Building Europe’s AI Factories: How Nebius, Azur Datacenters and Inflect Are Reshaping Northern France’s Digital Infrastructure

After two years dominated by investment announcements, sovereign AI initiatives and multi-billion-dollar capacity plans, attention across the industry is shifting decisively towards execution. The question is no longer whether demand for AI compute exists: it is whether developers can actually deliver the power, cooling and infrastructure required to support it.

That was the central theme of a panel discussion at the infra/CAPITAL Summit 2026 in Paris, where two executives involved in one of Europe’s most significant AI infrastructure transactions of 2025 came together to walk through how it happened. The discussion brought together Philbert Shih, Managing Director at Structure Research, Emmanuel Vannier, Founder, CEO and CTO of Azur Datacenters, and Mike Nguyen, CEO of Inflect.

The project at the heart of the conversation: a 240MW AI campus in Béthune, Northern France, developed by Azur Datacenters for Nebius, one of the most closely watched neocloud operators in Europe. This highly complex transaction was initiated and facilitated by Inflect, a digital infrastructure marketplace and advisory firm. Inflect identified Azur Datacenters as a highly capable operator to meet Nebius’s requirements and structured the transaction. The deal concluded in late 2025 and represents one of the largest dedicated AI infrastructure transactions completed in Europe to date.

France’s Expanding AI Infrastructure Footprint

The Béthune project arrives as France rapidly strengthens its position within Europe’s AI infrastructure landscape. According to data presented by Structure Research, France has emerged as one of the continent’s most active markets for large-scale AI infrastructure development, benefiting from strong government support, substantial nuclear generation capacity and growing interest from hyperscale, neocloud and sovereign AI operators.

The country’s pipeline now extends well beyond traditional markets Paris, and Marseille. Structure Research identified multiple large-scale developments either announced or under construction, including Data4’s proposed 1GW campus in Cambrai, OPCORE and EDF’s 700MW initiative around Paris, Sesterce’s AI deployments, Microsoft expansions, and DataOne and Core42’s Grenoble project. Developments in Béthune, Cambrai, Grenoble, Châteauroux, Calais, and Dunkirk reflect how AI infrastructure is becoming a national story rather than a metropolitan one.

Unlike several neighbouring markets facing power allocation restrictions, France continues to benefit from a substantial nuclear generation fleet capable of supporting large industrial electricity consumers, an advantage that is becoming increasingly valuable as AI developers search for locations capable of supporting hundreds of megawatts of capacity.

Finding the Real Deal

As the leading digital infrastructure marketplace and advisory firm, Inflect operates much differently than a traditional datacenter broker. Where conventional brokers represent one side of a transaction, Inflect operates a two-sided marketplace that connects AI compute operators, neocloud providers, hyperscalers and enterprise customers with datacenter capacity globally, covering deals from 10kW to 1GW+. One of the many transactions Inflect has facilitated is this Datacenter deal between Nebius and Azur Datacenters for the 240MW AI campus in Béthune, Northern France.

As AI infrastructure demand accelerated throughout 2025, Inflect was processing requests from operators, neoclouds and capital sources across the market. Many would never close. Nguyen was direct about what that looked like from the inside.

“Half of them, maybe two-thirds, turned out to be, shall we say, less than real deals, for fundamental reasons,” he said. “Power availability is often talked about, but how ‘fundable’ a project is became a primary driver on why projects failed. There are a multitude of reasons why a project fails or becomes successful. We just have to be ready to move quickly and with conviction to help fill in the gaps.”

That pattern recognition is built into Inflect’s two-sided structure. “We built a marketplace, complemented with our infrastructure advisory,” Nguyen said. “Many of the neoclouds that we represent also sell services in our marketplace. That helps us understand who is buying from them, which gives us insight into what capacity they need and where. It also helps us sell their story to the datacenter providers, and their capital partners. In this market, a customer has to compete with other qualified parties to even get a chance at winning capacity.”

The Nebius project with Azur Datacenters emerged as one of the opportunities with the right elements to move forward. Inflect was working on behalf of a customer with very specific requirements, sourcing the right capacity in the right location with the right operator.

“When it comes to the go-to-market side of this industry, you see a lot of people. But really, deals come down to the right provider, the right power, and a good story for the project to be fundable.” – Mike Nguyen, CEO, Inflect

Azur Datacenters: Building AI Factories from the Ground Up

For Emmanuel Vannier, the Béthune project is the result of more than a decade of evolution, from small edge datacenters to one of the most technically advanced AI infrastructure developments in Europe.

“We have been in the small edge datacenters for a decade,” Vannier said. “We started doing HPC liquid cooling for the largest European operators. Then came the AI factory era, in 2024.”

The negotiations that led to the Nebius agreement were not straightforward. Vannier described 2025 as a year of extensive exploration across multiple potential customers, many of which never progressed to a signed deal.

“Half of them, maybe more, were not real deals; they never went to agreement,” he said. “But we did a real one with Nebius.” Negotiations concluded just before Christmas 2025.

The Execution Problem

AI compute operators at scale cannot wait eighteen to thirty six months for a greenfield facility to clear permitting, secure a grid connection and complete construction. For Nebius, timeline was a hard constraint. When Inflect matched Nebius with Azur Datacenters, a deciding factor was that Azur had already solved the hardest problem: an existing building with power infrastructure and an established grid connection in place.

“Everything was live when we decided to go,” Vannier said. The only permit required was for external cooling equipment, a process that took five weeks.

“Usually, we propose to our customers six, seven or eight months for the first delivery, with two months early access,” he said. The project is being delivered in three phases, with the first tranche available in July.

Why AI Facilities Are Fundamentally Different

Much of the discussion focused on a misconception that Vannier believes remains widespread: the assumption that AI infrastructure can simply be adapted from conventional colocation designs.

“If you look at the power side, the medium voltage, the high voltage infrastructure, it is different,” Vannier said. “If you look at the cooling, the piping, it is vastly different.”

Cooling infrastructure has become a defining design challenge. While liquid cooling has received most of the industry’s attention, Vannier argued that air-cooled network components present equally significant engineering demands.

“Everybody is talking about liquid cooling density,” he said. “But there may be 160kW air cooling racks just for the switches. That is probably more challenging than bringing liquid cooling to the compute.”

The scale of infrastructure required has also driven Azur to look beyond conventional datacenter supply chains entirely. The company is sourcing cooling systems from the oil and gas industry rather than from traditional vendors.

“We are buying from the manufacturers, not from the datacentre equipment vendors,” Vannier said. “My coolers are 2,500 square metres on the ground, twelve metres high.”

“It is a real factory.” – Emmanuel Vannier, Founder, CEO and CTO, Azur Datacenters

Designing for What Comes Next

Vannier’s view is that developers focused on current hardware requirements are already behind.

“My feeling is that most designers are not really understanding what they need to deliver for the next generation of Nvidia hardware,” he said. “They are preparing to just catch up. We are preparing for the future.”

Infrastructure, Vannier argued, must be designed years before the hardware it will ultimately support reaches the market. Operators are already ordering equipment for mid-2027 delivery with a view to being operational by end-2027 or mid-2028, building for hardware generations that have not yet been formally announced.

“Maybe they do not know today, but next year they will ask for the next generation,” he said. “Before you need it, I know what you need: liquid cooling, air cooling, density, heavy racks, medium voltage close to the data hall, battery storage. They just do not know it yet.”

That anticipation extends to the requirements themselves, which continue to shift as AI hardware evolves. “The requirements change all the time,” Vannier said. “You have to be constantly available for the changes.”

Nguyen frames the challenge from the market side. “Perhaps that they are building to today’s standards,” he said. “But what about tomorrow’s?”

Partnership as the Model

For both Nguyen and Vannier, the Béthune project reflects an approach to partnership that distinguishes real transactions from the noise in the market.

Nguyen described Inflect’s role as being in service to the transaction, with proximity to all sides of every deal, understanding buyers deeply enough to represent their story credibly to providers, and understanding providers well enough to identify customers who are serious and well-capitalised.

“This is one of the most extraordinary periods we have seen,” Nguyen said. “We are doing so much at scale, constantly changing, you are hearing about projects delayed or failing everywhere. The key is being open to the challenges, speaking hard truths throughout the process, and making sure your partners understand what they are actually getting into.”

Vannier described a financing model designed to absorb the speed and uncertainty that characterises this market. With a return on investment target of five years, significantly tighter than the twelve-year investment-grade structures common in traditional datacenter development, Azur is structured to move faster and take more calculated risk.

“We are fast moving,” Vannier said. “We will take risk. We can invest ahead of the customer and share a little of that risk, asking for some commitment at the beginning, and then growing together.”

A Test Case for Europe’s AI Ambitions

The Nebius/Azur campus in Béthune is one project among many currently proposed across Europe. But it offers a meaningful benchmark for how the industry is maturing, from announcement to execution, from theoretical capacity to operational infrastructure.

The industry’s traditional assumptions about datacenter design, procurement and deployment are being challenged at every level by AI workloads that demand unprecedented power density, industrial-scale cooling, and fundamentally different engineering approaches.

For Vannier, the shift is clear even to those who have not yet fully processed its implications: “Everything is similar, but everything is different.”

The question now is which developers, operators and advisors understand the difference well enough to execute. As France’s pipeline continues to expand and projects such as Béthune move towards delivery, execution is proving to be the most valuable capability in the emerging market for AI factories.

About Inflect

As a digital infrastructure marketplace and advisory firm, Inflect is vastly different from the traditional datacenter broker. The company connects AI companies, neocloud providers, hyperscalers, and enterprise customers, with datacenter capacity and network services across North America, South America, Europe, and Asia, and facilitates datacenter deals from 10kW to 1GW+.

Inflect’s advisory services cover the full transaction lifecycle: site selection, power procurement, capital structuring, fit-out, and network connectivity. By aligning with all sides of each deal, Inflect provides customers with visibility into qualified supply and helps datacenter operators identify creditworthy, well-capitalised customers competing for scarce capacity.

In 2025, Inflect facilitated the transaction between Nebius and Azur Datacenters for a 240MW AI campus in Béthune, Northern France. Inflect is led by CEO Mike Nguyen and is headquartered in the United States.

Panel Participants

Philbert Shih, Managing Director, Structure Research (Moderator)

Mike Nguyen, CEO, Inflect

Emmanuel Vannier, Founder, CEO and CTO, Azur Datacenters

Media Contact
Company Name: Inflect
Contact Person: Chanyu Kuo
Email: Send Email
Country: United States
Website: www.inflect.com/