The public debate about AI is almost exclusively focused on the cognitive: chatbots, text generation, knowledge work. That is understandable — it is where most people encounter the technology today. But it obscures a parallel development that in the longer term could be equally transformative: the acceleration of physical automation and the geopolitical race that is driving it.
In my own essays so far, I have emphasised how AI is changing cognitive work, institutions and the labour market. This text complements that picture. The point is simple: if you only look at software, you miss half the shift. The actor who controls the physical automation infrastructure — robots, sensors, manufacturing capacity, supply chains — will have a different kind of power than the one who merely uses models that someone else has built.[1]
Why the physical wave is coming now
Industrial robotics has existed for decades. What is new is that AI makes robots more flexible, cheaper to programme and useful in more environments. Previously, each new task required expensive specialist programming. With AI-based control, machine vision and world models, systems can learn tasks faster and adapt to variation that previously required human judgment.
At the same time, the cost of hardware is falling. Cheaper sensors, better servomotors and more standardised platforms mean that robotics is no longer reserved for car manufacturing and semiconductor factories. It is starting to become accessible for food production, logistics, small-scale manufacturing and care.
It is the convergence of cheaper software and cheaper hardware that is now accelerating the wave of physical automation. And it is doing so unevenly — with dramatic regional differences.[2]
The geopolitical landscape
During 2024, an estimated half million new industrial robots were installed globally. The total operating base exceeded 4.6 million units. But the distribution is telling.[1]
China accounted for 295,000 of the installations — 54 per cent of the entire world's new robots in a single year. The country's robot density has doubled in four years to 470 per 10,000 industrial workers, which now exceeds both Germany and Japan. Chinese manufacturers have increased their domestic market share from 28 to 57 per cent. China treats AI and robotics not as an industry issue but as a national operating system: the AI+ Action plan targets 70 per cent AI adoption in research, industry and the public sector by 2027.[3]
Europe installed around 85,000 units, but the trend is declining — minus 8 per cent. Despite a strong engineering tradition and countries like Germany with high automation levels, the continent is hampered by fragmented regulation, lack of computing infrastructure and difficulty scaling rapidly. Europe leads in ethics and regulation but lags in physical implementation.
North America accounted for roughly 40,000 installations. The US strength lies in frontier research and venture capital, but the country has long been weaker at building physical infrastructure. That is now changing through a "Fortress" strategy: reshoring of semiconductor manufacturing and precision automation, driven by geopolitical uncertainty rather than cost alone. The geographical centre is shifting from the Rust Belt to a new south-west corridor in Texas, Arizona and New Mexico.
Japan and South Korea still have the world's highest robot density per capita and strong component manufacturing. India, with only 9,100 installations, is instead betting on using AI for large-scale social impact and trying to skip older development phases.[1][3]
What this means for a small open economy
Sweden is rarely mentioned in these global comparisons. Our industrial base is strong but small. We have advanced manufacturing, high digital maturity and a tradition of automation. But we do not have the scale, computing capacity or raw material base required to be an independent player in the race for physical automation.
That is not necessarily a problem — if we understand what it means. Sweden cannot compete with China on volume or with the US on frontier research. What we can do is be skilled users, integrators and niche producers. But that requires understanding the dependency chain: which components, platforms and systems we import, from whom, and what happens if supply is constrained.
It also requires us to take the physical side of the AI debate as seriously as the cognitive side. Today, the Swedish AI discussion is dominated by questions about data, models, ethics and skills. These are relevant questions. But there is no serious discussion of automation infrastructure, industrial resilience and what it means that Europe's share of global robot installation is shrinking.
The convergence
The really decisive dynamic is not cognitive automation or physical automation taken separately. It is the convergence. When AI models can control robots, plan production flows, optimise maintenance and coordinate supply chains in real time, a new kind of production system emerges that is qualitatively different from earlier automation.
In that system, software and hardware merge. The one who has only the models but not the machines is dependent. The one who has only the machines but not the intelligence loses competitiveness. That is why the question of who owns the automation infrastructure — not just the algorithms — becomes a strategic question for every industrial nation.
What I see from the factory floor
In my own projects, I already see the two worlds beginning to meet. AI helps with process analysis, experiment design and decision support. But in the background, a slower movement is underway: new sensors, more data from production lines, better integration between control systems and analysis tools. Step by step, the factory's information flow becomes denser and more accessible for AI-based control.
It is not dramatic in daily life. But it is building capacity that at some point could change what is possible to automate in physical environments. Anyone who does not keep up with that development risks waking up in a world where competitors not only think faster but also manufacture, distribute and adapt faster.
Conclusion: the other half of the shift
The AI debate needs to grow. The cognitive side is important and well discussed in these essays. But the physical side — robotics, manufacturing capacity, supply chains, geopolitical positioning — has equally great strategic significance.
Sweden has good conditions if we understand our position: we are not a superpower in the automation race, but we have industrial expertise, digital maturity and institutional strengths that could be valuable if directed correctly. What is required is that the physical automation question be lifted from the shop floor specialist discussion to the same strategic level as the rest of AI policy.
Source notes
The source notes below support the regional data and strategic trends discussed in the essay.
- International Federation of Robotics (IFR), World Robotics 2025. Global installation figures, regional distributions and robot density per 10,000 employees. IFR press release; IFR robot density.
- CIO, The geopolitics of AI and robotics: China, the US and Japan (2025), on strategic differences between regions. CIO. See also IIoT World on regional automation trends 2026. IIoT World.
- Xpert.Digital, AI strategies in a global comparison (USA, EU, Germany, Asia, China). Xpert.Digital. On China's AI+ Action and systemic integration, see also Stanford HAI, AI Index Report 2025. Stanford HAI.