Funder: European Commission
Due Dates: February 16, 2027 (Application deadline) | October 20, 2026 (Opening date)
Funding Amounts: €15 million per project; €15 million total topic budget; one expected grant
Summary: Supports research into predictive AI scaling methodologies and tools that improve training, inference, and computational efficiency.
Key Information: Participation is restricted to entities in specified eligible countries, with additional ownership and control restrictions.
Description
The European Commission funds this Horizon Europe Research and Innovation Action under the Frontier AI Initiative and Apply AI. It supports the scientific foundations, methodologies, and tools needed to predict AI model performance before major computing investments. The goal is to improve training and inference, deploy limited computational resources efficiently, and help European AI labs compete at the global frontier.
Proposals should deliver a toolset for scaling AI models, rather than focus only on model training. Research may address automated scaling laws across modalities and tasks, architecture search, dataset mixture optimisation, automated data quality assessment, and hyperparameter tuning for stable large-scale training. Frameworks should work across diverse hardware architectures, avoiding optimisation for particular proprietary systems, and methodological advances should be validated with advanced AI labs and high-performance computing centres.
