Advanced computing in Europe: what the EU strategy includes and how to check its progress

The expression “advanced computing” can refer to different resources and capabilities; it should not automatically be treated as a synonym for supercomputing. On the European Commission’s digital policy page, the topic is linked to computing technologies and high-performance computing, and is presented as part of the ambition for Europe to play a leading role in supercomputing during the Digital Decade. That wording describes a political priority, not a measure of success already achieved. (European Commission) The distinction matters because policy language often describes a direction of travel, while evidence about achievement has to come from information about what has actually been delivered and used.

To read that priority precisely, it helps to distinguish three levels. The first is technical capacity: machines and computing resources able to handle demanding workloads. The second is access and organisation: who can use those resources, under what conditions and with what support. The third is effects—for example, scientific advances or industrial applications—which cannot be inferred simply from the existence of infrastructure. This division is an editorial framework for reading the material, not an exhaustive legal definition of the EU’s concept. It is intended to clarify the different questions that can otherwise become conflated when broad terms are used in policy communications.

The three levels prompt different questions. Technical capacity concerns which resources exist and which tasks they can address; access concerns the practical conditions under which users can make use of them; and effects concern what follows from that use. Keeping these questions separate prevents an infrastructure figure from being interpreted as proof of broad access or of benefits that have already materialised. It also makes it easier to identify what information is missing before reaching a general conclusion about the policy. A statement that a system exists, for instance, does not on its own tell us who can use it, how frequently it is used or what outcomes can be attributed to that use.

The Commission presents supercomputing as a central part of the account of advanced computing. In this context, a system’s performance can be expressed in floating-point operations per second, known as FLOPS. The Commission’s explanatory page describes the term and compares it with the capacity of everyday devices; that explanation helps readers understand the unit, but it cannot by itself support a rigorous comparison between different facilities or workloads. (European Commission) A performance figure is meaningful only when readers can tell what system and what measurement it refers to. Understanding the unit is a useful first step, not a substitute for understanding the context in which the figure was produced.

A FLOPS figure offers a reference for operations, but how it should be interpreted depends on which system and which measurement are being described. The value of a number therefore does not lie solely in its seeming size: readers need to know which facility it relates to and what comparison is being made. The explanatory description is useful for becoming familiar with the concept; to judge different systems, however, data are needed that place the figures in a comparable context. This means that a single number should not be detached from its measurement conditions or used as a universal verdict on capability.

EuroHPC JU—the European High Performance Computing Joint Undertaking—offers another point of observation: its communication about five years of activity describes a network of supercomputers, artificial intelligence factories and quantum computers. This is an institutional description of the areas the organisation highlights, not an independent assessment of their performance or proof that every component has produced the same benefits. (EuroHPC JU) The list gives information about the ecosystem’s scope as the organisation presents it; on its own, it does not provide a common measure for comparing each component or its particular results. That distinction is important when moving from a description of components to any claim about the performance of the ecosystem as a whole.

It is therefore worth avoiding a list of technologies presented as though they all had the same maturity or function. Classical supercomputing, quantum computing and computing resources associated with artificial intelligence are related areas in EuroHPC’s strategy and communications, but that relationship does not imply technical equivalence. A useful comparison would need to specify the task, system, period and metric being used. Without those details, labels such as “more powerful” or “more advanced” can conceal important differences between objectives and workloads. One resource might be designed for one kind of task while another serves a different purpose; a broad label cannot establish how they compare for a particular use.

An announcement, an award of funding, the installation of a machine and its actual use represent different stages. Funding or a call for proposals tells readers about an opportunity or support mechanism; it does not automatically show that equipment has been installed, is available to users or has generated results. Similarly, mentioning a network of systems does not by itself reveal its utilisation level, service continuity or distribution among countries and user communities. A report about an early phase should therefore not be summarised as if it described the final state of a resource. Keeping the stages distinct makes it possible to describe progress accurately without turning a planned activity into a completed one.

To check the status of a resource, readers should look for documents that answer specific questions: Which system or service is being described? Which organisation operates it? On what date is its status reported? Does the document discuss selection, construction, commissioning or access? What conditions limit use? EuroHPC’s key documents page is an entry point to materials from the organisation; the next step is to identify the specific document and check its date and scope rather than relying only on a press release. (EuroHPC JU) Checking the date also helps prevent information about an earlier stage from being presented as a description of the current situation. Scope, in turn, helps establish whether a document concerns a specific resource or a broader initiative.

A practical scale for classifying evidence

• Stated priority: appears in a strategy, communication or explanatory page.
• Planned or awarded support: documented in a call, decision or funding instrument.
• Deployed capacity: an identifiable source reports installation, commissioning or availability.
• Use and results: information is available about users, work carried out or attributable results, with a clear method and period.

These stages are not interchangeable. The evidence for one initiative may be strong at one stage and incomplete at another. Presenting them separately makes it possible to describe progress without turning an announced budget into an operational machine or an operational machine into demonstrated impact. The scale is not intended to assess the quality of each initiative; it helps readers identify the type of claim supported by each document and where the conclusions that can be drawn from it end. One source may document financial support well without yet providing information about use, while another may describe activity without measuring its aggregate effects. Recording those limits makes an explanation more precise, not less informative. To assess results, it is also useful not to skip stages in the account: first identify what was funded or deployed, then what use is documented, and finally what evidence connects that use to a result. If one of those links is missing, the conclusion should retain that uncertainty rather than fill the gap with assumptions.

The question “Is European advanced computing progressing?” has no single answer unless “progressing” is defined. It might mean increasing the availability of resources, widening access, improving technical performance, supporting research or producing economic and social effects. Each interpretation calls for different indicators and sources. Counting communications or initiatives is not enough, by itself, to answer any of these questions conclusively. A narrower question—for example, whether a resource is available or what kind of activity has been reported—can be answered with more specific evidence, provided its scope is stated. This is not simply a matter of choosing a more cautious tone: the evidence needed to establish availability is different from that needed to establish impact.

A transparent assessment should document, at a minimum, the unit of analysis, period, source and calculation method. If utilisation is reported, available capacity must be distinguished from capacity actually consumed. If scientific publications are described, readers need to know what link is being made to the funded resource. If an industrial benefit is attributed, it is useful to identify who measured it and what comparison supports the attribution. These checks do not deny achievements; they prevent claims from being given a greater scope than the evidence permits. They also make it possible to understand what an indicator represents—and what it does not represent—instead of treating it as a universal measure of progress.

EuroHPC has reported user awards at its 2025 user days, including recognition of user contributions and work. This can be an indication of activity within the ecosystem, but an award is not a representative indicator of total utilisation, average productivity or aggregate impact. Broader data and an explicit methodology would be needed to support conclusions of that kind. (EuroHPC JU) Evidence of activity is relevant, but it is not, by itself, an impact assessment. The distinction allows the reported recognition to be described accurately without asking it to answer questions it was not designed to measure.

When reading a claim about advanced computing, readers can apply this short sequence:

1. Locate the primary source: prefer the responsible institution’s document over a third-party summary.
2. Identify the stage: distinguish an objective, funding, construction, availability, use and results.
3. Set the date and scope: check when the information was published and which system, programme or territory it concerns.
4. Look for the indicator and its method: verify what is counted, over what period and against what comparison.
5. Separate the data from the conclusion: an observed figure does not automatically prove causality or represent the entire programme.

This sequence also helps reveal when a source does not answer the question being asked. If a document confirms a priority but says nothing about availability, it cannot be used to prove that a system is already operational. If it reports activity without a method for measuring effects, that activity can be described, but it should not be turned into an estimate of impact. The wording of the conclusion should correspond to the type and scope of the evidence found. A careful account can still be informative: it can state what is known, identify who reported it and explain what remains unestablished.

Institutional communications are valuable for understanding priorities, structure and announcements, but they should be read with their purpose in mind. A policy page summarises objectives; a regulatory document may define roles and rules; an activity report may describe achievements. These are not interchangeable sources. When a claim appears only in an institutional communication, it is reasonable to attribute it to that institution and avoid presenting it as independently confirmed. Attribution makes it possible to report what the organisation says without implying that a different source has verified it. It also helps readers distinguish an organisation’s account of its work from an external assessment of that work.

What the sources consulted allow us to conclude

The material reviewed supports saying that the EU links high-performance computing to its digital ambition, and that EuroHPC presents supercomputers, AI factories and quantum computers in its institutional account as parts of its ecosystem. It also provides routes for locating further documentation. It is not, however, sufficient to quantify deployment comprehensively, compare performance across systems or attribute aggregate scientific and economic impacts. This qualification sets the limits of the conclusion; it does not invalidate the existence of documented priorities and activity. The available sources can establish how the institutions frame the subject and point readers towards additional records, while leaving other questions open.

The conclusion therefore depends on keeping the political objective, institutional description and evidence about results separate. The information consulted makes it possible to identify priorities and reported activity, but it is not a substitute for a comprehensive assessment based on comparable data and explicit methods. To extend the analysis, readers can consult the specific EuroHPC documents and verify which aspect each one covers, without assuming that a single page answers every question about infrastructure, access and effects. A conclusion proportionate to the available evidence can acknowledge what is documented while making clear where further information would be needed.