New StepUp Startups report: Scaling Europe’s open-source AI landscape
A new report, The European Open-Source AI Landscape, has just been released by the StepUp Startups project. The report offers a snapshot of where Europe currently stands in the global AI race, and explores how faster adoption of open-source AI could boost competitiveness while strengthening the EU’s digital sovereignty.
Open-source AI refers to models, tools and datasets whose components (such as source code, model weights and documentation) are openly available. This transparency lowers the barrier for universities, public institutions, startups and established companies alike, allowing them to develop and deploy AI without being tied to proprietary systems.
According to the report, open-source is already embedded across the AI tech stack: more than half of developers rely on open models, datasets or tools on a regular basis. It also supports safer AI, giving researchers the visibility they need to test and evaluate models in depth.
A fast-moving global race
Since 2022, the number of publicly released AI models has more than doubled, and more of them now come with open weights. This gives developers far greater insight and control. Open models are also rapidly catching up with proprietary systems and, on some benchmarks, even matching expert-level performance.
Europe’s strengths: Collaboration and trust
Europe comes to this challenge with real advantages. The region boasts top-tier research, close collaboration between academia and industry, and major contributions to tools like scikit-learn, spaCy and PyTorch, which underpin thousands of commercial AI systems worldwide.
Many European AI startups are also releasing open-source models under open-source licences, providing building blocks that others can use and scale.
Yet adoption in European businesses remains relatively low: in 2024, just 14% of EU firms were using AI. To tackle this, the EU launched its Apply AI strategy in October 2025, focused on increasing uptake across European companies. The report highlights that open-source AI can play a key role in closing this ‘use gap’, offering transparent, reusable, and cost-effective tools aligned with EU values.
Scaling open-source AI: Europe’s opportunity
Access to computing remains a major challenge for European innovators. The 19 EU-funded AI Factories and expanded EuroHPC supercomputers aim to change that, giving startups and SMEs free access to the GPU capacity they need to develop European models, including open-weight systems.
This public infrastructure is already bearing fruit. In September 2025, Latvian SME Tilde launched TildeOen LLM, a 30-billion-parameter open-source language model, trained using 2 million GPU hours on the EuroHPC LUMI supercomputer.
A path towards sovereign and competitive AI
With inference costs falling by more than 99% in just two years, AI is rapidly becoming low-cost and widely accessible. Combined with Europe’s research strengths, multilingual talent and regulated approach, open-source offers a real opportunity for the EU to carve out a leading position.
Rather than focusing only on building the largest frontier models, Europe can take the lead by developing open, trusted, multilingual and sector-specific AI aligned with its industrial and societal priorities. Doing so will require investment, easier access to compute, and targeted support to help organisations – particularly SMEs – adopt and deploy open-source AI effectively.
Download the report to explore the full findings.
The post New StepUp Startups report: Scaling Europe’s open-source AI landscape appeared first on EU-Startups.
