The National AI Delta Plan was presented in late November 2025 at the request of the Dutch Ministry of Economic Affairs. Its central message is alarming: the Netherlands risks losing control over AI, even as this technology is set to shape the economy, security, and democracy. The authors warn that the Netherlands is overly dependent on American and Chinese AI companies and that, without its own infrastructure, economic value creation will flow abroad.
The four pillars
The plan contains 52 recommendations, structured around four themes:
1. Technological foundations:
The development of a computing capacity plan (mapping, on the demand side, how much computing power the Netherlands and Europe will need under different scenarios), major investments in data centres and energy supply, and accelerated permitting in AI computing zones (enabling rapid expansion of computing capacity by removing bottlenecks in energy and grid infrastructure).
2. AI adoption and literacy:
The Netherlands is lagging far behind in AI use and adoption, driven by high levels of scepticism and a lack of knowledge. The plan calls for large-scale reskilling: all teachers, politicians, and civil servants should receive AI training. It also advocates large-scale responsible adoption within government and education, and exploring partnerships with private actors, as other countries have done, for example with OpenAI or Mistral.
3. A competitive AI ecosystem:
Measures to retain and attract talent, including revising the expat scheme, establishing an AI hub in Amsterdam, easing dismissal law for highly paid positions, and reforming public procurement strategies so that startups and scale-ups have better access. Efforts are also proposed to attract AI companies to the Netherlands and to develop career pathways that allow technical talent to further develop within government.
The plan includes several recommendations to strengthen the Dutch entrepreneurial climate for AI innovation. First, public procurement should be reformed to give innovative companies better access to the annual €113.6 billion in government contracts, which currently largely go to large firms. This is inspired by the UK example, where a new procurement strategy led to forty times more SME engagement and significant savings in 2024. Second, the plan proposes special economic zones (such as low-regulation zones) where companies can experiment with innovations in the physical domain—such as autonomous vehicles, drones, and robotics—with temporary exemptions from existing regulation. Successful innovations from these zones should be translated into national regulation through a structured feedback loop. Third, the plan calls for a clear definition of startups and scale-ups as a basis for innovation policy, taking into account sectoral differences in capital intensity and growth patterns, so that deep-tech companies with long development cycles are not excluded from schemes by arbitrary revenue or employment thresholds.
4. Democratic embedding:
The plan advocates the establishment of a National AI Impact Institute to monitor effects on work, well-being, and democracy. It also calls for a permanent infrastructure of citizens’ assemblies, in which representative groups of 100–200 randomly selected citizens are systematically involved in major AI-related decisions concerning public infrastructure, data use, and societal conditions. Inspired by international examples such as Taiwan, this approach should begin in 2026 with a first citizens’ assembly on a specific AI topic. Finally, measures are proposed to counter AI-driven manipulation and disinformation.
Institutional proposals
Central to the plan are the appointment of a State Secretary for AI within the Ministry of Economic Affairs and the creation of a National AI Strategy Agency, inspired by the UK’s Sovereign AI Unit. This agency would be responsible for strengthening the Dutch position across the global AI stack: ecosystem development, talent, data, infrastructure, security, and international partnerships. In addition, the plan calls for the establishment of a Dutch ELLIS institute as a European flagship for AI research, and a National Agency for Disruptive Innovation (NADI) modelled on DARPA.
The ELLIS institute should collaborate with existing Dutch structures such as ICAI and ELSA labs, place strong emphasis on valorisation through spin-offs and industry collaboration, and provide researchers with access to sufficient computing power to remain internationally competitive.
The NADI should achieve disruptive breakthroughs through high-risk R&D programmes involving teams from academia and industry. It should have sufficient autonomy, entrepreneurial leadership, and operate with performance-based evaluations so that programmes can rapidly scale up or be terminated.
Reception and criticism
The plan has been positively received by some as ambitious and urgent. At the same time, there is substantial criticism. First, it is seen as a wish list without clear prioritisation, attempting to address everything at once without making clear choices about where the Netherlands aims to excel. Second, the plan is predominantly economic and technological in nature, with fundamental questions about ownership, democratic control, social inequality, and public values receiving insufficient attention. Third, it reflects a form of technological tunnel vision, in which spectacular technical solutions are overvalued compared to social, political, or institutional solutions that are less “exciting.” Fourth, the plan is perceived as primarily a wish list of the tech elite, focusing on tax advantages, relaxed dismissal law, faster data centre permitting, and increased access to risk capital.
