Greece’s DAEDALUS Supercomputer Gives Pharos AI Factory a New Compute Backbone

Greece has unveiled DAEDALUS, a powerful supercomputer in Lavrion near Athens that will serve as the foundation for the country's AI innovation hub, the Pharos AI Factory. This move aligns with the European Union's broader EuroHPC initiative to broaden access to advanced AI resources for researchers, startups, SMEs, and public institutions.

DAEDALUS boasts an impressive 85.69 petaflops of sustained performance and a peak performance of 115 petaflops. Built by Hewlett Packard Enterprise utilizing NVIDIA GH200 Grace Hopper technologies, the supercomputer is operated by Greece's National Infrastructures for Research and Technology (GRNET).

For businesses, DAEDALUS' significance extends beyond its raw computational power. By providing a domestic high-performance computing backbone, DAEDALUS enables Pharos to foster AI innovation domestically. The success of any AI project using DAEDALUS will depend on various factors, including access programs, support services, and project requirements that Pharos may offer.

DAEDALUS and Pharos serve distinct yet interconnected roles. DAEDALUS is the physical computing system, while Pharos is the Greek AI Factory initiative that aims to use this capacity to accelerate AI research and innovation. It also connects Greece to the wider European AI Factory network and related EU initiatives.

However, high-performance hardware alone does not guarantee a smooth AI project deployment. Companies need a viable use case, suitable data, technical expertise, and a pathway to integrate resulting models or workflows into their day-to-day operations. DAEDALUS can support demanding research and AI workloads that would be challenging to run on conventional business infrastructure, but it is not a one-size-fits-all solution.

The supercomputer is funded through a EUR 36 million acquisition project, with the EuroHPC Joint Undertaking covering approximately 35% of the cost. The remaining funding comes from Greece's National Recovery and Resilience Plan, Greece 2.0, and NextGenerationEU. The Pharos AI Factory has a total funding of EUR 30 million, including about EUR 15 million in EU contributions.

While the computing capacity of DAEDALUS can significantly benefit specific AI initiatives, it may not be necessary for all business use cases. Many projects, such as integrating existing AI services into customer support, sales, or internal knowledge workflows, might not require this level of infrastructure. However, for organizations with truly compute-intensive AI problems, Pharos could be a valuable resource.

At this stage, businesses should avoid treating the announcement as a guarantee of immediate, low-cost compute access. Practical questions remain regarding access calls, eligibility rules, support programs, technical preparation, data controls, and the necessity of high-performance computing versus managed cloud tools or existing AI services.

Organizations with well-defined, compute-intensive AI problems should closely follow Pharos developments and assess whether its programmes align with their needs. Those who do not require high-performance computing should still consider how a successful pilot could be integrated into a repeatable business process.