“It’s the most humbling thing I’ve ever seen,” Ford’s chief executive Jim Farley said about his recent trip to China. After visiting a string of factories, he was left astonished by the technical innovations being packed into Chinese cars. “Their cost and the quality of their vehicles is far superior to what I see in the West. We are in a global competition with China, and it’s not just EVs. And if we lose this, we do not have a future at Ford.”
It’s not just the Americans feeling the pinch. The supposedly invincible German automotive industry is being dismantled in real-time. BYD sold 47,000 vehicles across Europe in Q3 2025, up 238% year-on-year. Chinese brands collectively captured 11.3% of Europe’s EV market in the first half of 2025, more than double their 4.8% share a year earlier. In Norway, Chinese EVs claimed 23% market share in September alone.
Just cheap copies?
Whilst Tesla’s Model 3 costs €42,990 and BMW’s i4 is €57,500 in Germany, BYD’s Seal sells for €31,990. These aren’t cheap knock-offs. Reviewers praise BYD’s build quality and interiors. They’re great products at dramatically lower prices.
Volkswagen announced in November it would close its Zwickau EV plant, the first closure of a German factory in the company’s history, citing “unsustainable competition from subsidised Chinese imports selling 30% below our production costs.”
European governments haven’t exactly been shy with automotive support—Renault’s €5 billion loan guarantee, VW’s €1.4 billion from Lower Saxony, billions in consumer EV subsidies across the continent. Both sides subsidise. But China is winning because they build better automated factories and iterate faster.
Stellantis CEO Carlos Tavares warned in October: “Chinese manufacturers operate with 25–30% cost advantage that we cannot match with current structures.”
This isn’t hypothetical future competition. BYD opened its Hungarian factory in July 2025 and announced a €2 billion investment in Spain. They’re not exporting to Europe—they’re building here.
How We Lost the Lead
For decades, Germany led the world in industrial automation. German manufacturers pioneered robot deployment, achieving the highest robot density globally. That automation advantage powered Germany’s reputation for precision manufacturing.
But in the last decade, China deployed at massive scale. From 189,000 industrial robots in 2014 to 2.03 million today—a tenfold increase. In 2024 alone, China installed 290,000 robots. China now has 470 robots per 10,000 manufacturing workers, surpassing Germany’s 429. The United States manages 295. The United Kingdom just 111.
China jumped from ninth to third place globally in robot density in four years whilst maintaining the world’s largest manufacturing workforce of 37 million workers. At current installation rates, China adds the UK’s entire robot stock every four months.
Executives describe visiting Chinese “dark factories” so heavily automated that lights are unnecessary. Xiaomi operates one in Beijing producing 10 million mobile phones annually with zero human labour. One phone every three seconds, 24 hours a day, 365 days a year.
Europe didn’t lose automation capability. We lost momentum deploying it at scale.
Physical AI: The Next Transition
China’s current advantage stems from traditional industrial automation—the robots that have existed for 60 years, deployed at extraordinary speed and scale.
But the next wave threatens to make that advantage permanent.
Physical AI represents flexible, language-directed robots that understand natural language commands and generalise across environments. Tell a robot “fold the towel” versus “stack the dishes” and it understands and executes. It works in new settings without prior training. This didn’t exist before 2023.
Chinese institutions already lead the research. Tsinghua’s world models won Best Paper at Robotics: Science and Systems 2025. ByteDance collaborates with top universities on Vision-Language-Action models. Chinese research groups publish prolifically on physical AI foundation models.
At current trajectories, China doesn’t just dominate manufacturing automation. They’re positioned to own Physical AI infrastructure entirely—the foundation models, the robot manufacturing, the deployment scale.
For European manufacturing, this looks terminal. Not just losing current battles, but being shut out of the next generation entirely.
Unless we understand what Physical AI actually requires.
What Customers Actually Buy
Dark factories don’t emerge from foundation models alone. They require sophisticated applications: quality inspection systems that predict defects, process optimisation tools that coordinate hundreds of robots, predictive maintenance platforms that prevent failures, warehouse software that adapts to changing inventory patterns.
Physical AI applications are the transformation layer that turns general-purpose robots into productive factories.
And applications are what customers buy. BASF doesn’t purchase foundation models. They buy AI inspection systems for chemical plants. Volkswagen doesn’t license robot platforms. They buy process optimisation for assembly lines. Hospitals don’t deploy general-purpose robots. They buy surgical planning systems.
This pattern holds across technology waves. Applications captured value in PCs (Microsoft over IBM), in the internet (Google over Cisco). Infrastructure enables, but applications concentrate value because applications own customer relationships.
Foundation models and robot platforms will commoditize. Value accrues to applications solving specific industry problems—quality inspection, predictive maintenance, warehouse coordination, surgical planning, pharmaceutical optimisation.
And Europe is good at building applications.
Europe’s AI Category Leaders
Just look at the AI companies that have come out of Europe in the past few years.
Synthesia serves 90% of the Fortune 100, defining the global standard for enterprise AI video generation. ElevenLabs set the worldwide benchmark for AI voice synthesis, serving 60% of Fortune 500 companies. Wayve is building the leading application layer for autonomous driving behaviour models. N8N competes directly with American workflow automation platforms, winning globally. Lovable demonstrates European capability in developer-facing AI applications.
These aren’t just unicorns. They’re AI category leaders. Global standards in their domains. They competed against well-funded American companies on product quality and execution speed—and won.
Physical AI creates the same opportunity. But they are not pure software. They require understanding mechanical systems, manufacturing processes, industrial constraints. Europe’s engineering heritage matters here. And this is what enables European founders to create global leaders in manufacturing, logistics, construction, agriculture, energy, and healthcare.
The European Ecosystem
Isn’t AI all happening in Silicon Valley? Synthesia, ElevenLabs, Wayve, N8N, Lovable suggest otherwise. These aren’t exceptions—they’re part of a broader pattern.
European venture capital has been growing rapidly over the past decade. From 2015 to 2024, European VC investment totalled €426 billion—ten times the €43 billion invested in the entire decade before. European unicorns grew from 72 in 2015 to 358 today, a fivefold expansion. In 2025, nearly a quarter of new unicorns globally are European companies. Europe’s VC ecosystem grew at 13% CAGR over the past decade, outpacing the US’s 8%.
The founders exist. The capital exists (although we’d like more at the growth phase). The customer opportunities exist across sectors representing over $50 trillion in combined market value. European manufacturers, hospitals, construction companies, and pharmaceutical firms provide immediate deployment opportunities and urgent demand.
Through Physical AI to Manufacturing Competitiveness
The executives returning from Chinese factory tours are right to be concerned. European manufacturing must compete with China’s dark factories. The cost advantage is real. The automation gap is widening.
Physical AI applications are how Europe closes that gap. Quality inspection systems, process optimisation tools, predictive maintenance platforms enable European factories to compete. And these same applications sell globally—to manufacturers, logistics companies, pharmaceutical firms worldwide.
Building these companies isn’t choosing between manufacturing competitiveness and software value. It’s achieving both.
Europe already builds AI category leaders. We led industrial automation for decades. Physical AI applications are how we reclaim that leadership whilst building the next generation of global category leaders.
The dark factories are spectacular. Europe will build them too. Physical AI applications are how we get there.