Europe may have lost the first phase of the artificial intelligence race. It does not control the largest cloud platforms, the most powerful foundation models, or enough of the specialised computing infrastructure on which advanced AI depends. But the next phase may be decided somewhere very different: on factory floors, in warehouses and inside machines.

As AI moves from generating text and images to perceiving, deciding and acting in the physical world, advantage may shift toward those who control industrial data, simulation systems and the software used to train robots. That creates an opening for Europe.

The paradox: Talent without scale

Europe’s AI position is a paradox. It has a large talent base, deep industrial expertise and some of the most automated factories in the world. Around 325,000 AI professionals work in Europe, while European AI startups raised more than $40 billion between 2025 and the first half of 2026.

Yet Europe remains weak where scale matters most. It attracts far less late-stage AI funding than the United States, holds only around 3% of global AI patents, and depends heavily on foreign computing infrastructure. Although it hosts roughly 16% of the world’s data centres, it controls less than 5% of the specialised computing capacity used for advanced AI.

Bodies, not yet brains

On the factory floor, however, the picture is different. Western Europe operates around 267 industrial robots for every 10,000 manufacturing workers. Germany has approximately 449, making it one of the world’s most automated manufacturing economies. This installed base gives Europe a potentially valuable starting point, but not yet a strategic advantage. A factory can operate thousands of robots without controlling the software, data or intelligence that makes them valuable. Many European machines still rely on old programmable logic controller code, proprietary interfaces and fragmented systems. Operational knowledge is often trapped inside individual factories, equipment vendors and engineering teams.

Europe has the bodies. It does not yet control enough of the brains. That distinction is becoming more important as European manufacturing faces stronger Chinese competition, labour shortages and pressure to raise productivity. Physical AI could help modernise factories and protect Europe’s industrial position. But simply buying more robots will not be enough. The strategic question is who will own the systems that train, coordinate and improve them.

Why simulation changes the game

This is where synthetic data and simulation become important. Training robots in the real world is slow, costly and potentially dangerous. A simulated factory can generate millions of examples by changing objects, layouts, lighting and physical conditions automatically. Robots can practise rare or hazardous situations without interrupting production or damaging equipment.

Recent research demonstrates the potential. One system trained on around 1.8 million synthetic robot trajectories achieved approximately 80% success on real-world manipulation tasks without physical training data or real-world fine-tuning. Other approaches can turn a single human video into a three-dimensional environment and generate robot-training examples from it.

Simulation will not eliminate real-world testing. Factories are too complex, and safety requirements too demanding, for virtual performance alone to be sufficient. But a large part of the expensive training process could move into simulation, leaving real-world deployment for validation, adaptation and continuous improvement. And that changes where value may accumulate. The most important assets in robotics may no longer be hardware alone, but synthetic datasets, industrial world models, simulation tools and deployment platforms.

Where Europe should play

Europe risks repeating the cloud era: European industry adopts foreign systems while the data, platforms and economic value remain elsewhere. Its large robot base becomes a strategic advantage only if European companies can connect machines, standardise industrial data and turn factory expertise into reusable training environments.

Europe does not need to lead every layer of AI. It should focus on the areas where it already has differentiated strengths: manufacturing, robotics, industrial automation and safety-critical systems.

Manufacturers should treat operational data and digital factory models as strategic assets. Technology companies should build software that connects legacy machinery with modern AI. Policymakers should support shared test facilities, simulation infrastructure and interoperable standards. Investors must provide the late-stage capital required to turn promising robotics companies into global platforms.

The window is opening

Investment is already moving in this direction. In June 2026, a German robotics company announced a funding round of around $1 billion to expand humanoid robots and an AI training platform. A 2,300-square-metre Physical AI training centre is also being developed at Munich Airport to train robots and generate learning data in Europe.

These initiatives matter because they point beyond hardware toward control of the training infrastructure itself. Europe already has the engineering talent, industrial environments and machines. The question is whether it can turn those assets into scalable software, data and platforms before foreign companies do.

Europe already has the robots. It must now build the intelligence behind them.