By Teti Igoumenidi

tigoumenidi@naftemporiki.gr

With the new Daedalus supercomputer inaugurated in Lavrio, Greece is securing the computing power needed to develop advanced artificial intelligence (AI) models and reduce its reliance on foreign infrastructure.

Daedalus is the most powerful supercomputer ever installed in Greece, with a €58.9 million budget and a performance ranking of 31st worldwide. Its operation will enable the scientific and business communities to carry out demanding research projects and process large volumes of data domestically. One of its key advantages is the ability to process sensitive or confidential data without having to transfer it abroad. Research institutions, healthcare providers, public bodies, and businesses will use the infrastructure to develop AI applications locally in Greece, while adhering to security and data protection standards.

Business access to the supercomputer will be supported by Pharos AI Factory, Greece’s AI factory. Under current plans, its services are expected to become generally available to businesses in March 2027. The inauguration will be followed by a period of technical testing and system configuration, while Pharos AI Factory continues its preparations.

Start-ups and small and medium-sized enterprises (SMEs) will be offered free access to computing power, subject to the constraints imposed by EU state aid rules. Charges based on computing resource usage time are being considered for large businesses. The full range of services, access terms and pricing policy have yet to be finalised.

Daedalus has been installed in the historic building of the former power station at the Lavrio Technological and Cultural Park, operated by the National Technical University of Athens. Its measured performance reaches 85.69 petaflops, equivalent to approximately 85.69 quadrillion floating-point operations per second. In the Green500 global ranking of energy-efficient supercomputers, it ranked 23rd in June 2026.

Liquid-cooled technology

The system is based on Nvidia Grace Hopper GH200 technology and uses liquid cooling. Its management and technical operation have been assigned to the National Infrastructures for Research and Technology (GRNET), which is also responsible for implementing the project.

Hewlett Packard Enterprise undertook the supply and installation of the computing equipment, with a €41 million budget, while Mainsys carried out the electromechanical installations and data centre fit-out, with a €17.9 million budget. Both amounts include VAT. The project was funded by the Recovery and Resilience Facility and the European High Performance Computing Joint Undertaking (EuroHPC).

With a €30 million budget, Pharos AI Factory is one of the first seven AI factories selected by the European Union. It is funded 50% by EuroHPC, with the remainder covered by national funds. Its focus areas include healthcare, sustainable development, the Greek language and culture.

According to the company’s presentation, organisations will be able to receive guidance on AI adoption, support in commercialising applications, and access to collaboration networks and business acceleration programmes.

They will also have access to computing resources, secure testing environments (sandboxes), trusted runtime environments and infrastructure for deploying applications in production. The planned services also include access to AI models, curated datasets and open data.

To support the commercial exploitation of the initiative, the company Hellenic AI Factory S.A. has been established, with the Growthfund holding a 70% stake and the Ministry of Digital Governance and Artificial Intelligence holding the remaining 30%.

Planned activities include an online training programme on November 10, in cooperation with Endeavor Greece and Nvidia. The programme is expected to involve 45 selected participants, who will receive training in building AI agents, connecting them to large language models and using related tools. It will also include practical exercises in modifying an existing application.

Second supercomputer planned for Kozani

Meanwhile, GRNET is advancing plans for a second supercomputer, to be installed at the University of Western Macedonia. The organisation’s board approved the launch of a public consultation on the tender for the supply and installation of computing equipment, storage and software, with a budget of €21.57 million, including VAT.

The amount concerns this specific subproject and does not represent the total investment. GRNET is expected to seek an increase in the budget and an update to the timetable through an amendment to the project inclusion decision.

The new system will operate in a green data centre integrated into GRNET’s RE-Cloud network, with a focus on AI, cloud services and the analysis of sensitive data. The facility will feature IT equipment with a power capacity of 1.2 to 1.5 MW and a photovoltaic installation with a capacity exceeding 2 megawatt-peak (MWp).

Funding

The subproject is funded by the Just Transition Fund, and the contract implementation period is set at 12 months from the date of signature.

As Digital Governance and Artificial Intelligence Minister Dimitris Papastergiou has said, the plan is to train models on Daedalus and run them on the supercomputer in Kozani, allowing the two infrastructures to operate in a complementary manner.