PAL Robotics Built Humanoids for 20 Years Before It Was Cool. They Are Still in the Game.

In 2004, three engineers in Barcelona started building humanoid robots. Mark Zuckerberg had just launched Facebook from a Harvard dorm room. YouTube did not exist yet. The iPhone was three years away. The idea that humanoid robots would become a multi-billion dollar industry attracting the world’s biggest tech companies was, generously speaking, a fringe position.

Davide Faconti, Francesco Ferro, and Luca Marchionni founded PAL Robotics in Barcelona that year. They were not chasing a trend. There was no trend to chase. They were building humanoid robots because they believed in humanoid robots, at a time when the entire field was dismissed as academic curiosity.

Twenty-two years later, PAL Robotics is still building humanoid robots. They have shipped approximately 10 humanoid units. They employ between 80 and 150 people. They operate from Barcelona with no venture capital backing, no billionaire investors, and no factory capable of producing thousands of units per year.

They are also the only humanoid robot manufacturer from Europe in the global race tracker.

That number - 10 units - deserves to sit alongside the numbers from the rest of the tracker, because the contrast is the story.

The gap between Europe and everyone else

The robotsin.life race tracker currently tracks 15 companies building humanoid robots. Six are Chinese. Five are American. One is Canadian. One is Norwegian (though 1X Technologies manufactures in Norway and operates from Palo Alto). One is South Korean. And one is European.

One.

PAL Robotics, based in Barcelona, Spain, with its 10 shipped humanoid units and its team of fewer than 150 people, is the entire European humanoid robotics manufacturing base represented in the tracker. That is not because the tracker has a bias against European companies. It is because there are no other European companies building humanoid robots at any meaningful scale.

China has shipped over 13,000 humanoid units from six companies. The United States has shipped over 2,000 from five companies. Europe, home to 450 million people, the world’s third-largest economy, and some of the best engineering universities on the planet, has shipped 10. From one company. In Barcelona.

This is not a manufacturing problem. It is a structural problem. And understanding PAL Robotics - what they built, how they survived, and why they never scaled - is the key to understanding why Europe is absent from the humanoid race.

The robots PAL actually built

PAL Robotics has produced an impressive range of robotic platforms over two decades, even if the unit volumes remain small. Each platform tells part of the story.

REEM (2006). PAL’s first humanoid, a full-size bipedal robot designed for service applications. REEM could navigate indoor environments, recognize faces, and interact with people through speech. It was one of the first European humanoid robots built outside a university lab.

REEM-C (2013). An open-source full-size humanoid platform that became one of the most widely used research humanoids in European academia. Standing 1.65 meters tall and weighing 80 kg, REEM-C runs entirely on ROS (Robot Operating System) and was specifically designed for research institutions that needed a fully programmable humanoid platform. It remains in active use at universities across Europe.

TIAGo (2015). A mobile manipulation robot with a single arm on a wheeled base. TIAGo became PAL’s most commercially successful product, deployed in research labs, hospitals, and elder care facilities. It is not a humanoid in the strict sense, but it represents PAL’s bread-and-butter revenue.

TALOS (2017). A full-size bipedal humanoid (1.75 meters, 95 kg) with torque-controlled joints, designed for research into whole-body control and dynamic locomotion. TALOS is arguably the most technically sophisticated humanoid robot ever built in Europe. It has 32 degrees of freedom and can perform dynamic walking, manipulation, and multi-contact tasks.

ARI (2020). A social robot designed for human-robot interaction research, deployed in healthcare, retail, and hospitality research settings.

Kangaroo (2024). PAL’s latest bipedal platform, representing a new generation designed with commercial deployment in mind.

The breadth of this portfolio is remarkable for a company of PAL’s size. Over 22 years, with fewer than 150 people, they designed, engineered, manufactured, and supported six distinct robotic platforms spanning service robots, research humanoids, mobile manipulators, social robots, and bipedal locomotion platforms. Most American robotics startups with ten times the funding have built one platform.

The problem is not what PAL built. The problem is how many they built. Or rather, how few.

How PAL survived without venture capital

PAL Robotics did not raise a $675 million Series B like Figure AI. They did not receive strategic investment from BYD or Samsung. They were not acquired by Hyundai for $1.1 billion. They survived for 22 years the old-fashioned way: by selling robots and winning research grants.

PAL’s revenue model has three pillars.

Direct robot sales. PAL sells its platforms - primarily TIAGo, REEM-C, and TALOS - to universities, research institutions, and government agencies. A single TALOS unit reportedly costs several hundred thousand euros. REEM-C and TIAGo units are priced lower but still represent significant research equipment purchases. At these price points and volumes, total annual revenue from robot sales is likely in the single-digit millions of euros.

Research contracts. PAL participates in EU-funded research projects through Horizon Europe (formerly Horizon 2020 and FP7 before that). These multi-year, multi-partner projects provide steady funding for R&D. PAL has been involved in dozens of EU framework program projects over the past two decades. Each project typically provides hundreds of thousands of euros over two to four years.

Service and support. Ongoing maintenance, software updates, and technical support for deployed robot platforms generate recurring revenue.

This model works for survival. It does not work for scale. EU research grants typically fund technology development, not mass production. University customers buy one or two units, not hundreds. Government research contracts have long timelines and modest budgets. The entire funding model is designed to keep a small company alive while advancing the state of the art, not to build a factory that produces 10,000 humanoids per year.

Compare PAL’s funding history to what happened across the Atlantic and the Pacific in the past three years alone. Figure AI raised $1.85 billion. AgiBot raised $140 million with automotive giants as strategic investors. Tesla redirected an unknown but surely enormous portion of its manufacturing and AI resources toward Optimus. 1X Technologies raised over $125 million in a single Series B round.

PAL had none of that. And the absence of that capital is not a failure of PAL’s founders. It is a failure of Europe’s innovation ecosystem.

The EU research robotics ecosystem

To understand why PAL exists in its current form, you have to understand the ecosystem that shaped it. European robotics operates within a specific institutional framework that has real strengths and critical weaknesses.

The strengths are genuine. Europe has world-class robotics research. ETH Zurich, EPFL, TU Munich, Imperial College London, CNRS in France, IIT in Italy, KTH in Sweden, and dozens of other institutions produce fundamental breakthroughs in locomotion, manipulation, machine learning, and human-robot interaction. The quality of European robotics research is equal to or better than anything produced in the US or China.

Europe also has a mature regulatory environment and a cultural emphasis on safety, ethics, and human-centered design that produces robots genuinely suited for deployment around people. PAL’s robots are designed from the ground up for safe human interaction, not as factory machines adapted for human environments.

The weaknesses are structural. EU research funding flows through framework programs (currently Horizon Europe) that are designed for collaborative, multi-institution research projects. These programs are excellent at producing papers, advancing knowledge, and training researchers. They are poor at turning research into products and products into companies that ship at scale.

The funding cycle looks like this: a consortium of five to ten partners from multiple EU countries submits a proposal. If funded, the project runs for two to four years. At the end, the consortium produces deliverables, publications, and prototypes. Then the project ends. The researchers move to the next project. The prototypes usually go into a lab somewhere.

There is no European equivalent of DARPA, the US Defense Advanced Research Projects Agency, which funds high-risk, high-reward technology development with explicit commercialization pathways. There is no European equivalent of China’s MIIT guidelines that directed provincial and municipal governments to create humanoid robot industrial parks and provide manufacturing subsidies. There is no European equivalent of the American VC ecosystem that poured $1.85 billion into a single humanoid robot company.

Europe has the research talent. Europe has the engineering capabilities. Europe does not have the mechanisms to turn those into scaled production.

What it means to be the only one

PAL Robotics is not just a company. It is a data point about an entire continent’s strategic position.

Consider what the absence of European competition means in practical terms.

Supply chain dependency. When European industries eventually need humanoid robots for manufacturing, logistics, elder care, and service applications, they will have to buy them from American or Chinese companies. Europe will be importing a critical automation technology rather than producing it. This is the solar panel story, the battery story, and the semiconductor story repeating itself in yet another domain.

Brain drain. European robotics researchers who want to work on commercial humanoid programs at scale have to leave Europe. They go to Figure AI, to Tesla, to 1X, to Boston Dynamics, or to Chinese programs. The research institutions produce the talent. The talent leaves. The institutions train more talent. That talent also leaves. The cycle accelerates.

Standards and regulation. Europe tends to set global standards through its regulatory weight (the “Brussels effect”). If Europe is not manufacturing humanoid robots, its ability to shape the standards and regulations around them weakens. The standards will be written by the companies that build the robots, which are in San Jose, Austin, Shanghai, and Hangzhou - not in Barcelona or Munich.

National security. Humanoid robots will eventually have applications in defense, border security, disaster response, and critical infrastructure maintenance. A continent that cannot build its own humanoid robots will be dependent on foreign suppliers for these applications, just as Europe discovered its dependency on foreign suppliers for energy, semiconductors, and cloud computing.

The customer base that kept PAL alive

PAL’s survival for 22 years without venture capital is itself a noteworthy achievement. The company survived because it found a niche that, while small, was durable: the European research institution.

European universities and government research labs have steady (if modest) budgets for robotics equipment. They need humanoid platforms for locomotion research, manipulation studies, human-robot interaction experiments, and AI development. They prefer platforms that run on ROS, that are fully documented, that come with technical support from engineers who speak the same language (both literally and in terms of research culture), and that comply with European safety standards.

PAL’s REEM-C and TALOS fill this niche almost perfectly. They are high-quality, well-documented, ROS-native humanoid platforms backed by over two decades of engineering expertise. For a European research lab, buying a PAL robot is straightforward. Buying a Chinese humanoid requires navigating import regulations, dealing with different software ecosystems, and accepting documentation that may not be available in European languages.

But this niche is also PAL’s ceiling. Research institutions buy single units. They do not place orders for a hundred. The total addressable market for European research humanoid platforms might be a few hundred units over a decade. That is enough to keep a small company alive. It is not enough to build a manufacturing base that can compete with AgiBot’s Shanghai factory or Tesla’s Fremont lines.

What would it take to change this

The question that hangs over PAL Robotics and European robotics more broadly is straightforward: what would it take for Europe to become a real player in the humanoid race?

The honest answer is uncomfortable. It would require a combination of changes that the European system is structurally resistant to making.

Capital. Someone would have to invest hundreds of millions of euros in a European humanoid robotics company, not as a research grant, but as commercial investment aimed at building manufacturing capacity. The European Investment Bank, national development banks, or private investors would need to make a bet comparable to what Figure AI’s backers made. Nothing in Europe’s current investment landscape suggests this is imminent.

Industrial partners. A European automaker, electronics manufacturer, or logistics company would need to do what BYD did for AgiBot: invest as a strategic partner with guaranteed deployment sites. Volkswagen, Stellantis, Siemens, Bosch, or Inditex could theoretically play this role. None of them has made a comparable move in humanoid robotics.

Speed. EU framework programs run on multi-year cycles. Grant applications take months. Project evaluations take months. Consortium formation takes months. In the time it takes to get a Horizon Europe project approved, AgiBot builds and ships a thousand robots. The timelines are incompatible with the pace of the current race.

Risk tolerance. European investors and institutions are famously risk-averse compared to their American and Chinese counterparts. Humanoid robotics is a high-risk, high-reward domain. The willingness to fund a company that might fail spectacularly but might also create a new industry is far lower in Europe than in Silicon Valley or Shanghai.

PAL in the context of the race

Here is the uncomfortable truth that PAL Robotics embodies: being early does not matter if you cannot scale.

PAL was building humanoid robots before Unitree existed. Before AgiBot existed. Before Figure AI existed. Before Tesla announced Optimus. Before anyone in China’s current humanoid wave had shipped a single unit. PAL had a 12-year head start on Unitree and a 19-year head start on AgiBot.

Unitree has shipped 5,500 humanoid units. AgiBot has shipped 5,200. PAL has shipped 10.

The head start did not matter because PAL operated in an ecosystem that was designed for research, not production. The Chinese companies that now dominate the shipment numbers entered the market later but entered with manufacturing infrastructure, strategic industrial partners, government support for commercial deployment, and a venture capital landscape willing to fund hardware at scale.

This is not a criticism of PAL’s founders or employees. They built excellent robots under severe constraints. They kept a company alive for 22 years in a domain that has killed countless startups. They contributed significantly to the global state of humanoid robotics knowledge. Their open-source contributions through REEM-C and their ROS packages have benefited researchers worldwide.

But survival is not the same as competition. PAL Robotics survived. It did not compete. It could not compete, because the ecosystem around it was never designed to let it compete.

The lesson for Europe

PAL Robotics is both a success story and a cautionary tale, depending on what you measure.

If you measure technical achievement, longevity, and contribution to the field, PAL is a remarkable success. A small company in Barcelona, without venture capital, built some of the best humanoid robots in Europe over two decades.

If you measure market impact, manufacturing capacity, and competitive position in the global humanoid race, PAL is a warning sign. Europe has effectively zero presence in a market that Goldman Sachs projects will reach $38 billion by 2035. The continent that invented the word “robot” (from the Czech “robota,” coined by Karel Capek in 1920) has exactly one humanoid manufacturer in the tracker, and that manufacturer has shipped 10 units.

PAL Robotics built humanoids for 20 years before it was cool. Now that it is cool, the company finds itself in the paradoxical position of having more experience than almost anyone in the field and less market impact than companies that are two or three years old. The experience is real. The robots are real. The 22 years of engineering knowledge are real. What is missing is everything around the company: the capital, the industrial partners, the policy framework, and the market conditions that would let that experience translate into scale.

Europe’s single humanoid manufacturer is still in the game. The question is whether the game will wait for them.