US and China together control roughly 90% of all global compute. And they house the winning LLMs. US + CHINA · 90% 10% for Europe and the rest of the world United States China Leading in the models and in the chips they run on. Winning on speed: state support, loose rules and cheap energy. Europe is trailing (badly) 2
DeepMind and AI unicorns like Synthesia and ElevenLabs Around them, top universities and the capital that follows. Talent does not have to relocate to King’s Cross Clustering. Talent, unicorns, US tech, change jobs. capital and research in one postcode. The UK also has a minister for AI. One point of contact, one agenda. Financial Times · John Gapper · King's Cross is the Silicon Roundabout of AI 4
Sweden vs the Netherlands Sweden counts 41 unicorns on 10.5 million inhabitants. The Netherlands lands at 15, with 17.5 million. Stockholm as scale-up capital Founders of Spotify, Skype and Klarna put their money back into the next generation. Pay-it-Forward. Those who made it fund the next generation. That is what On top of that a layer of tech hubs, investors, universities and talent. produces unicorns like Lovable and Legora. DutchStartup.ai · The Stockholm mystery: how Sweden became Europe's unicorn champion 5
blocks are already here. 01 02 03 04 Talent Research Multinationals Infrastructure Highest AI talent density in Europe. World-class institutes. Booking, Prosus and ASML based here. Excellently connected, open to the world. But advantages mean nothing without action. We have to get moving. 6
sector data, not with the biggest model. 01 Martech European leader in AI martech. 02 Fintech Amsterdam is the fintech capital of the continent. 03 Health & biotech World-class medical research as the starting point. 04 Agtech & logistics Second largest agricultural exporter in the world, plus the mainport of Europe. 7
2/3 89% of multinationals use AI 21% For SMEs it is one in seven. builds on an American model 10.9 Our own AI factory opens mid2027. AI scale-up ratio European average: 31%. AI professionals per 10,000 The densest talent pool in Europe. Sources: Techleap, CBS 8
Studio Hyra) The platform that makes the Dutch AI startup sector visible. Partner & SVP at DEPT® Set up the European data practice, member of the global and European MT. Tech entrepreneur FD Gouden Gazelle · finalist EY Entrepreneur of the Year · Deloitte Fast50. Willem Blom Driven by AI, innovation, sport, family and sustainability. 11
makes the Dutch AI startup sector visible, findable and connected. Startups, founders, investors and ecosystem partners in one place. Database & dashboard Interactive map All AI startups, searchable. AI innovation across the Netherlands. Founder stories Content hub The people, with video interviews. Articles, analyses and news. Events calendar Ecosystem All the relevant AI events. Knowledge institutes and investors. Community upkeep Decentralised management by the startups themselves. 13
COMPANY VALUATION WHAT THEY BUILD CuspAI $2.6bn AI for materials discovery. Amsterdam and Cambridge, platform July 2026 MIRA. General Intuition $2.3bn World models for robotics, trained on gameplay from Medal. Nearfield Instruments €1.4bn Axelera AI over $1bn January 2026 Chip inspection from Rotterdam, with roots at TNO. June 2026 Energy-efficient AI chips for the edge. The layer beneath the models. February 2026 9
2023, Series B 2024 AI for protein design Amsterdam, founded in 2021 by a Dutch-Swiss team. Software that designs antibodies, enzymes, peptides and vaccines with machine learning. THE MOAT Named Startup of the Year 2024 by MT/Sprout. Their own wet lab in Amsterdam. Models On customer data the claim is explicit: sequences and experimental data stay with the customer, and only train that customer's model. are trained and validated on real experimental data, not on a generic large model. Amsterdam · own wet lab · Series B in 2024 15
Thomas Wolf (Hugging Face) AI assistant for doctors Turns conversations between doctor and patient into structured notes, and finds relevant data back in the electronic health record. SOVEREIGNTY Works with seven EHR vendors. Positions itself as an intelligent layer on top of existing European cloud partners and MDR systems, not as a replacement for them. certification as a core choice, not an afterthought. Of the three, the most Its own language model, trained specifically on Dutch medical language. sharply argued example. Michel Abdel Malek (founder, anaesthesiologist who also codes) 16
Computer vision for inventory management Amsterdam. The restaurant manager films the storage room. The model pulls inventory data out of it and flags anomalies: an unusually high tomato use stands out immediately. MOAT IN THE MAKING The domain knowledge sits in the team. Both founders come out of operations at Uber and Every video trains the model further, have ties to family-run hospitality businesses. under constantly changing conditions. Where a restaurant's footage ends up has not been spelled out publicly. Harpreet Singh and Sebastien Pradier (founders, both ex-Uber) 17
02 Winning customer trust Data quality & due diligence European companies deliberately pick a local AI partner. A model is only as good as its training data. GDPR and the AI Act are not a burden, they are a selling point. Governance weighs heavily with investors assessing your IP. 19
& 27001 Data lineage & synthetic Sovereign infrastructure The new ISO 42001 framework for Startups fully document the Storage on certified European AI management systems works as provenance of training data. Where cloud hubs, such as Exoscale or an immediate tick in the box, privacy is a bottleneck, they switch local AWS clusters, keeps data closing large enterprise B2B to high-quality synthetic data. flows within EU jurisdiction. contracts faster. 20
TOOL VALUE FOR AI STARTUPS Synthetic data Syntho Generates privacy-proof synthetic data to train models without risking real personal data. Privacy & security Roseman Labs Enables Multi-Party Computation: collaborating securely on sensitive data without parties seeing the raw source. AI Act compliance AI Heroes / ActCheck Scans and frameworks to meet the transparency and documentation requirements of the EU AI Act. Data lineage Collibra / Datahub Maps data flows automatically, from source system to final AI output. 21
21% against 80% Data out of the Dutch polder share of AI startups that scale up · the Netherlands versus the US From startup to scale-up Cradle validates in its own wet lab. We need more risk-taking founders and more venture capital. Deeptech already proves it: 12% of the ecosystem, 40% of the scale-ups. Delphyr trains on Dutch medical language. They win with data that physically sits here. 23
mark3t1ng
tonyrice
mkilby
katiecoart
twada
productmarketing
eileencodes
davidcarrasco
inesmontani
jonoalderson
marktimemedia
sonatard
axbom
