Spain’s THEKER raises $85 million: Europe finally produces its own contenders in general-purpose robotics

The race for artificial intelligence is no longer being fought solely inside data centres. After three years dominated by language models, computing infrastructure and software agents, a new frontier is now attracting capital: physical intelligence.

Barcelona-based startup THEKER has announced an $85 million funding round, equivalent to approximately €73 million. The transaction is the largest Series A ever completed in European robotics. Co-led by CRV, the round also includes Samsung, Cathay Innovation, LVMH, 20VC, Henkel, Korelya and Sonae.

Viewed in isolation, this financing could appear to be simply another funding round in a sector already generating a steady flow of announcements. Seen from a broader perspective, however, it illustrates a much deeper transformation: the emergence of a new generation of European companies seeking to apply advances in artificial intelligence to real-world operations.

Coming only hours after announcements from Germany’s NEURA Robotics, the growing prominence of autonomous-defence companies such as Alta Ares, and industrial demonstrations by Chinese players led by Unitree Robotics, THEKER’s funding round confirms that robotics is becoming one of AI’s principal fields of application.

Artificial intelligence moves beyond the screen

The first phase of the AI revolution focused on automating cognitive work. Models developed by OpenAI, Anthropic, Google DeepMind and Mistral AI demonstrated their ability to produce text, generate code, analyse documents and support complex decision-making. At the same time, hundreds of billions of dollars were committed to building the computing and energy infrastructure required to sustain this growth.

This first phase created an intelligence capable of understanding. The second seeks to create an intelligence capable of acting.

That is precisely the ambition of general-purpose robotics. Where traditional industrial robots perform predefined tasks within perfectly controlled environments, new generations of systems are seeking to adapt to the variations of the real world. They must recognise different objects, manage unexpected situations, modify their behaviour and learn from experience.

In other words, the same principles that enabled the emergence of foundation models are now being applied to robotics.

Building a physical-intelligence platform

Founded in 2022 by Carla Gómez Cano and Jiaqiang Ye Zhu, THEKER develops robots designed to operate in complex industrial environments without requiring constant reprogramming.

The company claims an “AI-native” approach in which adaptability lies at the heart of the system. Its robots can adjust to production changes, handle different product references and incorporate new operational parameters without undergoing the lengthy integration cycles traditionally associated with industrial robotics.

THEKER’s real product, however, is probably not the robot itself. As in generative artificial intelligence, value is shifting towards the software layer. Visual perception, decision-making, motor control, data collection, continuous learning and operational orchestration combine to form a platform that resembles an operating system for the physical world more than a conventional industrial machine.

The comparison with language models is becoming increasingly relevant. Where OpenAI aims to become a universal cognitive layer, companies such as THEKER are seeking to become a universal operational layer capable of interacting with industrial processes.

A funding round extending far beyond a Spanish startup

The composition of the funding round is probably its most revealing feature.

CRV has historically been associated with several major successes in US software, including DoorDash, Mercury and Vercel. Its investment in THEKER reflects the growing interest of software investors in robotics.

Hardware was long regarded as a difficult, capital-intensive sector that was less scalable than software. AI is gradually changing this equation. Investors now see the emergence of platforms capable of generating network effects built on data and continuous learning.

Samsung’s participation also sends an important signal. This is the Korean group’s first investment in a Spanish startup. The decision reflects the growing attention being paid to the convergence of electronics, artificial intelligence and automation.

The most surprising investor, however, is probably LVMH. At first glance, the connection between a luxury group and an industrial-robotics startup may seem unexpected. Yet it reflects a profound shift in industrial priorities. Logistics-flow management, quality control, product personalisation, flexible automation within workshops and operational optimisation all have the potential to transform the luxury sector’s industrial models over the long term.

LVMH’s investment illustrates a conviction that extends far beyond its own operations: robotics could become a strategic infrastructure comparable to the role artificial intelligence now plays across digital functions.

Europe seeks to move higher up the value chain

For several decades, Europe held a strong position in industrial automation. Germany built global leaders in industrial equipment. Switzerland established itself across several specialised segments. France developed recognised expertise in industrial software and certain robotics technologies. Yet integrated robotics platforms capable of matching American or Chinese ambitions remained rare.

This situation is changing rapidly. Several European companies pursuing comparable ambitions are now emerging alongside THEKER. NEURA Robotics is developing a complete robotics ecosystem around its Neuraverse platform. ANYbotics is expanding in autonomous industrial inspection. Wandercraft is extending its work on advanced robotic systems, while Exotec continues to demonstrate Europe’s ability to industrialise robotics platforms at scale.

The leading US players remain formidable. Figure AI, Physical Intelligence and Skild AI are attracting hundreds of millions of dollars to build the future foundation models of robotics.

The difference is that Europe is no longer seeking merely to manufacture components. It is now attempting to build complete platforms.

The robot is no longer the real strategic asset

As in generative AI, the debate often focuses on models or visible machines.

Yet value could become concentrated elsewhere. Today’s artificial intelligence leaders enjoy a considerable advantage because they accumulate enormous volumes of data and possess the infrastructure required to exploit it. Robotics follows a similar logic.

Every movement performed inside a warehouse, every object handled on a production line, every error corrected and every environment explored generates new data. This information makes it possible to improve system performance and gradually reduce deployment costs.

The result is a learning loop comparable to the one observed in language models. The central question therefore becomes less about the robot itself than about the data it accumulates.

The industry’s future leaders could be those controlling the largest volumes of physical interactions from the real world.

General-purpose robotics still has something to prove

Investor enthusiasm should not obscure the remaining questions. The term “general-purpose robot” has become ubiquitous in startup presentations across the sector. Industrial history, however, has often demonstrated the economic superiority of highly specialised and perfectly optimised systems.

A production line does not need a robot capable of performing one hundred different tasks. It needs a robot capable of executing one critical operation perfectly, millions of times without error.

Industry imposes constraints with which the software world has little familiarity: safety, availability, maintenance, precision and reliability.

The next challenge: deployment

Another blind spot for the industry is industrialisation. Technology demonstrations are multiplying. So are funding rounds.

The essential question, however, remains largely unanswered: how many robots are actually being deployed at scale? The recent history of artificial intelligence shows that technology is not always the main obstacle. In many cases, the difficulty lies in integration with existing systems, user adoption and the transformation of operational processes. Robotics could follow the same trajectory.

The challenge over the coming years will not simply be to build more intelligent robots. It will be to deploy them across thousands of factories, warehouses and infrastructure sites where every service interruption carries an immediate economic cost.

As the United States advances, China accelerates

Another question remains relatively absent from European debates. Most analyses focus on competition from the United States, yet China’s progress could prove equally decisive.

Companies such as Unitree Robotics and UBTECH Robotics benefit from unique proximity to the country’s supply chains, manufacturing capabilities and industrial infrastructure.

China already holds a dominant position in several strategic hardware segments. If it succeeds in reproducing this advantage in AI-powered robotics, the competition could rapidly extend far beyond technology alone.