
The European Union has launched a tender to establish up to seven large-scale artificial intelligence (AI) computing centres across Europe, in an effort to strengthen the
bloc's technological independence and accelerate its ambition to become a global AI leader.
The initiative, backed by up to €10bn in EU and national public funding, is expected to attract at least €20bn in private investment.
The planned AI Gigafactories will provide computing infrastructure for start-ups, technology companies, small and medium-sized businesses, researchers and public authorities working on advanced AI systems.
The facilities will combine high-performance AI processors with software, cloud technology, high-speed connectivity and energy-efficient data centres.
They will operate alongside Europe's existing network of 19 AI Factories, which the European Commission says will help strengthen the bloc's technological resilience and strategic autonomy.
The EU says the new infrastructure will allow advanced AI models to be developed and trained within Europe and under European rules on data protection, safety, security and ethics.
Who can apply?
Consortia bringing together private companies, public bodies, investors and other partners will be able to submit proposals.
Gigafactories can be built in a single EU member state, either at one location or across several sites. Projects can also involve several countries through distributed computing facilities.
Applications will be assessed through a competitive selection process.
Up to seven facilities
The European High Performance Computing Joint Undertaking (EuroHPC JU) and participating member states will jointly procure computing capacity from the selected facilities.
Eighteen EU countries have signed an agreement to take part in the joint procurement: Croatia, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Poland, Portugal, Slovakia, Spain and Sweden.
The programme will support up to seven Gigafactories through two development phases and two funding lots.
The first lot will fund up to four projects. Each will be eligible for up to €100m in EU funding during the first phase, followed by up to €400m in the second phase.
Each facility in this group will initially have to deploy at least as many advanced AI processors as are currently installed in Europe's most powerful AI Factory. That capacity will have to rise to at least three times the current level during the second phase.
The second lot will support up to three projects, with each eligible for up to €200m in EU funding in the first phase and a further €800m in the second.
Facilities selected under this lot will have to deploy up to twice the number of advanced AI processors currently installed in Europe's most powerful AI Factory, with capacity rising to at least four times that level in the second phase.
Participating member states will match EU funding, while the public investment is intended to encourage the much larger private investment needed to build and operate the facilities.
Focus on AI hardware
The Commission says access to advanced computing hardware is increasingly important as demand for AI grows worldwide.
The Gigafactory consortia will be able to purchase hardware during either development phase from suppliers based in Europe or other countries considered like-minded by the EU.
Some procurement may be directed towards European start-ups and scale-ups, with the aim of strengthening the continent's digital supply chain.
As part of the follow-up to the EU-US trade agreement, the European Commission has also signed letters of intent with US technology companies AMD, Nvidia and Qualcomm.
The agreements are intended to ensure that the consortia can secure access to the advanced hardware needed to build the new AI infrastructure.
The Commission says the Gigafactories will be a key part of its wider strategy to increase Europe's computing capacity and reduce its dependence on technology and infrastructure developed outside the bloc. Photo by mikemacmarketing, Wikimedia commons.
