Europe’s AI paradox: big ambitions on an overloaded power grid

AI may seem limitless, but the infrastructure that powers the technology is anything but. At the recent IOWN Global Forum event, FUTURES Amsterdam, an international summit where over 170 tech companies are designing the next generation of AI infrastructure. Speakers outlined the dilemma currently facing Europe’s IT infrastructure. Vito Mabrucco, the event’s master of ceremonies, set the tone from the outset in his welcome address: “We are now beginning to realise that we are running up against hard physical limits. AI cannot continue to evolve if power grids are overloaded. That is simply the new reality.”

Two different perspectives emerged on how Europe can push beyond these limits. Stijn Grove, managing director of the Dutch Data Center Association, advocates a different approach to spatial planning and energy supply. Dr. Katsuhiko Kawazoe, president of the IOWN Global Forum, sees the solution in the All-Photonics Network, which uses light rather than electricity to process and transmit data.

‘We’re sold out’

Space is a scarce resource. “When you look at data centre capacity worldwide, and particularly in Europe and the Netherlands, we are simply sold out,” says Stijn Grove. “Demand is many times greater than supply.” This shortage is evident in Frankfurt, for example, where available capacity has fallen from a healthy vacancy rate of 10 to 12 per cent to just 3 per cent.

The power grid is another obstacle. Europe accounts for 15 per cent of global data centre electricity consumption, yet capacity will need to triple over the coming years to remain internationally competitive. However, permits and grid connections are making that growth difficult. “When I started at the industry association, obtaining a permit took a maximum of one year. Now it takes five. In the past, you could get electricity from the grid straight away; now you have to wait ten years. That is becoming the norm almost everywhere in Europe,” Grove explains.

According to Grove, the solution lies in a radical rethink of spatial planning. “Electricity has become the new location factor. It is the real bottleneck holding back the development of digital infrastructure and AI.” Data centres should therefore no longer be built primarily in overcrowded cities, but in locations where renewable wind power comes ashore. He also advocates the creation of dedicated “AI zones” with permits arranged in advance, following the examples of France and Finland.

Smart system integration is also crucial. Equipped with enormous battery storage facilities, data centres should act as active buffers for the power grid. When wind power temporarily drops, they can switch to their own reserves, immediately easing pressure on the grid.

Computing with light

While Grove outlines the limits imposed by physical space and electricity, Dr. Katsuhiko Kawazoe offers a solution rooted in the underlying architecture of the technology itself.

“Enormous computing power is required to process large volumes of data. As a result, energy consumption is now going through the roof.” The way forward is photonics. Traditional IT infrastructure relies on electricity to process information, generating substantial heat and energy losses. Switching to optical networks can largely overcome this problem. “Optical technology has a wonderful advantage over electricity,” Kawazoe explains. “Even as transmission distances or frequencies increase, power consumption barely rises.”

Since NTT (Nippon Telegraph and Telephone, the Japanese telecommunications company that founded IOWN) invented the first optical transistor in 2019, the IOWN Global Forum has been working with more than 170 companies to develop a radically new infrastructure. As Kawazoe made clear during his presentation, the targets are ambitious: “Compared with today’s conventional networks, we are aiming for 100 times greater energy efficiency, 125 times more transmission capacity, and 200 times lower latency.”

The challenge

Europe therefore faces a twofold challenge: its physical infrastructure must expand to keep pace with growing demand for AI, while new technologies must reduce energy consumption. Mabrucco aptly summed up the need for action: “Ambition is crucial, but we must address these constraints in practice now — through photonics, decisive policymaking and our power grids. We are currently part of a global revolution.”