Insight
4 min read
The message from Invest-NL, the national financing and development institution, is as straightforward as it is ambitious. The Netherlands can play a competitive role in the world of artificial intelligence within five years, but not by entering the race for language models. That race has already largely been decided. The real opportunity lies in what comes next. 'Don't follow the Americans and Chinese by also developing large language models, find the niches,' says Jelle Vaessen, deeptech fund manager at Invest-NL, in a recent interview. [1]
That vision is the core of a new form of industrial policy. Rather than casting a wide net, targeted investments are being made in a coherent ecosystem of companies that excel in the hardware enabling 'next generation AI'. This includes the chips, sensors and networks that allow AI to understand and influence the physical world.
One of the most important niches is the development of energy-efficient AI chips that can run directly on devices, commonly known as 'edge AI'. While large language models require enormous, energy-hungry data centres, demand for efficient, local computing power is growing steadily. The International Energy Agency forecasts a doubling of data centre energy consumption over the next five years, which only amplifies the call for efficient alternatives. [1]
A prime example of this strategy is Eindhoven-based chip developer Axelera AI. This company, in which Invest-NL's Deep Tech Fund invested €15 million, launched its 'Europa' chip in October 2025. [2] [3] This AI Processor Unit (AIPU) delivers 629 trillion operations per second (TOPS), computing power previously only found in data centres, at a fraction of the energy consumption. The technology is promising enough that the European Commission awarded the company a €61.6 million grant in March 2025 to strengthen European digital autonomy. [4] Axelera AI is regarded as a serious European challenger to US giant Nvidia in the field of edge computing.
Another critical technology is photonics, in which chips operate using light rather than electricity. This makes data transfer exponentially faster and more energy-efficient, an essential prerequisite for increasingly complex AI applications. Here too, Invest-NL has taken a strategic position with an earlier investment in Smart Photonics, also from the Eindhoven Brainport region. [5] This company is one of the few independent producers of photonic chips in the world and forms a key component of the European Chips Act, which aims to reduce dependence on Asian chip manufacturing.
The combination of companies such as Axelera AI and Smart Photonics creates a unique Dutch ecosystem. One company produces the ultra-fast, energy-efficient AI computing cores; the other provides the infrastructure to transport data at the speed of light. Together, they form the hardware backbone for the next generation of AI.
Perhaps the most forward-looking niche is the focus on a new type of AI that is not trained on text from the internet, but on data from the physical world. Stefan Leijnen, AI expert at the University of Applied Sciences Utrecht and author of the 'Deepdive AI' report for Invest-NL, calls these 'world models'. [1] These models are fed data from cameras, radars, thermal and acoustic sensors, enabling them to learn to understand the physical world. Think of AI that can drive autonomously, control robots in a factory, or monitor crop quality.
"That new type of AI, the so-called 'world models', is not based on digital text, as are the large language models that OpenAI and Google are making their mark with," says Leijnen. [1] He believes that Dutch companies, with their strong position in high-tech manufacturing and sensor technology, have a unique advantage here. It is no longer about who has the most text data, but about who can collect and process the richest and most diverse data from the physical world.
Invest-NL's strategy is more than a series of individual investments. It is a deliberate attempt to build a coherent, sovereign European AI ecosystem, with the Netherlands at its heart. The idea of an 'AI giga factory of a new kind' is one example of this: not a centralised data centre, but a network of smart devices and servers running on Dutch chip technology. [1]
It is a long-term vision that requires patience and boldness. "We are only now beginning our search, mapping out the ecosystem," Vaessen acknowledges. [1] But the direction is clear. Rather than embarking on a futile pursuit of yesterday's AI, the Netherlands is opting for a leading role in tomorrow's critical technologies. By focusing on the niches in which it has historically been strong, hardware, engineering and systems integration, the Netherlands can become an indispensable link in the global AI chain. Not the largest, but the most strategically positioned.
This article is an interpretation and analysis based on the original article \"Nederland kan met gerichte investeringen serieuze AI-speler worden, denkt Invest-NL\", written by Jan Fred van Wijnen and published in Het Financieele Dagblad on 26 November 2025.
"}}]
CuspAI, founded in 2024 and operating from Amsterdam and Cambridge, uses generative AI to design new materials that can be discovered up to ten times faster than through traditional methods. The company has raised $130 million to date and works with partners including ASML, Meta and Hyundai.
Rotterdam-based Crownstone is building an indoor positioning system using Bluetooth modules mounted behind power outlets that draw power from the electrical grid. The system localises people and objects at room level and targets applications in healthcare, safety and building management.
The Dutch healthcare and health sector counts 78 active startups in 2026, a large share of which deploy AI for diagnostics, hospital logistics and home care. Funding is concentrated among a handful of players; broad adoption and regulation remain the principal bottlenecks.
