Greece's AI Strategy: How DAEDALUS and PHAROS Could Shape Its Future

On October 19, 2026, Greece inaugurated DAEDALUS, a new supercomputing facility intended to revolutionize its approach to artificial intelligence. The system, procured through a contract with Hewlett Packard Enterprise, offers sustained performance of 85.69 petaflops and peak performance of 115 petaflops. It utilizes HPE infrastructure and over 2,000 NVIDIA GH200 Grace Hopper Superchips.

The inauguration of DAEDALUS signifies a significant milestone in Greece's AI strategy, as it provides access to computing resources that were previously inaccessible. However, its strategic significance will be determined by the projects that can reliably utilize the system and the institutions capable of retaining the resulting expertise.

Greece focuses on specific priority domains such as health, culture, language, and sustainability, which will guide the assessment of outcomes and prevent treating every new AI activity as evidence of national progress.

The operation of DAEDALUS is managed by GRNET within an international consortium, with access managed jointly by EuroHPC and the DAEDALUS consortium based on investment. The system's hardware and software are supplied by HPE and NVIDIA, respectively, which brings substantial capabilities but also ties parts of the system's operation to these suppliers. Therefore, while Greece has made significant investments, it remains dependent on external technological providers.

To evaluate the next phase of Greece's AI strategy, four indicators are proposed: published access and utilization records, completed project results, sustained usage by public bodies and firms, and staff retention and maintained software. These indicators will help determine whether the system's capabilities are being effectively utilized and whether they are leading to tangible benefits in specific domains such as health, cultural preservation, language processing, or sustainability.

The policy response should focus on ensuring continued support, staff competence, and clear options for future workloads, rather than simply relying on the aggregate performance of the supercomputer.