ALIA: Spain's LLM and Technological Sovereignty
· AI · Alejandro Cantero Jódar

ALIA: The Future of AI in Spain
The ALIA artificial intelligence model, developed by the Spanish government, represents a significant advancement in technological sovereignty and innovation in Spain. This project, coordinated by the Barcelona Supercomputing Center (BSC), aims to promote the use of Spanish and co-official languages in the development and deployment of AI. With an additional investment of 150 million euros for its integration into companies, ALIA stands out for its transparency and openness, key characteristics for democratizing access to AI for citizens, public administration, universities, and businesses. This article explores ALIA's capabilities, its impact on the Spanish economy and society, and its positioning in the broader context of AI development and adoption.
The Role of the Barcelona Supercomputing Center
The Barcelona Supercomputing Center (BSC) has played a crucial role in developing ALIA, using its MareNostrum 5 supercomputer to train the model. This effort reinforces Spain's technological sovereignty and represents an important step in the broader movement toward national and regional autonomy in AI development. The investment in ALIA is part of the National Artificial Intelligence Strategy 2024, which aims to modernize Spain's productive fabric and public services.
The Importance of Supercomputers in AI
Supercomputers like MareNostrum 5 are essential for advancing artificial intelligence. These machines can process vast amounts of data and perform complex calculations at unprecedented speeds, which is fundamental for training advanced AI models. The computational power of supercomputers allows researchers and developers to explore new frontiers in machine learning, natural language processing, and other areas of AI.
In the European context, only Italy is actually in the top 10 of the world's 500 most powerful supercomputers, according to the Top500.org website. This position highlights the importance of investing in supercomputing infrastructure to maintain global technological competitiveness. Supercomputers not only drive innovation in AI but also have applications in various fields such as medicine, engineering, and scientific research.
The MareNostrum 5 supercomputer, located at the BSC, is one of the most powerful in the world. Its processing capacity has been crucial for training ALIA, enabling the handling of billions of parameters and tokens. This extensive training has given ALIA a deep understanding of the language and cultural nuances of Spain, making it one of the most advanced AI models in Europe.
Capabilities and Training of ALIA
ALIA has been trained on a vast corpus of data, including classical literature, official documents, and scientific content, allowing it to deeply understand the language and cultural nuances of Spain. The model, with 40 billion parameters, has been trained on 9.2 trillion tokens in 35 European languages, making it one of the most advanced of its kind in Europe. This extensive training was possible thanks to the computational power of MareNostrum 5, one of the world's most powerful supercomputers.
The training of ALIA involved processing billions of words and phrases, enabling the model to learn and understand a wide variety of contexts and linguistic nuances. This process was fundamental in developing an AI model that not only understands language but can also generate coherent and precise text in multiple languages. ALIA's ability to handle 35 European languages makes it an invaluable tool for the continent's linguistic and cultural diversity.
At the moment, ALIA keeps training on BSC, but you can download the intermediate checkpoint from HuggingFace here.
Investment and Practical Applications
The Spanish government has announced an additional investment of 150 million euros to integrate AI into Spanish companies. Of this investment, 20 million euros will be allocated to 500 use cases in small and medium-sized enterprises, aiming to create a more innovative and competitive ecosystem. This initiative seeks not only to improve business efficiency but also to be part of the broader trend where governments and large entities develop their own AI models to achieve technological autonomy.
ALIA is already being applied in several public projects, including an internal chatbot for the Tax Agency and another for the healthcare sector to diagnose heart failure early. These projects demonstrate ALIA's potential to improve the efficiency and quality of public services in Spain. Additionally, ALIA is expected to be used in other sectors such as education, where it can help personalize learning and improve educational outcomes, and in scientific research, where it can accelerate the discovery of new knowledge.
Geopolitical and Cultural Implications
Beyond its practical applications, ALIA also has significant geopolitical implications. In a context where Europe is increasing its defense budget, the AI technology developed by ALIA can have a relevant impact on the continent's security and defense. However, the competition over which AI models will be used within and outside different countries, regions, entities, or enterprises may transcend into a cultural struggle. These models are often shaped by the internal culture of their developers, as has already been observed with models from OpenAI and Deepseek. The biases and cultural inclinations embedded in AI models could lead to significant influence over how information is processed and presented in different parts of the world.
Technological sovereignty is a crucial aspect of modern geopolitics, and ALIA represents a small but significant step in Spain's and Europe's journey toward greater technological independence. By developing and controlling its own AI technology, Spain can reduce its dependence on foreign suppliers and better protect its national interests. Additionally, collaboration with other European countries in developing ALIA can strengthen technological and economic ties within the European Union.
Rather than a definitive leadership position in AI, Spain's investment in ALIA represents just the beginning of a larger trend. Governments and large organizations worldwide are recognizing the necessity of having their own AI models to maintain autonomy in the digital age. The debate over which models prevail in different contexts will continue to shape AI's influence on society, making transparency and ethical considerations essential factors in future developments.