Data Centers: Future-Proofing and Sovereignty in the Age of AI

Digital technology is a key component of growth and innovation, and with important infrastructure consequences. The supply and demand of data centers are expected to growth 10-15% annually due to the surge in data consumption, notably from AI and cloud services.

Indeed, according to the IEA's 2025 Energy and AI report, data centre electricity consumption is set to more than double to around 945 TWh by 2030, which is slightly more than Japan’s total electricity consumption today. In addition to this are challenges due to water consumption, temperature, and location.

At Infraday 2025, crucial questions were brought forward: How can we ensure data sovereignty while integrating resource constraints and access to energy and water, all while maintaining proximity to users, the holy grail for financial markets and other algorithm users?

Introducing the segment by connecting to the previous panel on geopolitical impacts, Thomas Reynaud, CEO of Iliad, the fifth largest telecom operator in Europe, argued that it is not too late for Europe to achieve digital sovereignty, illustrating the current challenges and what needs to be done in his keynote speech followed by a Q&A with Rebecca Smith from Natixis CIB's Green & Sustainable Hub.

Thomas Reynaud stressed the fundamentally strategic aspect of digital sovereignty, as it covers how we get informed and connected to the world. It is a relevant issue for Iliad which has operations in Poland and the Baltic states, regions identified by Jean-François Robin in the previous panel as primary targets should Russia go to war with Europe.

"Energy and data are evolving into diplomatic tools", he noted, and telecom infrastructure is already under threat, as evidenced by cable cuts and cyberattacks. Beyond that, he emphasised that AI raises a burning question for civilisation: "AI models impose the worldview of those who train them. If we do not control our infrastructure, we will not be able to defend our values," he noted.

Significant concerns therefore surround data storage locations and access to General Purpose (GP) computing resources. Iliad's response first consists in keeping its cloud services free from extra-European regulations. “For Scaleway, our cloud provider subsidiary, sovereignty relies on two fundamental principles: immunity to extraterritorial legislation and independence in the underlying technology stack”. He advocates for establishing a purely European GP ecosystem while maintaining ties with US and Chinese technology through cooperation.

Regarding another key topic of the day AI, Reynaud explained that we need four key elements: brain power, data centers, clean and competitive energy, and Graphic Processing Units (GPUs). The main pain point is GPUs, for which we must maintain good international cooperation. He remembered that, since 2023, the Group has also invested in the largest AI compute capacity available in the commercial market in Europe.

Regarding energy and water requirements, Thomas Reynaud emphasized the critical access to grid connectivity, particularly the speed of execution for connections to power distribution networks : he particularly welcomed the introduction in France of a "fast track" for electricity connection.

Access to clean, but also competitive, energy is key. “The cost of AI is closely tied to energy prices, which presents a significant challenge for the sector in Europe”, he noted. When access to the grid is saturated, he favors proximity to usage locations with data centers powered by captive power plants.

What does he expect from infrastructure investors? Long-term commitment and a deep understanding of their business. From public authorities, he calls for public procurement rather than subsidies, along with incentives to bring “the best European engineers back from the West-Coast to develop European AI”.

Continuing the discussion on the challenges associated with ongoing shifts in the market structure, particularly in light of AI disruption, Robert Wallin, Managing Director, Telecom & Tech Industry Group and Laëtitia Fournier, Head of Data & Innovation, invited Ralph Cho, Co-CEO of Apterra Infrastruture Capital; Richard Bienfait, Chief Financial Officer of Stack Infrastructure; and Alexander Oyaert, Chief Investment Officer of Data4, to address these pressing issues from both lender and operator perspectives.

So far, the data center industry has been fueled by exponential data consumption alongside high revenue predictability. Multibillion USD/EUR developments are being agreed upon and financed, supported by long-term contracts - often spanning 10, 15, or even 20 years - with investment-grade hyperscale tenants like Amazon, Google, Microsoft, META, and Oracle.

Thousands of data centers have indeed emerged worldwide, particularly in the United States, which processes more than 40% of global data, leading to a fivefold increase in capacity over the past 20 years.

With both supply and demand expanding at an impressive rate of 10-15% annually, a crucial question arises: how much capacity can be absorbed in the long term? Is there a risk for a glut in the market over time, as some of the older data centers are coming to their contractual end. How far will hyperscalers choose to extend?

Furthermore, and the question is exacerbated by the explosion of AI: will future needs align with those of today? Can existing data centers be repurposed to meet changing demands? As, for instance, we know that the Graphic Processing Units (GPUs) required for AI usage are heavier and denser than traditional cloud servers, requiring enhanced cooling systems and more robust land foundations.

One major concern for the AI sector is, at this point, the return on massive investments, considering that the equipment inside an AI Data Center could be 10 times the cost of the value of the infrastructure, compared to up to 3 times for typical Cloud Services. What if AI developers should not utilize all this capacity over time? Many smaller developers of AI applications are relying on NeoClouds and other non-IG offtakers for GPUs as a service. How do DC investors look at the business opportunity beyond the IG hyperscalers?