Where Should Europe Build AI Gigafactories?
The strongest industrial cluster may not be the best place to plug in Europe’s next computing giant.
Europe’s AI Gigafactories — Signal III
Europe’s AI Gigafactories will need more than land and political support. They will need dependable electricity, cooling, connectivity, skilled operators and customers willing to pay. Finland, Spain and Germany offer different combinations of these assets. Brainport raises another question: should Europe place its largest computing facilities where intelligence is applied, or where infrastructure can sustain it?
The location contest is therefore a test of what Europe wants these facilities to do. The EuroHPC call allows single-site, multisite and cross-border projects. That creates room for a European architecture, provided geography follows function.
🟦 What makes a location viable beyond the announcement?
A grid connection with a delivery date matters more than an attractive plot with an ambitious masterplan.
A credible site needs electricity at a workable long-term cost, sufficient network capacity, cooling that remains effective during extreme weather, resilient fibre connections and permits that can be delivered. It also needs engineers, maintenance capacity and space for expansion. Renewable generation nearby does not automatically mean electricity can reach the facility when required.
The technical proposition must match the commercial one. Bulk training can take place far from the eventual user, provided data can be moved securely and efficiently. Inside the training cluster, however, the accelerators need high-bandwidth, low-latency connections. A distributed European network cannot treat its internal connections as an afterthought.
Local acceptance belongs in that calculation. A project competing with housing and existing businesses for scarce infrastructure must explain its public value.
The decisive asset is deliverable capacity. Everything else is a promise.
🟦 Why Finland and is cheap power enough?
Finland offers something harder to manufacture than enthusiasm: experience in operating demanding computing infrastructure.
Its government backed a Nokia-led consortium’s expression of interest in an AI Gigafactory in 2025, citing clean electricity networks and available land. Separately, Kajaani’s existing LUMI ecosystem demonstrates the value of industrial infrastructure, cooling expertise and heat recovery. CSC operates there in a former paper mill, with an industrial power network and connections to district heating.
Kajaani’s AI Factory and the proposed Gigafactory are distinct projects. The former provides operational evidence; it does not establish the location or ownership of the latter.
Finland’s strongest argument is therefore a combination of energy, expertise and infrastructure that can support intensive computation. The remaining commercial test is whether a facility can attract sustained demand across Europe.
A favourable electricity bill cannot replace a customer base.
🟦 Why Spain, if renewable power cannot guarantee continuous supply?
Spain’s opportunity lies in connecting its energy transition to a new industrial role.
Renewables supplied 55.5% of Spanish electricity generation in 2025, according to Red Eléctrica. The Spanish Gigafactory proposal combines Móra la Nova in Tarragona with San Fernando de Henares near Madrid. The government authorised a €719 million investment through SETT to join the consortium.
Móra la Nova offers a development site owned by the public land agency INCASÒL. The wider strategic opportunity is to connect that physical capacity with commercial and technological networks around Madrid. But two addresses do not establish a coherent operating system. Their respective functions, connections and costs must justify the arrangement.
The energy test is equally concrete. Solar and wind advantages must translate into dependable supply at the connection point. Cooling must work through hot periods, with a credible water budget and an acceptable local impact.
A closed cooling loop inside the facility does not settle how heat is rejected outside it. Evaporative cooling can save electricity by consuming water; dry cooling can reduce on-site water use while increasing electricity demand in hot conditions. Those choices affect operating margins.
Spain can turn energy capacity into computing capacity. It still has to prove the conversion works every day.
🟦 Why Germany, if industrial strength does not guarantee viable economics?
Germany’s strongest argument is the industry that could use the machines.
Bavaria’s Blue Swan proposal in Schweinfurt explicitly connects computing infrastructure with research and industrial applications. The Bavarian government identifies available land, grid capacity and fibre connectivity as advantages, with the Leibniz Supercomputing Centre providing operational expertise.
There is already a practical illustration of that demand. Deutsche Telekom’s Industrial AI Cloud opened in Munich in February 2026, serving applications including robotics and engineering simulation. It was developed independently of the EU Gigafactory programme.
The tension is between industrial potential and bankable demand. An automotive supplier interested in AI is not automatically a customer for large-scale model training. Companies may need smaller workloads, integration support or sensitive data to remain within their own systems.
Germany’s industrial base creates an opportunity. Contracts and competitive operating costs make it investable.
🟦 Does Brainport need to host a Gigafactory to benefit from one?
Brainport’s case is strongest where AI meets precision engineering.
Its semiconductor, photonics and advanced manufacturing ecosystem offers a plausible setting for industrial model development, simulation and testing. The region brings together companies such as ASML and NXP, specialised suppliers and research institutions. That makes Brainport worth examining as an application and technology node.
The physical siting question is much harder. Eindhoven reports that grid congestion threatens housing, businesses and public facilities; a connection pause has applied in Eindhoven-West since July 2026. A major computing installation would have to establish how its electricity requirements could be accommodated alongside those pressures.
Brainport is considered here as a strategic possibility. A strong industrial ecosystem alone does not establish a viable Gigafactory site.
A credible alternative would connect local expertise and application facilities to bulk computing elsewhere. Model training and many engineering simulations can run remotely. Time-critical robot control and machine responses may need computation on the device or nearby. Industrial AI therefore requires a division of functions, rather than a requirement to place every workload beside the factory floor.
Brainport’s value may be measured in the intelligence it helps develop, rather than the megawatts it consumes.
🟦 Who would own the infrastructure, and who would control access?
Hosting a Gigafactory does not automatically give a region control over it.
The EuroHPC programme accepts consortia and project companies combining businesses, public entities and investors. Public funding is structured partly around purchasing computing access, creating an anchor customer for facilities led by industry.
Spain’s proposed joint venture brings together ACS, Telefónica, Banco Santander and SETT. Finland’s expression of interest was led by Nokia. These examples show why land ownership, investment, technical supply and operational control must be examined separately. A consortium participant is not necessarily the eventual owner of every asset.
The decisive questions concern who sets access conditions and carries the risk when utilisation disappoints. The building and grid connection may serve for decades; the accelerators face a much faster competitive cycle. Older chips can remain useful while newer systems change the economics of frontier workloads.
Who pays for the next generation? A viable consortium needs a funded renewal plan, with responsibility for upgrades agreed before construction begins. Public support should secure usable capacity for European researchers and businesses, including those outside the host country.
The best location is the one whose infrastructure and governance can keep delivering European access after the opening ceremony.
Credit
Artwork: OpenAI Image Generation (2026), created for Altair Media.
Caption
A symbolic puzzle of Europe’s AI infrastructure. Finland, Spain, Germany and Brainport represent different location possibilities, while the surrounding pieces highlight the energy, cooling, connectivity and governance conditions that determine whether ambition can become operational capacity.
