Digitalisation of science stands as a prominent megatrend. The process of gathering, transferring, analysing and eventually understanding data requires digital tools and infrastructure. Artificial Intelligence has a growing role in assisting decision making and information retrieval. As datasets expand in size and resolution, computing capacity requirements surge due to simulation and modelling needs. Data is generated in fast cycles and the amount of data is growing rapidly. Competent individuals are in high demand, also across industries, revealing a scarcity in human resources in many areas, emphasising the need to set additional requirements in education and training in digital skills.

HPC

Despite rapid advancements in processor technology, large supercomputer installations such as FUKAGU FUKAGU 
https://www.fujitsu.com/global/about/innovation/fugaku/  in Japan or Frontier Frontier 
https://www.olcf.ornl.gov/frontier/  in the USA aim to prolong systems lifespans beyond the typical five years of most HPC centres. Extending the longevity of HPC technology can be enabled by investing in code optimisation and higher processor performance. This is possible since today code efficiency varies considerably, with some cases utilising only a small percentage of the hardware potential. The trend for the largest installations is to run longer in production than before.

In HPC, the rise of AI applications such as Large Language Models (LLM), has been a clear advancement, boosting the use and development for GPUs and related software. Applications such as ChatGPT divide opinions and impact developmental trajectories. Concerns from five years ago about GPU usability due to the lack of applications and high application porting efforts, seem not warranted anymore.

Traditionally, HPC systems have been purchased and maintained by national and regional centres, research organisations, or companies. However, in the future, owning HPC systems and hosting them locally may not remain the most efficient strategy for numerous service providers. This is the case especially when a change is identified in the behaviour of users which may want to shop for HPC resources the same way they presently do with cloud. Huge differences in operational cost, mainly due to the variation in electricity price, will drive the choices of future data centre locations and attract user communities or even nations to collaborate and possibly share cost and resources for joint HPC ownership and operation.

QUANTUM COMPUTING

In the future, the HPC systems will include quantum processors alongside traditional ones. However, it will take some years before quantum computers achieve the reliability necessary to solve practical problems. Nonetheless, ongoing technology and software developments are paving the way. Quantum computers are complementary solutions to supercomputers and will not replace them. Their problem-solving scope remains in fact limited to specific application areas yet to be fully explored. However, in these specific applications, quantum computing performance is expected to far surpass traditional supercomputers.

Although quantum computing applications are still in the future, Europe is actively advancing technology development and competence building. EuroHPC has funded a number of quantum projects (six initiated from a call launched in 2022, one from an earlier one), where various types of quantum computers are built and integrated to supercomputer systems. These projects collaborate with quantum computer vendors and technology centres aiming not only to advance technology and devices but also to explore their integration with supercomputer systems. The development of software stack and user interfaces will play a key role in future quantum integration.

DEFINING & DESIGNING DEMONSTRATORS AND TESTING OF NEW TECHNOLOGIES

The European HPC partners are globally at the forefront of the evaluation of new technologies, assessing their relevance for operations of new architectures and their efficiency in solving existing and novel scientific and technical challenges. EuroHPC and PRACE members can set up collaborative frameworks where new technologies are evaluated against specific needs of communities, particularly addressing industrial needs on the basis of proxy and benchmarking applications. New Horizon Europe funding boosts European research in data, computing, and AI technologies: a new set of calls has been launched worth over € 290 million from the 2023-2024 Horizon Europe Digital, Industry, and Space work programme.

GLOBAL COLLABORATION

ICT infrastructures have been recognized as “a crucial asset underpinning European research and innovation policies”.Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, ICT infrastructures for e-science
https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52009DC0108&from=DA  Significant progress has been made in the deployment of pan-European ICT infrastructures for research support across multiple disciplines. Opportunities for synergy are expanding beyond European borders due to increasing global collaboration in research. Most RIs have been collaborating at the global level since their inception, with installations and research teams distributed around the world. Similar trends are foreseen at the e-infrastructure level, facilitated by agreements between the EU and other regions of the world. Digital Partnerships have been signed with Japan, South Korea and Singapore; specific agreements on AI and computing with the USA. What is more, discussions about the inclusion of countries outside Europe such as Canada, Australia, Japan, South Korea and Singapore in Horizon Europe are underway. A wider framework would thus be formed, enabling global collaboration on digital R&D, including e-infrastructure collaboration at a global scale.

On the connectivity side, two fibre sea cable projects, namely Far North Fibre and Polar Connect, have been initiated. These projects will establish connections to Asia via the northern route, with links to USA and Canada. Similar projects are running in other European regions, such as the BELLA-programme BELLA-programme 
https://www.bella-programme.eu , which connects to Latin America, enabling faster data transfer globally.

GREEN TRANSITION

Green transition is the emerging trend in all digital services. Innovations to diminish the carbon footprint and reduce electricity consumption are being devised both at the data centre level, often yielding the largest benefits, and within chip design.

Power Usage Efficiency (PUE) indicates the amount of electricity required for HPC systems cooling, which ideally is as close to 1 as possible. A value of 1 means that all electricity powers the system rather than overheads such as cooling. Advanced methods such as free cooling with outside air (especially in cold regions) can save energy and cost with lower carbon footprint. Another evolving trend, particularly in liquid-cooled systems, involves utilising waste heat for district heating, lowering both the carbon footprint and expenses.

At every level, there’s a growing emphasis on environmental factors. The impact on the environment depends on the type of electricity used (renewable or non-renewable), how efficiently it is used (high or low PUE) and additional emissions reduction strategies (e.g. waste heat utilisation). To achieve as low a carbon footprint as possible is a major criterion in new data centre projects, as well as in other activities. Beyond the carbon footprint, focus needs to shift on carbon handprint, acknowledging actions taken for a positive climate impact. Projects like DestinE, aiming to study and develop actions against climate change, are good examples of efforts to improve the carbon handprint.