Big Data, Big Compute: HPC Provides Computing Power for AI

The development of powerful large language models (LLMs) places high demands on computing capacity and energy efficiency. While leading global technology companies rely on their own supercomputers and specialised data centres, Europe must expand its strategic capacities in High-Performance Computing (HPC) in order to keep up with global competition.

Modern LLMs are not only data-intensive, but also extremely computationally demanding. Their development and subsequent operation require an enormous amount of computing power, which can currently only be provided by specialised HPC systems. So far, American and Chinese providers have dominated the market for AI models – a fact that presents Europe with the challenge of developing its own trustworthy and competitive solutions.

This requires not only the further development of algorithms, but also the provision of sufficient computing resources in the form of high-performance data centres. Government investment, an innovation-friendly regulatory environment and targeted support for research institutions play a key role here. The current developments in the efficient design of LLMs (see e.g. deepseek) do not fundamentally change this.

Andreas Weiss, Managing Director of eco – Association of the Internet Industry, emphasises: ‘Europe urgently needs its own high-performance computing capacities in order to remain independent and competitive in the field of generative AI. Without this essential infrastructure, we run the risk of permanently lagging behind in the development of powerful LLMs or being degraded to consumers altogether. Our goal must be to build scalable and safe AI infrastructures across Europe that meet our high standards of data protection, trustworthiness and innovation.’

Building high-performance and sustainable HPC infrastructures is therefore not only a technological necessity for Europe, but also a strategic one. ‘Without its own supercomputing resources that are optimised to meet the requirements of LLMs, there is a significant competitive disadvantage in terms of digital sovereignty,’ explains Weiss and adds: ’Increased investment and coordinated efforts by business, science and politics are therefore required to create the basis for independent and innovative European AI development.’