Spain has taken a decisive step to compete for one of Europe’s future AI gigafactories. The Council of Ministers has authorized the participation of the Spanish Society for Technological Transformation (SETT), known as SEPI Digital, in a newly created company aimed at developing and managing the Spanish bid for the upcoming European call.
The project will have a public-private structure and a multisite proposal, with Móra la Nova in Tarragona and San Fernando de Henares in Madrid as the intended locations to house the infrastructure. The initiative falls under European InvestAI, which Brussels aims to use to reduce technological dependence on the United States and China in advanced computing capacity for training and deploying large AI models.
The company will be majority-controlled by private capital. Banco Santander, ACS, and Telefónica will jointly control 47%, each holding 15.67%. Multiverse Computing will add 4%, bringing private participation to 51%. SETT will hold 47.99%, and the Generalitat de Catalunya will initially hold 1%. The structure aims to combine public support, financial capability, industrial experience, and technological strength.
An infrastructure to compete in AI, not just a data center
The term “gigafactory” might sound exaggerated, but it describes a category distinct from a conventional data center. These facilities are designed to host hundreds of thousands of GPUs or cutting-edge accelerators, with ultra-low latency networks, advanced cooling, high electrical density, and specialized software to train and operate complex AI models.
The difference is not only in size. A traditional data center can host enterprise applications, storage, private cloud, or digital services. An AI gigafactory is built for far more demanding loads: training large language models, computer vision, scientific simulation, multimodal models, industrial AI, and large-scale inference deployments.
| Element | Conventional Data Center | AI Gigafactory |
|---|---|---|
| Main Load | Cloud, hosting, applications, storage | Training and inference of advanced models |
| Dominant Hardware | General-purpose servers | GPUs and AI accelerators |
| Internal Network | High capacity, but less specialized | Low latency, high bandwidth between nodes |
| Cooling | Air or hybrid solutions | Liquid cooling and high-density designs |
| Electric Consumption | High, but variable | Very high and concentrated |
| Strategic Use | Business digitization | Technological sovereignty and scientific-industrial capacity |
For Spain, the goal is not just to build a large facility. The ambition is to be part of the European advanced computing network that enables companies, universities, research centers, administrations, and startups to train models without relying entirely on foreign infrastructure. This dependence has become especially visible with US hyperscalers’ dominance and China’s closed ecosystem.
Santander, ACS, and Telefónica: three profiles for one shared goal
The involvement of Santander, ACS, and Telefónica is intentional. Each company contributes a different piece to the project. Santander signifies financial capacity, access to large clients, and experience in banking digitalization. ACS brings expertise in infrastructure, construction, engineering, and major industrial projects. Telefónica adds connectivity, data centers, enterprise services, cybersecurity, and experience with critical networks.
Multiverse Computing provides a more specialized layer, focused on advanced software, quantum computing, and algorithms for complex problems. Its 4% stake is smaller in capital but significant in technological terms, bringing the consortium closer to developing AI tools and applications.
| Partner | Planned Participation | Likely Role |
| Banco Santander | 15.67% | Funding, enterprise clients, digital services |
| ACS | 15.67% | Infrastructure, construction, industrial execution |
| Telefónica | 15.67% | Connectivity, cloud, cybersecurity, tech operations |
| Multiverse Computing | 4% | Advanced software and AI tools |
| SETT / SEPI Digital | 47.99% | Public backing and strategic state participation |
| Generalitat de Catalunya | 1% | Regional and institutional anchor |
Most private partners aim to send a clear signal to Brussels: the project is not solely reliant on public funds. Behind it are capable companies that can execute, finance, operate, and generate demand. This will be crucial in European evaluation, as AI gigafactories must attract users, use cases, talent, and industrial projects to avoid underutilization.
Móra la Nova and San Fernando de Henares: a multisite proposal
The Spanish bid is not conceived as a single isolated location. It will be multisite, with Móra la Nova playing a key territorial role in Tarragona, and San Fernando de Henares serving as an associated node near Madrid. This architecture aims to combine land availability, connectivity, proximity to talent, access to companies, and regional balance.
Móra la Nova’s choice has obvious industrial and political implications. The Ribera d’Ebre area seeks economic renewal amid the potential nuclear activity shutdown. A project of this scale in AI could serve as a transformational driver, though its actual impact depends on execution, skillful employment generation, local suppliers, and integration with the region.
San Fernando de Henares offers closeness to Madrid’s metropolitan area, one of Europe’s key hubs for connectivity, cloud, tech companies, and data centers. Combining both sites may strengthen the bid if it can present a coherent technical architecture rather than just a sum of locations.
| Location | Value to the project |
| Móra la Nova | Land, regional reindustrialization, Catalan anchor |
| San Fernando de Henares | Proximity to Madrid, connectivity, business environment |
| Catalonia | Industrial, scientific, and technological potential |
| Madrid | Cloud, enterprise, and communication hub |
| Spain | Geographical position, renewables, and international connectivity |
| Europe | Need for independent capacity compared to US and China |
Energy will be a central concern. An AI gigafactory requires sufficient power, supply stability, long-term energy agreements, and a clear sustainability strategy. Without energy, GPUs cannot run. Without proper cooling, the infrastructure is not viable.
Almost €4 billion for a broader European race
The Spanish bid aims to mobilize nearly €4 billion in public-private investment, though estimates suggest the project could scale higher with EU backing. The government has authorized a maximum contribution of €719.85 million through SETT, funded from digital transformation and European funds.
Europe plans to create up to five AI gigafactories within InvestAI, backed by a dedicated fund of €20 billion. These infrastructures are meant to complement existing AI factories around European supercomputers, such as those linked to EuroHPC. The key difference is scale, with gigafactories targeting much larger capacities for frontier models and industrial applications.
| Magnitude | Relevant Data |
| Maximum contribution from SETT | €719.85 million |
| Public national share | 47.99% |
| Total private participation | 51% |
| Regional government’s share | 1% |
| Planned public-private investment | Around €4 billion |
| EU InvestAI fund for gigafactories | €20 billion |
| Planned gigafactories in Europe | Up to five |
The size of this investment underscores that AI has become an industrial policy issue. While Europe has excelled in regulation, research, and talent, it has relied heavily on others for access to large computing infrastructures. Without independent capacity, European companies are forced to train and deploy models on foreign platforms, incurring costs, dependency, and sovereignty concerns.
The key will be who can utilize this capacity
One of the main challenges is ensuring the gigafactory doesn’t become a facility reserved only for large corporations. The public value of the project depends on enabling SMEs, startups, universities, research centers, governments, and industrial firms to access computing power on fair terms.
This involves more than just installing GPUs. It requires defining access models, pricing, priorities, usage guarantees, development tools, technical support, interoperability with existing supercomputers, and clear data policies. Prioritization sectors might include health, industry, energy, defense, language, science, automotive, public administration, or European languages models.
| Pending Questions | Why It Matters |
| Which companies will access? | Impacts real-world economic influence |
| At what cost? | Determines SME access |
| Which models will be trained? | Impacts sovereignty and applications |
| How will data be protected? | Crucial for regulated sectors |
| How much energy will be used? | Impacts sustainability and local viability |
| What talent will be attracted? | Decides if it becomes an ecosystem or just infrastructure |
| Connection to EuroHPC? | Prevents duplication, optimizes usage |
The risk with such projects is building infrastructure without sufficient ecosystem support. To succeed, the gigafactory must integrate with universities, tech centers, software companies, cloud providers, industry, public administrations, and developer communities. Hardware alone is necessary but not enough.
Technological sovereignty with very real costs
The narrative of technological sovereignty is compelling, but it must be tempered with realism. The most advanced GPUs and accelerators will continue to rely heavily on U.S. suppliers. HBM memory mainly comes from South Korea and the US. Cutting-edge semiconductor manufacturing depends on Taiwan, the Netherlands, Japan, and other players. A European gigafactory won’t eliminate this dependency but could reduce it at the operational and access layer.
Furthermore, AI infrastructure involves very high recurring costs: energy, cooling, maintenance, hardware upgrades, licenses, security, connectivity, specialized personnel, and technological refreshes. GPUs age quickly, and what’s state-of-the-art today might be outdated in a few years.
| Challenge | Impact |
| Energy | Affects location, cost, and sustainability |
| Hardware renewal | Requires ongoing investment |
| Talent | Needs highly specialized personnel |
| Security | Infrastructure will be critical and sensitive |
| Governance | Must balance public and private interests |
| Global competition | US and China already hold significant advantage |
| Effective utilization | Avoids costly underused infrastructure |
Spain’s advantage could lie in combining renewables, connectivity, geographic position, supercomputing experience, and partnerships with capable companies to generate demand. However, European competition will be fierce. Other countries also aim to host gigafactories and already have strong cloud, industrial, or energy ecosystems.
A strategic move, but no guaranteed victory
The approval of the company and initial public investment does not mean Spain has secured a European gigafactory. It means it arrives at the call with a well-structured ownership, a territorial proposal, and initial funding to stay competitive. The final decision will depend on Brussels, the technical quality of the bid, its sustainability, private investment committed, and execution capacity.
The involvement of Santander, ACS, and Telefónica lends significance to the project and raises expectations. If Spain wants this infrastructure to be more than just a symbol, it must prove that it can turn it into a useful platform for Europe’s economic and scientific fabric.
AI is no longer just about models or applications. It’s about energy, chips, data centers, networks, funding, regulation, and access. Through this partnership, Spain seeks to position itself in that deepest layer of technology — the most expensive, least visible, and hardest to improvise.
The Móra la Nova and San Fernando de Henares bid presents a significant opportunity but also raises inevitable questions about energy, territorial impact, governance, access, costs, and public return. The true difference between a strategic infrastructure and an overhyped promise will depend on how these questions are answered in the coming months.
Frequently Asked Questions
What has the Spanish government approved?
The Council of Ministers has authorized a maximum investment of €719.85 million through SETT to participate in the company presenting Spain’s bid for a European AI gigafactory.
Which companies will be part of the private capital?
Banco Santander, ACS, and Telefónica will collectively control 47%, each holding 15.67%. Multiverse Computing will add another 4%, bringing private participation to 51%.
Where would the gigafactory be located?
The Spanish bid will be multisite, including Móra la Nova in Tarragona and San Fernando de Henares near Madrid, as the intended locations for the infrastructure.
How much investment will the project mobilize?
The initiative aims to mobilize nearly €4 billion in public-private investment, with a maximum state contribution of €719.85 million.
Why is this important for Spain and Europe?
Because it would provide advanced computational capacity to train and deploy AI models, reducing dependence on foreign infrastructure and supporting companies, research centers, and administrations.
