Summary
- The ÆTHER consortium has applied for the European Commission’s AI Gigafactory call and is in advanced talks for two Strasbourg-region sites.
- The first two campuses would initially provide 42MW, with a further 40MW targeted within 12 months of commissioning.
- Longer-term ambitions above 400MW remain dependent on grid availability and proposals from French transmission operator RTE.
2CRSi has joined a European industrial consortium seeking to develop AI gigafactory infrastructure in the Strasbourg region, with the project company in advanced negotiations for two industrial sites intended to host large-scale AI computing facilities.
The ÆTHER consortium has applied for the European Commission’s AI Gigafactory call for proposals, which sits within the EU’s wider industrial and digital sovereignty agenda. Its founding members span energy, infrastructure, cloud computing, semiconductors, high-performance computing, and artificial intelligence.
The first two proposed campuses, known as FR-SXB1 and FR-SXB2, are planned around existing infrastructure and administrative approvals. FR-SXB1 is expected to begin operations during 2027 if its acquisition is completed by the end of October 2026. FR-SXB2 is expected to follow a few months later, subject to completion of its acquisition by the end of December 2026.
The initial deployment would provide a combined 42MW of electrical capacity across the two sites. Within 12 months of commissioning, the consortium intends to secure a further 40MW. Its longer-term ambition is to exceed 400MW of total electrical capacity, although that depends on grid availability and proposals from RTE, France’s transmission system operator.
Compute policy reaches the substation
The Strasbourg proposal moves the European AI sovereignty discussion into the physical layer. Large AI systems need more than domestic model development, public funding, or cloud procurement frameworks. They need land, substations, cooling systems, fibre, construction teams, power purchase strategies, security, and a credible operating model for dense, high-load compute.
Strasbourg gives the project a politically useful location. It sits close to European institutions and within the Rhine economic corridor, while the Grand Est region offers industrial sites, cross-border links, and a manufacturing base that can be connected to the consortium’s sovereignty pitch. Those advantages still have to be turned into grid capacity and buildable facilities.
The consortium is presenting ÆTHER as more than a data centre development. Its model combines high-performance computing, sovereign cloud, energy recovery, scientific research, industrial innovation, and regional development. That wider frame may strengthen the bid, but it also expands the delivery burden. A gigafactory for AI compute needs a coherent technology stack, a bankable energy strategy, and enough commercial demand to justify staged capacity growth.
The 42MW starting point is already material for an AI infrastructure project in France. The 400MW-plus ambition moves the proposal into campus-scale territory and puts grid sequencing at the centre of the plan. Identifying sites is only the first step. The harder work is securing transmission capacity, agreeing connection works, procuring electrical equipment, and phasing load in a way that fits both the grid and the commercial ramp-up.
Heat, density, and delivery
AI gigafactory projects are being asked to answer harder questions earlier than conventional cloud-region schemes. Dense GPU and accelerator environments change power distribution, cooling design, and resilience assumptions. Load profiles can be concentrated and less forgiving, while thermal design has to account for higher rack densities and potentially faster hardware refresh cycles.
The consortium’s references to energy recovery give the project a useful infrastructure hook, but heat reuse depends on local conditions. Useful heat export requires offtakers, network infrastructure, temperature compatibility, commercial agreements, and planning support. A district heating concept is only valuable if it survives the engineering and commercial tests.
Industrial zoning and existing infrastructure could help the first sites move faster than a greenfield scheme. They do not remove the constraints around high-voltage delivery, cooling plant, fire strategy, fibre diversity, emergency power, and operations. AI workloads may also require closer coordination between server design and facility design than older enterprise environments.
For 2CRSi, the project connects domestic server manufacturing with the infrastructure needed to host European AI workloads. The company designs and manufactures high-performance, energy-efficient servers, which gives it a different role from a pure real estate developer or colocation operator. Server architecture, deployment density, and facility energy design sit closer together in AI infrastructure than they did in older hosting models.
The next milestones will decide whether the proposal becomes an EU-backed industrial project or remains a well-framed ambition. Site acquisitions, RTE’s grid response, local approvals, cooling design, funding structure, and customer demand will determine how much of the stated capacity can be delivered. Europe has plenty of AI policy language. The Strasbourg bid will be judged by substations, permits, cooling loops, and commissioning dates.
