The global data center industry faces an unprecedented challenge that could determine which nations emerge as leaders in the artificial intelligence economy. As demand for AI computing power surges, the fundamental constraint isn’t land, capital, or even technology, it’s electrical power infrastructure.
Stephen Beard, Head of Data Centres at Knight Frank, has witnessed this shift firsthand over his decade-long career in the sector. What began as a niche alternative investment class has become a mainstream asset category attracting sovereign wealth funds and institutional capital at new scales.
“In 2015-2016, the average size data center in the UK was somewhere between six and 10 megawatts, costing roughly 80 to 100 million pounds,” Beard explains. “By 2025, the average size data center going forward is about 50 megawatts minimum. We’re selling development sites in London that are 200-300 megawatts. You’re talking about billion-pound investments in singular assets.”
This significant scaling has changed the investment landscape. Where data centers once competed with office buildings and industrial sheds for capital, they now require the backing of top-tier institutional investors and sovereign wealth funds to finance their development.
The Infrastructure Bottleneck
The power challenge stems from the evolution of cloud computing and the recent AI boom. Historically, data centers could source power from local distribution network operators at medium voltage levels. However, as requirements have grown rapidly, developers must now apply directly to national grid operators for high-voltage connections.
“What happened in 2022-2023 is everyone started to bypass the distribution network operators and go straight to National Grid,” Beard notes. “National Grid then got a wave of applications and suddenly said, ‘We can’t facilitate all this demand. We’re going to have to build new substations.’ But it takes seven or eight years to build new substations.”
This infrastructure lag has created a critical bottleneck across Europe. “In 2025, the UK, France, Germany, Spain, and much of Europe is out of power because we’re all waiting for national grid operators to build this new infrastructure,” Beard observes.
The timing couldn’t be worse. AI applications require significantly more power than traditional cloud computing, some estimates suggest a ChatGPT search consumes four to five times the electricity of an equivalent Google search. This increased demand coincides with the infrastructure constraints, creating what industry experts describe as a perfect storm.
AI’s Geographic Implications
The power crisis is changing the geographic distribution of data center development. Traditional data centers clustered around major population centers to minimize latency, the delay between a user’s request and the system’s response. However, AI’s machine learning phase doesn’t require the same immediacy of response, enabling more regional deployment strategies.
“The first phase of AI, the machine learning component, doesn’t need the immediacy of response, and that’s why you’ve seen announcements like Blackstone going up to Northumbria, Microsoft going to Leeds, and other fairly spurious locations,” Beard explains.
However, this regionalization may be temporary. As AI evolves beyond machine learning to inference and eventually predictive applications, proximity to end users will again become critical. “The next phase of AI is inference. As consumers, we need the immediacy of that response. Our view is that these AI requirements need to be where the existing cloud data centers are anyway,” Beard argues.
This evolution suggests that while some AI infrastructure may temporarily migrate to less populated areas with available power, the long-term trend will favor established data center hubs with robust infrastructure and proximity to major population centers.
Geopolitical Implications
The power shortage is creating new geopolitical dynamics in the global data center market. Countries with abundant power resources and streamlined regulatory processes are attracting significant investment from international developers and cloud providers.
The Middle East, particularly the UAE and Saudi Arabia, has emerged as a major beneficiary. “You’re seeing the rise of data centers at scale in Saudi and Abu Dhabi and Dubai. Why? Because we’ve got an abundance of power, an abundance of land, no red tape, and entrepreneurial leadership,” Beard explains.
High-profile meetings, including technology leaders’ visits to Saudi Arabia, underscore this shift. “Look at Donald Trump yesterday in Saudi Arabia, every single major digital infrastructure entity was there. Were they asked to go? No. Were they dragged to Saudi Arabia? Yes.”
European countries are responding with various strategies. France has developed a comprehensive AI strategy plan, while Italy has signed deals with Middle Eastern partners. The UK government has announced ambitious plans to become an “AI superpower,” though concrete financing and implementation details remain unclear.
The Sovereign Cloud Movement
Power constraints are accelerating the development of sovereign cloud platforms as countries seek to reduce dependence on American technology giants. The recent announcement of a German cloud platform, initially misreported as a “Lidl cloud,” represents a broader trend toward data sovereignty.
“It’s really about the sovereignization of data,” Beard explains. “It’s a modern-day Cold War, the weaponization of data and American dominance. Germany, as the most powerful economy in Europe, is saying, ‘Why are we just letting the Americans take all that benefit?’”
This trend extends beyond Europe. China maintains its own cloud ecosystem, while the UAE has developed G42 in collaboration with the US government. The UK has announced Nscale as its sovereign AI platform, though its development timeline remains uncertain.
Investment Implications
For institutional investors, the power crisis presents both challenges and opportunities. Projects with secured power connections command premium valuations, while those without face significant uncertainty.
Knight Frank is currently working on several major UK developments, including a one-gigawatt project in the northwest that could represent a 3 billion investment. However, power connection dates extending to 2030-2031 create substantial execution risk.
“Imagine buying a piece of land which you can develop a data center on in two years, but you’re sitting there waiting for six years for the power, no one is going to do that,” Beard notes.
This dynamic is driving capital toward markets with more reliable power infrastructure and shorter connection timelines, fundamentally altering global investment flows in the sector.
Technical Evolution
The AI boom is driving technical changes within data centers. Traditional cooling systems that rely on air handling units are being replaced by liquid-to-chip cooling systems to handle the higher heat loads generated by AI processors.
“The only difference with AI and cloud is that AI is a higher performance type of compute which requires a different type of cooling,” Beard explains. “Rather than relying on air handling units or chillers to blow air through the white space, we actually need to take liquid direct to the chip.”
This technical evolution requires additional infrastructure investment and specialized expertise, further consolidating the industry around operators with the resources and capabilities to implement these advanced systems.
Looking Forward
The data center industry stands at a critical juncture. The convergence of AI demand, infrastructure constraints, and geopolitical tensions is changing the global landscape in ways that will have lasting implications for technology deployment and economic competitiveness.
Countries that can solve the power equation while maintaining regulatory flexibility and attracting international capital will likely emerge as winners in the AI economy. Those that cannot may find themselves relegated to secondary roles in the digital transformation.
As Beard concludes, echoing Elon Musk’s assessment of the sector: “It’s a thriving sector, a really exciting sector, but without doubt, it’s the most dangerous.” The stakes for technological leadership, economic competitiveness, and data sovereignty have never been higher.
The resolution of the power crisis will ultimately determine not just where data centers are built, but which nations will lead the next phase of the digital economy. For investors, developers, and policymakers, understanding these dynamics is crucial for navigating the complex landscape ahead.
