EU launches call for seven AI gigafactories: infrastructure will be the critical issue

Introduction

• The European Commission opened its call to establish seven AI gigafactories on July 30, 2026.

• According to the published plans, the program anticipates EUR 10 billion in public funding and approximately EUR 20 billion in private investment.

• The facilities would be designed to accommodate at least 100,000 advanced AI processors.

• A key feasibility issue will be coordinating grid connections, critical power, redundancy and high-density cooling.

• Technical decisions must account for the full lifecycle, not only the initial IT capacity.

The European Commission's announcement of July 30, 2026 marks a new phase in the development of European AI capacity. Given the scale of the seven planned AI gigafactories, however, the issue is no longer limited to procuring processors or developing AI models. Equally important will be how quickly and reliably the selected sites can provide electrical power, heat rejection and the associated critical infrastructure.

The announcement therefore points to specific design and delivery tasks for the data center industry. Grid capacity, medium- and low-voltage distribution, backup power, cooling for high-power-density racks and integrated commissioning will need to be treated as parts of a single interconnected system.

Who is this analysis for?

This article is intended for data center operators, facility owners, investors, technical procurement teams, designers, consultants and colocation providers. Its focus is not AI software, but the physical infrastructure decisions that determine facility capacity, availability, scalability and maintainability.

The industry impact extends beyond these seven projects. The supply chain, grid and engineering requirements associated with AI gigafactories may also affect the schedules and resource needs of other European data center developments.

What happened on July 30, 2026?

The European Commission opened its call to establish seven AI gigafactories. The announced funding framework anticipates EUR 10 billion in public financing and approximately EUR 20 billion in additional private investment.

The planned facilities would accommodate at least 100,000 advanced AI processors. This number of processors creates substantial electrical power and heat-rejection requirements, while the infrastructure must continue to operate predictably during failures, maintenance and future capacity expansions.

The scale of the projects does not automatically imply a single correct technical architecture. The required design will be determined by a combination of site-specific grid conditions, the IT load profile, power density per rack, availability targets and the delivery schedule.

Why is power becoming a critical scheduling factor?

The electrical infrastructure of an AI gigafactory extends from the utility grid connection to power distribution at individual racks. It may include medium-voltage switchgear, transformers, low-voltage main distribution boards, UPS systems, generators, automatic and static transfer switches, PDUs and high-current busways.

The design task is not limited to adding up nominal power ratings. The assessment must include, among other factors:

• normal, backup and maintenance operating states;

• whether redundancy is genuinely maintained from end to end;

• protection selectivity and the ability to isolate faults;

• load ramp-up and partial-load operation;

• the physical and electrical reserves required for future expansion;

• operational access and safe maintainability.

Grid connections and equipment with long lead times may become part of the project's critical path. The electrical concept must therefore be coordinated with site assessment and IT capacity planning from an early stage.

Cooling cannot be separated from the IT architecture

The heat loads generated by high-density AI racks may differ from those typical of conventional enterprise server rooms. As a result, each project requires a separate assessment of whether air cooling alone is suitable, what proportion of liquid cooling is required and how the two systems will operate together.

Introducing liquid cooling involves more than selecting coolant distribution units or rack-side heat exchangers. The entire heat-rejection chain must be sized and coordinated, including process-water loops, pumps, heat exchangers, pipework, isolation points, leak detection, controls and external heat-rejection equipment.

Water quality, material compatibility, pressure and temperature ranges are critical considerations, as is the effect that the failure or maintenance of one unit may have on the active IT load. Cooling redundancy should therefore not be defined solely by equipment count. Shared pipe sections, controls and power dependencies must also be evaluated.

What does this mean for operations?

AI gigafactory infrastructure may create a greater number of interdependencies between electrical, mechanical and IT systems. An electrical transfer may affect pumps and controls, while a cooling constraint may require a rapid reduction in IT load. These relationships must be visible both in operating procedures and within the monitoring system.

Decision-makers should define in advance:

• which faults can be handled automatically and when operator intervention is required;

• which alarms require immediate load limitation;

• how maintenance can be performed without putting active capacity at risk;

• which spare parts and specialist skills must be available on site;

• which data must be retained for audits and incident analysis.

Effective operability requires clear system boundaries, current documentation, trained personnel and regularly tested emergency procedures.

A common mistake: treating interconnected systems as separate projects

A frequent risk is that the IT platform, electrical infrastructure, cooling and building management systems are designed through separate processes, with interfaces becoming visible only at a late stage. This can result in incorrect capacity assumptions, unsuitable control sequences or redundancy that exists only on paper because of a shared dependency.

Optimizing solely for the final peak capacity is also risky. A facility may operate at partial load for an extended period, so the controllability, efficiency and redundancy of the power and cooling systems must be verified across multiple load conditions.

Recommended next step

For potential project owners and infrastructure partners, the first step should be an integrated feasibility and capacity assessment. This should include at least the following work packages:

1. Baseline load model: Define IT capacity, power density per rack and the expected ramp-up schedule. 2. Site power audit: Assess the grid connection, distribution paths, short-circuit and protection selectivity conditions, and expansion capability. 3. Cooling concept: Coordinate air- and liquid-based heat rejection, hydraulic boundaries, backup states and external heat-rejection systems. 4. Redundancy analysis: In addition to N, N+1 or 2N targets, identify common points of failure and maintenance states. 5. Phased delivery: Clearly separate the first operational capacity from later modules. 6. Commissioning plan: Define factory, site and integrated tests, responsibilities, acceptance criteria and documentation.

Under the Digital Technologies engineering approach, these areas are not treated as separate product procurements. Electrical and cooling architectures, rack systems, structured cabling, monitoring, security systems and fire detection must be coordinated around shared operational objectives.

Availability cannot be verified without commissioning

Design documents and manufacturer data sheets alone do not prove that the complete facility will respond correctly to a real failure. In addition to individual equipment tests, the acceptance process must therefore include integrated scenarios.

These may include the loss of an incoming power feed, a UPS or generator transfer, the shutdown of a cooling unit, a communications failure or the use of a maintenance bypass. The objective is not merely a fault-free start-up, but evidence that the infrastructure operates safely, predictably and in a documentable manner across different operating states.

Conclusion

The EU call aims to create substantial AI capacity in Europe, but the success of the projects will not be determined solely by the number of installed processors. Grid connections, critical power, high-density cooling, redundancy and commissioning will collectively determine when the capacity becomes usable and the level of operational risk involved.

The most important practical lesson is the need for early integration. IT, electrical, mechanical and operational requirements must be brought together in a common model during the concept design stage. This provides the foundation for phased delivery, maintainable redundancy and operations that remain verifiable over the long term.

Sources

• European Commission press release: https://ec.europa.eu/commission/presscorner/detail/en/ip_26_1708

• European Commission – AI Gigafactories: https://commission.europa.eu/topics/competitiveness/competitiveness-coordination-tool-projects/ai-gigafactories_en

• Associated Press background article: https://apnews.com/article/eu-ai-gigafactories-china-us-data-center-88b83cd517a4d47c115605e636d0b3e4