Future computing power for AI requires vision and collaboration

Europe has the ambition to become more digitally sovereign. This means that European countries and organisations are able to make independent choices and maintain control over the digital infrastructure, data and technology they use. At the moment, most European countries and organizations are still very dependent on American technology and services made by Big Tech companies. That is why there are many initiatives to increase digital autonomy both in the member states and at the European level.

The importance of more computing power

This dependence on Big Tech and the lack of digital autonomy is particularly true for our data infrastructure, such as our data 'cloud' and the infrastructure needed for large-scale use of artificial intelligence (AI). According to estimates, America currently holds 78% of the computing power ('compute') for AI and Europe only 5%. This dependence on data and AI is therefore a major stumbling block to more digital autonomy.

To reduce this dependency, the European Commission wants to invest in AI and cloud infrastructure through the Cloud and AI Development Act (CADA). Among other things, it aims to triple the number of data centers, for example by making it easier to obtain permits, and to develop next-generation computing power to support AI and other data-intensive services.

Protests against data centers

However, developing digital technology and building the necessary computing power in the form of data centers is not without controversy. In the United States and the United Kingdom, the construction of data centers has now become a political factor: the construction of several data centers has now been halted as a result of local protests. Because of the gigantic energy consumption of data centers, we are also seeing first protests in the Netherlands, such as recently in Amsterdam.

While we already have major spatial challenges in the Netherlands, for example as a result of nitrogen deposition and grid congestion, there is now also a spatial and energy demand for data centers to provide the development of AI with computing power. Because more European computing power is needed for digital autonomy, there is therefore a risk that there will also be opposition to the construction of new data centers.

New forms of computing power

The development of increasingly data-intensive 'general purpose AI models', which require more and more computing power, not only requires the use of current compute technology, but also requires new compute technology that is more efficient. For example, there are already more efficient generative AI models, which do not require more and more computing power, but comparable performance with less computing power. An example is the Chinese generative AI model DeepSeek, which uses a specific architecture.

In addition, there is also a development that focuses on more efficient computing power that such AI models can use. This allows AI models to be developed and deployed more efficiently. This computing power is more efficient than the current generation of technology, because it uses emerging technologies such as quantum, photonics and neuromorphic computing. This should lead to hybrid computing systems that make use of all these forms of compute that build on each other's strengths. This combination of technologies is called 'Future of Compute'.

Dutch Future of Compute collaboration

In the Netherlands, too, 'Future of Compute' technology is being looked at. As a country, we are very well positioned with solid innovation ecosystems for the individual technologies. Examples include QDNL for quantum computing, AI Coalitie voor NL (AIC4NL) for AI, PhotonDelta for photonics, ChipNL and Holland High Tech for semicon, and NC-NL and NL-ECO for neuromorphic computing. All these innovation ecosystems operate on a Dutch scale, with key innovation parties working together.

However, the cooperation between these ecosystems could be better; at the moment, little work is being done on truly hybrid compute technology. An overarching vision of, and agenda for, the Future of Compute as it exists in the United Kingdom is also still lacking in the Netherlands. That is why a number of organizations, Digital Holland, TNO, SURF, imec, Holland High Tech and Invest NL, have called for the development of such a vision and agenda for the Netherlands.

To make Europe more digitally autonomous in the field of cloud and AI, which requires much more European computing power, and to prevent us from having to deal with similar AI backlash in the Netherlands as in America, focusing on future, more efficient computing technologies is therefore an interesting opportunity. Especially for the Netherlands, which is already firmly positioned in various future computing power clusters. The Netherlands is also looking at whether it is possible to make these emerging technologies work together. However, this requires a strong ambition and more cooperation between the various innovation ecosystems.