France’s Bull is doubling production capacity at its Angers factory as Europe attempts to turn its sovereign AI ambitions into physical industrial capacity.
Europe’s argument about AI sovereignty is rapidly becoming an argument about factories.
For years, the continent’s technology strategy has focused on data, regulation, cloud sovereignty and access to advanced processors. But the explosion in generative AI has exposed another dependency: possessing algorithms and computing chips means relatively little without the industrial capacity to turn thousands of processors, cooling systems, networking components and power infrastructure into machines capable of training and running frontier models.
In Angers, western France, Bull is betting that Europe will increasingly want to build those machines at home.
The advanced computing company has opened a new production building at its historic Angers factory, doubling manufacturing capacity as part of an €80 million reconstruction of the site. The facility will produce supercomputers, AI infrastructure and enterprise servers, with quantum technologies expected to join that list in future.
It is an industrial expansion, but one carrying distinctly geopolitical undertones.
Europe is pouring money into AI Factories, exascale computing and sovereign digital infrastructure as governments confront the strategic implications of relying heavily on technology manufactured or controlled elsewhere. Bull’s rebuilt factory represents the less glamorous but increasingly critical end of that strategy: actually manufacturing the infrastructure.
The new building covers 20,000 square metres, including 10,700 square metres devoted to production. Bull says the investment will double capacity while increasing productivity by 30%.
The wider reconstruction is scheduled for completion in 2028.
From AI policy to AI machinery
The timing matters.
Artificial Intelligence infrastructure is moving towards increasingly enormous, power-hungry systems containing tightly interconnected accelerators. Building them requires expertise stretching beyond conventional server assembly into cooling, power management, networking, systems integration and large-scale validation.
Bull wants Angers to become one of Europe’s industrial answers to that challenge.
The company describes the factory as the only industrial site in Europe capable of assembling large-scale supercomputers. It designs, integrates, assembles, tests and validates systems at the same location, bringing together capabilities that have become strategically important as nations reconsider where critical computing infrastructure originates.
Angers already has experience at the extreme end of computing.
The factory contributed to JUPITER, Europe’s first exascale supercomputer, designed to cross the threshold of one quintillion calculations per second. Bull also points to its involvement with machines occupying leading positions in the Green500 ranking of the world’s most energy-efficient supercomputers.
Next comes another wave.
In 2027, Angers is expected to deliver Alice Recoque, the next European exascale supercomputer hosted in France, alongside LUMI-AI, the computing system underpinning the LUMI AI Factory in Finland.
These are not simply faster computers. Systems such as these are becoming infrastructure for scientific research, industrial simulation, defence, climate modelling and increasingly AI.
That convergence between supercomputing and AI is changing what an advanced computing factory needs to be.
Bull says its new facility has therefore been designed around Industry 5.0 principles, with production areas capable of being reconfigured as technologies change. Greater automation, virtualisation and collaborative robots will form part of the manufacturing environment.
That flexibility could prove essential. AI hardware architectures are evolving at a speed that makes conventional long-lived production lines potentially restrictive. The factory built for today’s GPU-heavy AI clusters may have to accommodate different accelerators, cooling technologies and hybrid computing architectures within only a few years.
Quantum computing adds another variable.
Bull is bringing its work in supercomputing, enterprise servers and quantum technologies together at Angers while expanding its AI capabilities. Around 800 R&D engineers will support the technology developed for the site.
The result is effectively an attempt to create an advanced-computing production platform rather than a factory dedicated to a single generation of hardware.
Sovereignty meets the supply chain
There is also a political dimension that Bull makes little attempt to disguise.
The French state became Bull’s sole shareholder on 31 March 2026, making the company’s manufacturing capacity part of a wider debate about European technological independence.
Roland Lescure, French minister for the Economy, Finance and Industrial, Energy and Digital Sovereignty, described control over computing infrastructure as increasingly important in a fragmented geopolitical environment.
“In a world that is becoming increasingly fragmented, including in the digital sphere, it is imperative that France and Europe retain control over their own computing capabilities,” he said.
That idea of control is increasingly stretching down the technology stack.
A country can operate a sovereign cloud, protect its data and develop domestic AI models while still depending on overseas companies for processors, servers, networking equipment and manufacturing. Complete technological independence is neither realistic nor necessarily desirable, but governments increasingly want greater control over selected parts of the supply chain.
Bull’s strategy reflects that distinction.
Its Angers factory does not eliminate European dependence on global technology companies. Instead, it attempts to ensure that more of the engineering, integration, manufacturing and validation surrounding critical computing systems can happen inside Europe.
The company is now extending that proposition beyond its own products.
Opening the factory doors
Alongside the new building, Bull has launched an initiative called Open Factory.
The programme will make Angers’ industrial capabilities available to other companies developing technologies in AI, advanced computing and quantum systems.
The target is a persistent weakness in European technology: moving inventions out of research laboratories and into industrial-scale manufacturing.
Europe has world-class universities, research organisations and start-ups, but converting technical breakthroughs into high-volume products has frequently proved more difficult. Bull believes an existing industrial infrastructure for designing, integrating and validating complex computing systems could help bridge that gap.
That challenge is becoming more urgent as Europe attempts to translate billions of euros of investment in AI research into commercially deployable infrastructure.
Open Factory is intended to support technologies ranging from AI GigaFactories and neocloud infrastructure to hybrid systems combining high-performance and quantum computing.
Its first significant project illustrates the scale of that ambition.
Bull recently announced a strategic partnership with Foxconn to design and manufacture AI servers in Europe using NVIDIA’s Vera Rubin NVL72 platform. The machines are intended for AI Factories and cloud providers.
The arrangement is revealing. European technological sovereignty does not necessarily mean constructing an isolated European supply chain. NVIDIA remains a US company and Foxconn is headquartered in Taiwan. Instead, the strategy is about moving more manufacturing, engineering expertise and industrial capability onto European territory.
For Europe, that may be a more practical definition of sovereignty: not independence from the global technology ecosystem, but greater influence over where critical infrastructure is assembled and who possesses the expertise required to build it.
Bull says further Open Factory partnerships involving AI and quantum technologies will be announced in early 2027.
The power problem
There is another constraint looming over the expansion of AI infrastructure: energy.
The enormous electricity requirements of AI clusters have turned power availability into one of the defining limitations on data centre development. Manufacturing the machines themselves also carries an environmental footprint.
Bull has incorporated energy infrastructure into the Angers redevelopment.
More than 3,500 photovoltaic panels will cover approximately 7,000 square metres, providing a combined peak capacity of 1,500 kWp. The site will also use electricity supplied from renewable sources.
More unusually, heat generated by equipment will be captured rather than simply expelled.
Bull is targeting recovery of 5 GWh of waste heat annually, which will be injected into the local district heating network. The company estimates that this could provide energy equivalent to the requirements of around 600 households.
The redevelopment is also expected to reduce building maintenance costs by 30%.
Public funding has contributed to the project. It includes €1.2 million from France’s Ecological Transition Acceleration Fund for Local Authorities, €213,000 from ADEME’s Heat Fund and €1.5 million from the Pays de la Loire Regional Council. Local public development company Alter Cités is overseeing the overall redevelopment.
But the real test will not be the building itself. It will be whether Bull can keep its expanded production lines busy.
Europe’s hardware test
Europe’s AI ambitions are escalating rapidly. AI Factories, sovereign cloud programmes, national supercomputers and proposed GigaFactories are creating demand for computing infrastructure on a scale that would have looked extraordinary only a few years ago.
Manufacturing capacity now has to catch up.
“Europe must have the industrial capabilities required to produce tomorrow’s strategic digital infrastructures on its own territory,” said Bull CEO Emmanuel Le Roux.
That statement captures the wager behind Angers.
Since opening in 1963, the factory has survived several generations of computing, from mainframes through enterprise systems to supercomputers. Now it is being rebuilt for an era in which AI infrastructure is becoming both an industrial asset and an instrument of national strategy.
The important number may therefore not be the €80 million investment, the 20,000 square metres of new space or even the doubling of production capacity.
It is 1963.
Europe already knows how quickly leadership in computing can migrate elsewhere. Bull’s wager is that preserving a place in the next computing era requires more than research funding, regulation and ambitious AI strategies.
At some point, somebody still has to build the machines.
