Estonia digital innovation in high-performance computing

Estonia is considered one of the most advanced digital societies in the world. For further innovation, a carefully developed e-infrastructure must be a central pillar

When Estonia’s first computer cluster became available in 2008, its capacity was relatively modest by global standards – 42 Sun Fire nodes with eight cores each, 32 GB of RAM, 10 TB of shared storage, and only a small team responsible for maintaining the system. However, this marked only the beginning.

A few years later, the computing clusters of Estonian universities were merged into a shared national infrastructure. This development led to the creation of the consortium called the Estonian Scientific Computing Infrastructure (ETAIS). This consortium represents a nationwide e-infrastructure that provides cloud services, classical high-performance computing (HPC), and data repository resources. Its aim is to enhance the competitiveness of Estonian research in computing- and data-intensive disciplines by providing access to a modern scientific computing environment. All ETAIS services are also open to businesses and the public sector.

Today, ETAIS has grown to become the largest provider of computing power and artificial intelligence (AI) services in the Baltics. This rapid development can be regarded as a true success story. The Innovation Platform spoke with Ülar Allas, Head of the Estonian National HPC Competence Centre, and Aadi Tegova, Public Relations Specialist, about ETAIS’ success story.

In just a few years, ETAIS has become the largest HPC service provider in the entire Baltic region. What is the secret of your success?

Ülar Allas (UA): At ETAIS, our focus is not solely on hardware capacity, but also on how that capacity is organised and integrated. Rather than relying on a single supercomputer, ETAIS connects multiple computing centres into a unified system. This provides users with a single entry point to a significantly larger pool of computing resources.

ETAIS is also closely linked to major European initiatives such as the EuroHPC Joint Undertaking. Through this collaboration, our users can access world-class systems such as the LUMI supercomputer, greatly expanding available capacity beyond national limits.

In addition to infrastructure, ETAIS provides comprehensive support services, including technical assistance, training, and consultancy. A key priority for us is lowering the barrier to entry, ensuring that both businesses and researchers – regardless of their prior experience – can make effective use of HPC. Our aim is to bring together local and international computing resources within a single, user-friendly ecosystem.

Aadi Tegova (AT): ETAIS was established in 2008 in response to a practical need: Estonian researchers required greater computing capacity as well as skilled specialists to manage increasingly complex systems. Our rapid development has been driven both by the ambition to build a strong domestic infrastructure and by the continuous growth in demand.

Estonian users often prefer ETAIS to commercial cloud service providers for several reasons. Cost is an important factor: as ETAIS is supported by the state, we are able to offer services at a more competitive price. Security is another key consideration. We place strong emphasis on data protection and provide users with full transparency regarding how their data is stored and managed.

Which projects implemented on the infrastructure of ETAIS do you consider to be Estonia’s biggest success stories?

ÜA: Estonia has world-class researchers in data-intensive fields such as natural language processing, bioinformatics, genetics, personalised medicine, chemistry, big data, and AI. One particularly effective HPC user is the Chair of Natural Language Processing at the University of Tartu. Their work was recently recognised at the national level for the development of the translation engine Neurotõlge, which enables translation into small and endangered Finno-Ugric languages. This tool allows researchers to access and analyse materials that would otherwise remain incomprehensible to them, opening up new opportunities to study the history of languages and regions without requiring prior knowledge of the respective language. Moreover, as many Finno-Ugric languages are no longer widely spoken, such technologies play an important role in their preservation. The LUMI supercomputer was used to train the underlying language model, as the project required extremely large-scale data processing.

AT: Estonian language technologists are also developing a translation system for the hearing impaired that operates in Estonian sign language. This involves training artificial neural networks on hundreds of hours of annotated video material. As a result, Estonia is set to become one of the first countries in the world to develop a translation solution for a small national sign language. This is an important step towards a more inclusive society where technology better addresses the needs of all people.

ÜA: Estonian geneticists are another key success story. They are highly experienced users of HPC, as cutting-edge life sciences research increasingly depends on advanced computing capabilities. One prominent example is the Centre for Data-Driven Personalised Medicine, which brings together expertise in genomics, bioinformatics, data science, clinical medicine, public health, and socio-economic analysis. Their work focuses on applying AI and real-world data – from registries, biobanks, and electronic health records – to identify high-risk individuals at an early stage and develop targeted interventions to help prevent or delay disease.

These use cases describe what HPC can achieve in science. But how widespread is the use of supercomputers among Estonian companies?

ÜA: Despite its image as a highly digital nation, Estonia’s industrial base is still largely composed of small and medium-sized enterprises operating in low- and medium-technology sectors with relatively modest R&D intensity. However, this is gradually changing, and interest in HPC is clearly on the rise.

There are already several Estonian companies whose success is built on the effective use of HPC. One notable example is Better Medicine, a medical technology company that applies artificial intelligence in radiology to detect oncological findings. The company faces the complex challenge of training advanced AI models for lesion detection, measurement, and segmentation based on computed tomography scans. This requires substantial computational power to process large datasets, optimise model performance, and reduce training times. HPC resources significantly improve both the speed and accuracy of these models, enabling faster iteration cycles and continuous performance improvements. They also support scalability, allowing the company to efficiently handle growing data volumes as it expands its solutions to broader clinical applications.

AT: Another remarkable example is the Formula Student Team Tallinn. Each year, the team designs, builds, and competes with a high-performance formula-style racing car using cutting-edge technologies. A key aspect of their work is aerodynamic development, which relies on computational fluid dynamics simulations. Running a full 3D aerodynamic simulation on a standard computer would take so long as to be practically unfeasible. HPC capabilities dramatically accelerate this process, allowing simulations to run much faster and enabling multiple simulations to be carried out simultaneously. This significantly enhances both development speed and overall performance.

Estonia belongs to the consortia of both the LUMI supercomputer and the LUMI AI factory. What do these memberships mean for your country?

ÜA: Estonia’s membership of the LUMI supercomputer consortium means that the country is both a co-owner and a privileged user of one of the most powerful computing systems in the world. As a consortium member, Estonia gains access to state-of-the-art infrastructure that would otherwise be unaffordable on its own. This strengthens the country’s role in European digital sovereignty and competitiveness and supports its positioning as a digitally advanced and AI-focused nation. In practical terms, it moves Estonia beyond being merely a ‘digital country’ to becoming part of the high-performance computing core of Europe’s innovation ecosystem.

AT: Researchers at Estonian universities and research institutes can access LUMI computing resources either through Estonia’s national allocation or by applying for EU-wide competitive access. The same applies to companies, as up to 20% of LUMI’s resources are reserved for industry and SMEs.

ÜA: Estonia’s participation in the LUMI AI Factory – an AI-focused extension of the LUMI ecosystem – adds an additional layer of value. While the original LUMI consortium provides raw computing power, the AI Factory focuses on transforming that power into practical AI capability. For Estonian businesses, this enables faster, more cost-effective, and more accessible AI development. For researchers, it provides broader and more user-friendly access to advanced AI workflows. As a result, more researchers are able to make effective use of supercomputing for AI applications. At the same time, access is becoming increasingly democratised, with advanced AI capabilities no longer limited to large technology companies.

European countries are trying to encourage companies to more actively explore and use HPC opportunities in order to stimulate the economy. What is ETAIS’s experience with engaging companies?

ÜA: If a country fails to innovate, the competitiveness of its economy inevitably declines. Expanding the use of HPC technologies in business is therefore essential for driving innovation. Enhancing the competitiveness of the Estonian economy is, accordingly, one of the core objectives of ETAIS. HPC acts as a force multiplier: it accelerates innovation, reduces unnecessary costs, and provides companies with a strategic advantage.

AT: Our representatives take part in all major IT conferences and business festivals in Estonia to engage potential users. We deliver presentations, interact with participants, and work continuously to raise awareness of HPC. It is an ongoing effort. In addition, we publish success stories of Estonian companies that have adopted HPC, using these examples to inspire new users.

ÜA: Interest in ETAIS services has also been strengthened through demonstration projects carried out in collaboration with AIRE, the Estonian centre within the network of European Digital Innovation Hubs. These projects follow a ‘test-before-invest’ approach, enabling manufacturing companies to experiment with AI and HPC technologies in support of their R&D activities. AIRE’s team provides extensive support and funding for selected projects, while ETAIS contributes by offering access to HPC resources.

Interestingly, a significant proportion of Estonian companies using HPC have been founded by former scientists who first encountered supercomputers during their research careers.

ETAIS consortium partners participate in several international IT development projects. Which of these do you consider the most important from an HPC perspective?

ÜA: From an HPC perspective, I would highlight our participation in the development of the EuroHPC Federation Platform. This initiative is particularly significant because it represents a shift from isolated national supercomputers towards a federated, pan-European HPC ecosystem. Its importance lies less in new hardware and more in how HPC resources are accessed, integrated, and utilised at scale.

Traditionally, systems such as LUMI, Leonardo, MareNostrum and other European supercomputers operate under different policies, software stacks and access procedures. The upcoming platform will federate these heterogeneous systems into a single logical environment, providing a unified access point to EuroHPC resources. At present, workflows are often tied to a single supercomputer, but the Federation Platform will help overcome this limitation by enabling seamless data movement and workflow execution across multiple systems.

The broader goal is to transform HPC from a specialised research tool into a widely used infrastructure, thereby strengthening Europe’s competitiveness. It is far more than a simple operational upgrade.

AT: The University of Tartu is one of six consortium partners developing the EuroHPC Federation Platform, alongside CSC, IT4Innovations, Ghent University, GÉANT and NORDUnet. Specifically, the University of Tartu is responsible for the Allocations component – a federated resource allocation, monitoring and management service that enables users to manage projects and track computing resources across different EuroHPC hosting entities.

You mentioned that ETAIS pays a lot of attention to data protection and cybersecurity. Which international projects related to cybersecurity does ETAIS participate in, and what is the importance of these projects from an HPC perspective?

AT: We are currently involved in two cybersecurity projects. The first one is SOCCER – Security Operations Centre in the Central and Eastern European Region. Its primary objective is to establish and interconnect Security Operations Centres across universities, while enhancing the monitoring, detection, and response to cyber threats. Although it is not an HPC project per se, it clearly intersects with the HPC ecosystem, as it involves HPC centres as stakeholders and contributes to securing HPC environments. The capabilities of Security Operations Centres directly benefit HPC infrastructures by enabling effective monitoring of HPC clusters and the detection of anomalies in large-scale computing environments. This is an area where ETAIS’ specialists can contribute their expertise within a broader cybersecurity framework.

The second project is the Cybersecurity Certification and Assessment Tools initiative. This project focuses on developing a cybersecurity certification framework that supports the assessment and certification of HPC environments, alongside other types of infrastructure.

Estonia has often been described as ‘e-Estonia’, reflecting the government’s long-standing priority of facilitating citizens’ interactions with the state through ICT solutions. What are the latest national trends in this area, and how do you see ETAIS’ role in these developments?

AT: Estonia is currently transitioning from an ‘e-state’ to an ‘AI-driven state.’ While the earlier focus was on digitalising services, attention is now shifting towards making these services more proactive and intelligent. To support this transition, the Eesti.ai programme has been launched, with the aim of bringing the practical use of artificial intelligence into everyday life – including healthcare, education, business, and public services. For most citizens, the earliest tangible impact of these AI initiatives is likely to be seen in the healthcare sector.

ETAIS effectively serves as the ‘engine’ enabling AI development in Estonia. Implementing the Eesti.ai strategy requires significant computational capacity. ETAIS manages Estonia’s domestic HPC infrastructure and provides access to the LUMI supercomputer, making it a key partner in this ecosystem. Without ETAIS, Estonia would lack the capacity to train the advanced AI models required. It is estimated that by 2030, the country’s aggregate demand will reach approximately 11,600 GPUs.

ÜA: It is crucial that the infrastructure supporting these developments is located within Estonia. Firstly, public sector institutions and research organisations often handle sensitive data or state secrets that cannot be processed abroad due to legal and security constraints. Secondly, latency is a critical factor: many data-intensive public and private services require real-time performance, and geographically distant computing resources can introduce network delays, resulting in slower and less reliable services.

Are there enough HPC and data science specialists in Estonia to meet the growing demand?

AT: As time goes on, HPC skills are becoming increasingly important. It is therefore clear that HPC should already be integrated into IT and engineering education at the undergraduate level. At the University of Tartu, such courses were introduced several years ago. The most popular of these is the general High-Performance Computing course. Its aim is to provide students with a solid understanding of HPC, how it works, and its best practices and use cases, enabling efficient computation with large and complex datasets. The course is open to students at all levels of study and has even attracted participants from other universities.

We also offer a System Administration course, which takes a practical approach by placing students in the role of a system administrator. During the course, students take on tasks that a typical system administrator encounters on a daily basis, while gaining hands-on experience in systems, integration, monitoring, and DevOps.

ÜA: In addition, a new Quantum Computing Micro-Degree programme was launched at the University of Tartu this year. This programme provides a comprehensive and practical introduction, covering topics from the fundamental principles of quantum information to the development of simple quantum programs using the Python programming language.
We believe this initiative represents an important milestone in shaping Estonia’s quantum technology ecosystem and supports the development of the skills needed ahead of the wider adoption of quantum technologies.

Here’s one final question for both of you: if you had to dispel one myth about HPC, what would it be?

AT: HPC is often misunderstood, even within the technology sector. One common myth is that it is simply a faster version of a desktop computer. In reality, HPC is a coordinated network of many machines, not just a single, more powerful system.

Moreover, HPC is not only about speed – it is about scale and parallelism. These systems often consist of thousands of nodes working together, using specialised architectures and high-performance interconnects to solve problems that a single machine simply cannot handle.

ÜA: I have heard experts suggest that more compute cycles are always better. In practice, this is not the case. Hardware without the necessary expertise is an extremely effective way to waste resources. HPC is a tool for acceleration and amplification – and it can amplify both expertise and poor decision-making alike.

This article will feature in our upcoming High-Performance Computing Special Focus Publication.