AI Foundation models in science (GenAI4EU) (RIA)
Quick Facts
Email me updates on this grant
Get notified about:
- Deadline changes
- New FAQs & guidance
- Call reopened
- Q&A webinars
We'll only email you about this specific grant. Unsubscribe anytime. No spam.
Ready to Apply?
Get a personalized assessment of your eligibility and application strategy
See in 5 min if you're eligible for AI Foundation models in science (GenAI4EU) (RIA) offering max €45.0M funding💰 Funding Details
Funding Description – AI Foundation Models in Science (GenAI4EU)
Overview
* Call Identifier: HORIZON-CL4-INDUSTRY-2025-01-DIGITAL-61
* Action Type: HORIZON-RIA (Lump-Sum)
* Maximum EU Contribution per Project: €45 million (no predefined minimum)
* Submission Scheme: Single-stage; opens 22 May 2025, closes 23 Sept 2025 – 17:00 CET
* Scientific Domains (choose ONE):
1. Materials Science (A)
2. Climate-Change Science (B)
3. Environmental Pollution Science (C)
4. Agricultural Science (D)
What the Grant Funds
* Design, training and open-source release of AI foundation models (not limited to GenAI) tailored to the chosen domain.
* Data curation (FAIR-compliant), multimodal dataset acquisition/creation, semantic annotation & metadata standardisation.
* Access to and use of EuroHPC / HPC / cloud compute for model pre-training, fine-tuning and benchmarking.
* Multidisciplinary research addressing at least four downstream scientific use-cases and demonstrating model adaptability.
* Integration of domain knowledge (e.g., ontologies, knowledge graphs, RDF).
* Risk-assessment & mitigation of model misuse; legal & ethical compliance (SSH participation).
* Community-building, dissemination, training material, long-term maintenance plan and collaboration with EOSC & other EU infrastructures.
Lump-Sum Specifics
* The lump sum covers 100 % of eligible project costs; payment is triggered by successful completion of agreed work packages & deliverables – no cost reconciliation at reporting stage.
* Applicants must provide a realistic cost breakdown (person-months, subcontracting, equipment, travel, etc.) per work package in the proposal.
Eligibility Snapshot
* Consortium: Minimum 3 independent legal entities from 3 different Member States/Associated Countries (MS/AC).
* Participation from third-country organisations possible (with own funding unless automatically eligible).
* Universities, RTOs, large industry, SMEs, public bodies & NGOs are all eligible.
* Proposals must commit to open access of model weights, code and – where legally possible – training data.
* Projects should foresee gender equality plans for public-sector entities (mandatory HEU requirement), although sex-gender dimension in R&I content is optional for this topic.
Special Award Provision
To ensure a balanced portfolio, the Commission will fund at least 2 projects in domain A and 1 project in each of domains B, C, D, provided proposals pass all thresholds.
📊 At a Glance
Get Grant Updates
Get notified about:
- Deadline changes
- New FAQs & guidance
- Call reopened
- Q&A webinars
We'll only email you about this specific grant. Unsubscribe anytime. No spam.
🇪🇺 Strategic Advantages
EU-Wide Advantages & Opportunities – "AI Foundation Models in Science (GenAI4EU)"
1. Strategic EU-Level Added Value
• Critical-mass funding – Horizon Europe can mobilise budgets (EUR 8–12 m per project expected) that are seldom reachable for single Member States, enabling billion-parameter models and truly multimodal datasets.
• Balanced portfolio rule – Guaranteed support for at least 2 materials, 1 climate, 1 pollution and 1 agriculture project multiplies the chance for each domain community across Europe.
• Lump-sum grant – Simplified financial management lowers administrative burden, particularly benefiting SMEs, universities and RTOs in cohesion regions.
2. Cross-Border Data Assets & Repository Synergies
• Pan-European FAIR data troves (e.g. EOSC, Copernicus, EMODnet, ELIXIR) become instantly exploitable for training; no single country hosts all modalities required.
• Legal interoperability under the Data Governance Act and upcoming EU Data Act allows secure cross-border data pooling, vital for rare PFAS measurements or high-resolution agri-sensor streams.
• Harmonised ontologies – projects can shape EU-wide reference taxonomies (e.g. INSPIRE, Agri semantics), reducing future fragmentation.
3. Federated Compute & EuroHPC Privileges
• Priority access to EuroHPC JU petascale/exascale machines via EPICURE support service; training runs costing >€3 M on commercial clouds can be executed at marginal cost.
• Energy-efficient locations (Nordics, Iberia) align with the Green Deal, lowering the carbon footprint of model training by up to 60 % compared with US-based clouds.
• Cross-site federation – exploit GPU clusters in multiple Member States while complying with GDPR through on-prem inference nodes.
4. Standards, Governance & Trust
• Early compliance with EU AI Act high-risk rules; joint development of model cards, Data Protection Impact Assessments and red-teaming protocols sets EU industry benchmark.
• Open-source mandate dovetails with European Open Science Cloud policies, fostering reproducibility and lowering access barriers for Widening countries.
• SSH integration helps anticipate societal acceptance, greenwashing risks in climate models or dual-use in agri-biotech, reinforcing Responsible Research & Innovation.
5. Talent Pool & Interdisciplinary Collaboration
• Pan-European doctoral & post-doc mobility (Marie-Curie, EIT Digital) feeds directly into consortia, mitigating skills shortages in smaller Member States.
• Domain–AI matchmaking – Europe hosts world-class institutes in each target field (e.g. Max Planck, INRAE, Helmholtz, CNR, Imperial-EU nodes) enabling genuine interdisciplinarity that a national project could not assemble.
6. Industrial Competitiveness & Open Strategic Autonomy
• Materials & semiconductor sovereignty – EU-trained atomistic models shorten discovery cycles for critical raw-material substitutes, reducing Chinese/US dependence.
• Agri-tech & food security – models trained on diverse pedo-climatic zones (Mediterranean droughts to Nordic soils) give European agribusiness a unique decision-support edge.
• Climate services market – Copernicus downstream SMEs can integrate foundation models into commercial offerings, estimated €1.4 bn market by 2030.
7. Policy Alignment & Downstream Funding Pathways
• Feeds the GenAI4EU flagship under the 2024 AI Innovation Package, increasing prospects for scale-up funding via DEP, Digital Europe Test & Experiment Facilities.
• Synergy with Missions & Partnerships – outputs can plug into Mission ‘Adaptation to Climate Change’, Batteries Partnership or Circular Plastics; proposers can earmark pilot budgets already.
• Cohesion policy leverage – regions may finance local fine-tuning nodes with ERDF 2021-27 smart-specialisation funds once the open model exists.
8. Societal Impact & Public Trust
• Open public goods – pan-EU citizens and NGOs gain equal access, avoiding “AI divide” between large and small Member States.
• Multi-lingual interfaces – EU consortium can natively support 24 official languages, boosting uptake by local researchers and SMEs.
• Green compliance – cross-border life-cycle assessments of model training help meet Fit-for-55 targets and set global best practice.
9. International Cooperation Leverage
• Collective bargaining power – EU consortia join the Trillion Parameter Consortium as a bloc, influencing global governance of scientific GenAI.
• Reciprocal data exchange – agreements with NASA, DOE, UKRI are easier when backed by the Commission rather than single institutes.
10. Take-Home Message
Running GenAI4EU at EU scale unlocks unparalleled data richness, compute power, regulatory certainty and interdisciplinary talent, translating into faster scientific breakthroughs, stronger industrial competitiveness and higher societal trust than any isolated national effort could achieve.
🏷️ Keywords
Ready to Apply?
Get a personalized assessment of your eligibility and application strategy
See in 5 min if you're eligible for AI Foundation models in science (GenAI4EU) (RIA) offering max €45.0M funding