Funding coalesces to unify Europe’s compute, data, and research talent into a shared infrastructure for developing and applying advanced AI models across science and HPC.
Europe is tired of renting the future from other people’s datacenters but the EU Commission’s plan for RAISE announced yesterday treats that as a solvable problem.
The goal is to create a virtual institute that brings together Europe’s compute power, data, funding, and talent into a single framework for scientists, which will have hooks into existing EuroHPC systems and open sciences data spaces with a streamlined policy framework to sit on top.
The first milestone is fixed: a €108 million pilot, launching in Copenhagen in early November marking the formal start of the program.
The Commission notes that Europe’s share of global AI compute capacity sits below 5% while the US is around 75% and China around 15%. That imbalance shows up in who trains the frontier models, who sets evaluation practice, and who turns research into reality. RAISE is the response.
RAISE tackles this in two directions at once by focusing AI in science (making new discoveries) and also science for AI to (advancing core model capabilities). The target, as they spell out, is human centric, explainable, safe, and aligned with fundamental rights, but also architected to remove fragmentation and reach critical mass.
More specifically, RAISE plugs into the EuroHPC stack and stands up AI Factories in 2025–26 that more than triple AI capacity for European users. These are AI-optimized supercomputers (versus just bigger HPC clusters) surrounded by services for algorithm development, evaluation and validation of large models, and programming environments that do not make scientists reverse engineer the machine room.
Access policy gets a new lane called “AI for Science and for Collaborative EU Projects” that lets selected EU-funded work bypass a second peer review and go to the front of the queue.
The next step up from that are “AI Gigafactories”. Those are very large training and inference facilities with energy efficient datacenters, automation across the model lifecycle, and scheduling policies that make sovereign scale real rather than rhetorical.
Horizon Europe is putting up to €.600 million on the table to underwrite dedicated access in this pilot.
On the data front, the European Open Science Cloud becomes the default lane for research, with other Common European Data Spaces, the European Health Data Space, and sectoral utilities around it. RAISE adds Data Labs inside the AI Factories that federate sources, clean and enrich datasets, generate synthetic data where needed, and provide common building blocks and compliance help.
The point, which will be clear to HPC centers who have had to stitch together a cross-institution dataset and understand the challenge, is to make the data usable at scale and you do not turn every PhD candidate into a part-time privacy lawyer. RAISE will also map the strategic scientific data gaps and finance the curation and integration work to close them.
Another part of the program focuses on talent. The pilot funds Doctoral Networks on AI in science and builds Thematic Networks of Excellence that focus talent in areas like materials and biotech while keeping methods moving across fields.
The Commission keeps the guidance loop tight through the Living Guidelines on responsible use of generative AI in research, updated regularly to address plagiarism, fabricated citations, and provenance issues that arise as models move into the authoring layer.
The Joint Research Centre is tasked to create a Scientific AI Hub that monitors and evaluates models used in strategic scientific research, aligned with the European AI Office. That hub is where “trustworthy” stops being a tagline and becomes evaluation practice and metrics.
Horizon Europe has already put €6.4 billion into AI from 2021 to 2024, with another €1.6 billion in 2025 and about nearly one billion of that focused on AI in science.
The plan is to double AI funding by 2028, including AI in science, and to explicitly finance scientific laboratory automation and scientific foundation models that are kept current and shared across labs.
This is how you prevent promising prototypes from dying when a grant ends. You fund platforms that become community baselines and you keep them updated.
€108 million — RAISE pilot phase launching 3–4 November 2025 in Copenhagen under the Danish Presidency.
Up to €600 million — Dedicated access for scientists and startups to AI Gigafactories and AI Factories under Horizon Europe.
€6.4 billion — Horizon Europe investment in AI from 2021 to 2024.
€1.6 billion — Additional AI funding in the 2025 Horizon Europe Work Programme.
€700 million — Portion of the 2025 funding specifically for AI in science.
Commitment to double total AI and AI-in-science funding by 2028.
European Regional Development Fund — Support for national and regional AI and HPC infrastructure projects complementing RAISE.
Cohesion Policy funding — Additional dual-use AI investments through the Mid-Term Review and STEP mechanism.
€150 million — EIC investment in AI startups in 2024.
€400 million — Cumulative EIC AI investments from 2021 to 2024.
Ongoing MSCA funding — Doctoral Networks supporting AI research and training integrated into RAISE.
Next Multiannual Financial Framework — Dedicated RAISE budget for sustained expansion beyond 2028.
The European Commission anchors RAISE. Member States co-own it through European Research Area governance, with the future AI Board connected on policy.
EuroHPC JU is the operational partner for compute. EOSC and the sector data spaces are the substrate for research data. The European AI Office aligns rules and guidance. The Joint Research Centre runs evaluation. The European Innovation Council and the European Institute of Innovation and Technology connect startups and scaleups, while the Apply AI Alliance provides a broader industry and public sector forum.
Regulatory sandboxes that are mandatory in every Member State under the AI Act bring the research community into design and operation so that pre-market work is not slowed by guesswork. There is also a pledging campaign pointed at private and philanthropic actors to pull more capital into AI in science.
The Thematic Networks of Excellence organize around disciplines with heavy leverage and the EU has decided that materials science is a flagship.
The network aligns labs deploying AI for analysis, discovery, and testing, with standards like the coming Materials Commons and access to dedicated AI infrastructure and data management. The goal is to couple AI and automated laboratories so design, synthesis, and validation become a closed loop. That makes safer low-carbon materials, quantum-class components, and better batteries a function of tokens and robots rather than multi-year cycles.
The biotech track does the same for biology, where biological foundation models can predict structures, analyze complex systems, and generate new biological agents
Both tracks are tied into broader policy moves like an Advanced Materials Act in 2026 and a forthcoming Biotech Act, which gives the technical work a path to industrial scale.
The calendar is set. In Q4 2025 the pilot launches in Copenhagen. That start includes the coordination secretariat, the advisory board, the first Doctoral Networks and Thematic Networks, and the reserved AI compute allocations flowing through the new access lane.
In 2026 the Data Labs design and integration with EOSC and other spaces is delivered and a monitoring framework begins tracking uptake and impact.
In 2027 the AI evaluation hub comes online.
By 2028 the target is full deployment with doubled annual AI investment, normalized Gigafactory access for scientists and startups, and national programs aligned enough that a pan-European network behaves like a single system instead of a patchwork.
The Commission will report on implementation by the end of 2027.
About ninety percent of public R&D in the EU is funded nationally, which means RAISE only works if Member States align enough to avoid fragmentation killing momentum.
If Copenhagen is the point where promises become allocations, queues, and running jobs, then the metric that matters will look something like, how many scientists got priority access to AI-optimized machines or how many datasets crossed institutional boundaries through Data Labs with provenance intact. Or even questions like how many foundation models and automated labs shipped as shared infrastructure rather than one-off papers.
Europe wants to stop asking Big Tech for permission to do cutting-edge science at scale. RAISE is the plan to allow that as a choice.
