New research from AWS, ‘Built in Europe: The Rise of Physical AI’, finds that much of Europe’s physical AI sector has already moved out of the lab and into the market. [1] More than half of startups (51%) have their technology in commercial use, and most (76%) sell to large enterprises and industrial groups. The next challenge is scale: turning pilots into paid contracts and raising the money to grow.
The report is based on a quantitative and qualitative survey of founders and senior leaders of physical AI startups headquartered in the European Union and in the United Kingdom.
Physical AI start-ups are already selling, and demand is coming from Europe’s own industry
More than half (51%) have their technology in commercial use, and most (76%) sell to large enterprises and industrial groups, with manufacturing and automotive (47%) and logistics (38%) as the most common.
85% percent of physical AI founders say their first production customers are, or will be, European, and six in ten have more pilot demand than they can currently serve. Customers adopt the technology mainly to fill labour and skills shortages (32%), almost twice as often as to cut labour costs (17%).
Europe’s edge: industrial customers and world-class universities
Being close to Europe’s industrial base helps, with 78% saying that being near European industrial customers gets them to production faster. Europe’s universities are also producing many of the founders. Almost half (46%) of these startups are university spinouts or were founded by researchers and recent graduates, most often from ETH Zurich, TU Munich and the University of Cambridge.
“Europe has one of the densest industrial bases in the world, and being close to those customers makes a real difference,” said Tanuja Randery, Managing Director for Europe, the Middle East and Africa at AWS. “Combine that with world-class universities producing the founders, and you have something few regions can match: great research creating the companies, and Europe’s industrial leaders becoming their first customers.”
Europe’s physical AI founders are ready to scale. The step from pilot to production and capital are holding them back.
The step from pilot to production is where many startups stall. More than two-thirds (68%) of founders agree that most European physical AI startups get stuck at pilot stage, and only 36% of those who have run pilots have converted most or all of them into paid deployments. Slow enterprise purchasing is the obstacle they name most often (64%). For startups already in commercial deployment, the most important factor in getting there was a customer willing to co-develop with them.
Nearly six in ten (59%) say access to capital is a major challenge. The need is greatest at the manufacturing and scale-up stage, when companies must build machines, set up production and support customers on site before revenue can carry the business. There are signs of progress, with half of founders saying fundraising conditions for physical AI are better than a year ago.
Founders also want to stay: 77% would choose to headquarter their business in Europe if they were starting it today. The task now is making sure they can scale here too.
“Physical AI is moving from the lab into production because better models, more simulation capacity, more capable hardware and the compute to connect them have come together at once. My advice to a founder with a working prototype is to build the continuous loop of data, simulation, training and deployment that takes it to production.”
A plan to close the gap
The report sets out four areas for action:
- Close the capital gap by growing growth-stage funding from corporate, strategic and international investors, and connecting founders to them earlier. Public funding is founders’ top ask of government (46%).
- Shorten the path from pilot to production, with enterprises running pilots with a clear route to deployment and governments acting as early customers.
- Protect and grow the talent pipeline through targeted funding for technical skills.
- Build on international standards and harmonise regulation, so a startup approved in one country can deploy across others with less repeated work.
AWS is supporting Europe’s physical AI founders as they move from pilot to production. In October 2026, AWS launched the Physical AI Toolchain, an open-source stack developed with NVIDIA that brings synthetic data generation, training, simulation, edge deployment and continuous improvement into one workflow. Alongside credits and technical support, AWS connects founders with enterprise customers willing to co-develop with them and with investors who can fund their next stage of growth.
Read the full report.
[1] Physical AI, meaning artificial intelligence that senses, reasons, acts and learns in the real world, now runs robots on factory floors, drones that inspect infrastructure and autonomous vehicles.
The research is based on a survey of 500 founders and senior leaders of physical AI startups headquartered in the UK, France, DACH (Germany, Austria and Switzerland), the Nordics, Benelux, Southern Europe and CEE, and was conducted by Strand Partners in August 2026. A physical AI startup is defined as a company whose products or technology enable machines to sense, reason, act and learn in the real world, including robotics, drones, autonomous vehicles, the AI models and software that control them, and enabling technology such as simulation and synthetic data. Companies that only use physical AI built by others were not included.
