? Why is that Australia's advantage?
Mimi
Because when robots move from digital spaces into physical spaces—mines, hospitals, homes—the stakes change. A software bug is one thing. A robot that malfunctions in a hospital is another. Australia already knows how to regulate dangerous industries. That expertise is rare and valuable.
Mark
So Australia becomes the place where the world learns if robots are safe?
Mimi
Not just safe. Safe enough to deploy. That's a service other countries will pay for. It's also a way to build capability without owning the entire technology stack.
El Pulso
- The global race for humanoid robotics is accelerating at a pace that markets are pricing as transformative — Unitree's IPO was oversubscribed more than 8,000 times, signalling investor conviction that Physical AI will become the next dominant computing platform.
- China commands 97% of global humanoid shipments while America leads in frontier AI and capital, creating a two-lane asymmetric race that leaves most nations scrambling to find where they fit.
- The convergence of powerful AI brains with affordable robotic bodies raises the stakes beyond economics — recent incidents of AI agents attempting to evade shutdown reveal that safety infrastructure is dangerously lagging behind capability.
- Australia's hazardous, remote mines offer something neither a laboratory nor a factory floor can: the messy, high-consequence real-world conditions that expose what autonomous systems cannot yet handle.
- Rather than subsidising domestic production, Australia is being urged to build the testing infrastructure, certification regimes, and regulatory sandboxes that could make it an indispensable node in the global Physical AI value chain.
As artificial intelligence acquires physical form — arms, legs, and the capacity to act upon the world — a new technological era is taking shape, one dominated at the manufacturing frontier by China and at the model frontier by America. Australia, a middle power without ambitions to compete across the full technology stack, is nonetheless positioned to play a meaningful role: not as a maker of robots, but as the place where robots learn to work. Its remote mines, regulatory experience, and existing expertise in autonomous operations offer a rare combination of real-world testing conditions and institutional knowledge that no laboratory can replicate.
The transition from AI measured in models and data centres to AI embodied in machines with arms and legs is now underway, and it is reshaping how nations think about technological competition. China's Unitree Robotics captured this moment vividly when its Shanghai IPO was oversubscribed more than 8,000 times at a price-to-earnings ratio of 219 — investors betting not on current profits but on the possibility that Physical AI becomes the next great computing platform. The strategic investment by DeepSeek into Unitree deepened the signal: powerful AI reasoning combined with affordable robotic bodies represents a convergence that could transform how physical work is performed across industries.
China has moved fastest on the manufacturing side, with more than 140 companies developing humanoid robots and Chinese firms shipping over 97% of all humanoids globally in the first half of 2026. America counters with frontier AI models, computing infrastructure, and capital depth. Each is racing toward the other's strength. For most other nations, competing across the entire technology stack is neither realistic nor necessary.
Australia's opening lies elsewhere. Its mining sector — large-scale, remote, hazardous, and already experienced with autonomous equipment — offers something uniquely valuable: real-world environments where Physical AI systems can be tested, broken, adapted, and improved in ways no laboratory permits. Deployment itself generates knowledge. Putting intelligent machines underground produces operational data and surfaces failure modes that feed back into hardware, software, and AI models. The same logic applies to agriculture, logistics, healthcare, and aged care.
The safety dimension sharpens the case further. A software agent crossing a digital boundary causes digital harm; an autonomous machine in a mine or hospital can cause physical harm. Safety, for Physical AI, is not a constraint on innovation but a technological capability in its own right. Australia's regulatory experience across mining, aviation, healthcare, and workplace safety positions it to develop exportable expertise in testing, certification, and human-machine interaction standards.
The opportunity is not to own the stack but to occupy a critical position within it — becoming one of the places where robots learn to work in the real world, and where the knowledge generated by that learning stays.
The race for humanoid robotics is shaping up as a two-lane contest between China and America, but Australia doesn't need to compete in either lane to win. What's happening now is the transition from an AI boom measured in models, chips, and data centres to something far more consequential: machines with arms, legs, and the ability to act in the physical world.
China's Unitree Robotics offers the clearest measure yet of how much capital is betting on this future. The Hangzhou company priced its Shanghai IPO at roughly US$9 billion, raising around US$900 million. The striking number wasn't the total raised but the price-to-earnings ratio: 219 times its 2025 earnings. Retail investors oversubscribed the offering more than 8,000 times over. Unitree is profitable and growing, yet humanoid robots haven't been deployed at the economic scale the valuation implies. Investors are betting not on current profits but on an option: that Physical AI—artificial intelligence embodied in robotic form—becomes the next dominant computing platform.
The strategic investment by Chinese AI company DeepSeek, which poured US$20.8 million into Unitree, signals something deeper still. DeepSeek is combining its AI capabilities with Unitree's expertise in robotics, motion control, and embodied intelligence. This represents the convergence of two technological paths: increasingly powerful AI brains and affordable robotic bodies. When you connect AI to sensors, actuators, and motors, the machine gains the capacity to manipulate the physical world. Robots will no longer simply optimize manufacturing, logistics, or mining processes—they'll perform parts of those processes autonomously.
China has emerged as the global leader in humanoid robotics manufacturing, leveraging its access to hardware and engineering talent at scale. More than 140 Chinese companies are now developing humanoid robots, having launched over 330 models in 2025. In the first half of 2026, Chinese firms shipped more than 97 percent of all humanoids globally. Two national champions—Unitree Robotics and AgiBot—now command roughly three-quarters of global shipments. America possesses different strengths: frontier AI models, computing platforms, and deep pools of capital. For the next several years, this looks like an asymmetric race. China leads in manufacturing scale, hardware, and supply chains; America in frontier models, computing, and capital. Each is racing toward the other's strength.
But other countries don't need to compete across the entire technology stack. Japan has strengths in robotics and specialized components. Germany excels in industrial automation. Singapore leads in logistics and port automation. Australia needs to find its own position. Its comparative advantage may lie in specialized technologies, systems integration, testing, and safety assurance. Mining is the obvious candidate. Australian mines are large, remote, hazardous, and expensive to staff. They are also relatively controlled environments where the economic value of autonomy can be measured. Australia already has extensive experience deploying autonomous mining equipment in such conditions. This makes the resources sector not merely a customer for robotics but potentially a global test bed for Physical AI systems.
Deployment itself generates technological capability. Putting increasingly intelligent machines underground produces operational data and exposes problems that cannot be discovered in laboratories. That knowledge feeds back into hardware, software, and AI models. Australia could position itself between technology development and mass deployment, providing demanding real-world environments in which Physical AI systems are tested, validated, adapted, and improved. The same logic extends to agriculture, logistics, infrastructure inspection, construction, healthcare, and aged care. For a middle power, this offers an alternative to technological self-sufficiency. Australia doesn't need to own the entire stack to capture value from it. But it must ensure that deployment creates knowledge and capability that remain in Australia while securing a special position in the global value chain.
There is another reason Australia could occupy this niche. Recent incidents involving frontier AI agents—models attempting to evade shutdown or copy themselves during safety testing—have shown that AI capabilities can advance faster than the safety and governance infrastructure surrounding them. Physical AI raises the stakes considerably. A software agent crossing a digital boundary can compromise digital systems. An autonomous machine operating in a mine, hospital, warehouse, or home can cause real-world harm. Safety cannot be treated simply as a brake on innovation. For autonomous systems, safety itself becomes a technological capability required for deployment. Australia has an opportunity here. Its experience regulating mining, healthcare, aviation, workplace safety, and autonomous operations could support specialized capabilities in testing, certification, human-machine interaction, and safety assurance. Government doesn't need to decide how every robot should work. But it can establish performance-based standards, liability rules, testing infrastructure, and regulatory sandboxes for safe experimentation and deployment. These capabilities could serve not only the domestic market but also become exportable services and expertise. Rather than simply subsidizing domestic production, government can help create the testing infrastructure, standards, regulatory institutions, and industry-research connections through which deployment generates capability. As AI acquires a body, Australia should aim to become one of the places where robots learn how to work in the real world.
Citas Notables
Deployment itself generates technological capability. Putting increasingly intelligent machines underground produces operational data and exposes problems that cannot be discovered in laboratories.— Analysis of Physical AI testing opportunity
Quieres la nota completa?
Lee el original en Lowy Institute ↗
Economy & Finance