AI boom thrusts Europe between power-hungry data centers, environmental goals

The boom in artificial intelligence is driving an environmentally conscious shift in the way data centers operate, with European developers using more energy to keep up with the high-power chips from companies like the tech giants. Facilities are facing pressure to lower water temperatures. Nvidia.

AI is Demand increased by 160% Specialized chips used by AI companies are expected to increase energy use in the data centers that deploy them, according to Goldman Sachs research, an increase that could come at the expense of Europe’s decarbonization goals. There is sex.

High-performance chips, also known as graphics processing units (GPUs), are essential for training and deploying large-scale language models, a type of AI. These GPUs require dense computing power, generate more heat, and ultimately require chilled water to support reliable cooling of the chips.

According to Andrei Korolenko, chief product and infrastructure officer at Nebius, AI can consume 120 kilowatts of energy in just one square meter of a data center, which is equivalent to the power consumption of about 15 to 25 homes. This corresponds to the amount of heat dissipated. In preparation for the introduction of Nvidia’s Blackwell GB200 chip.

“This is very dense and requires a different solution from a cooling standpoint,” he said.

Michael Winterson, president of the European Data Center Association (EUDCA), warned that falling water temperatures would eventually lead to a return to the “fundamentally unsustainable situation of 25 years ago”.

“The problem that chip manufacturers are facing is [that] “AI is currently a space race led by the American market, where land rights, energy access, and sustainability are relatively low on the list of priorities, and where market dominance is key,” Winterson told CNBC. .

Major equipment suppliers in Europe are demanding that U.S. chip designers lower water temperatures to accommodate higher-temperature AI chips, said Herbert Radlinger, managing director of NDC-GARBE.

“Initially, everyone on the engineering side was thinking about liquid cooling being adopted to operate at higher temperatures,” he told CNBC, referring to liquid cooling technology, which is said to be more efficient than traditional methods. “This is shocking news, as I expected it to be.” Air cooling.

Energy efficiency is high on the agenda of the European Commission, which aims to achieve its goal of reducing energy consumption by 11.7% by 2030. The EU predicted in 2018 that data center energy consumption could increase by 28% by 2030, and the advent of AI is expected to increase that number. 2x, 3x In some countries.

Winterson said the drop in water temperatures is “fundamentally incompatible” with the EU’s recently launched Energy Efficiency Directive, which establishes a dedicated database for data centers of a certain size to publicly report their electricity consumption. ” he said. EUDCA has been lobbying Brussels to take these sustainability concerns into account.

Energy management company Schneider Electric frequently collaborates with the EU on this topic. Steven Carlini, chief advocate for AI and data centers and vice president at Schneider Electric, said much of the recent discussion has focused on different ways to source “prime power” for AI data centers and how they work with utility companies. He said he is focusing on the possibility of further cooperation.

The European Commission’s energy representatives have also been in contact with NVIDIA to discuss energy consumption and data center usage with respect to power usage and chipset effectiveness.

CNBC has reached out to NVIDIA and the committee for comment.

“In data centers, cooling is the second largest energy consumer after IT loads,” Carlini told CNBC in an emailed comment. “Energy usage will increase, but low water temperatures may not increase PUE (power usage efficiency), even if the chiller is running at a high rate.”

Schneider Electric’s customers, which are deploying Nvidia’s Blackwell GB200 superchip, are looking for water temperatures of 20 to 24 degrees Celsius, or 68 to 75 degrees Fahrenheit, Carlini said.

He added that this is comparable to about 32 degrees Celsius for liquid cooling, or about 30 degrees Celsius, which Meta suggests for the water temperature it supplies to its hardware.

Ferhan Gunen, Vice President, UK Data Center Operations Equinixtold CNBC that there are a number of concerns about AI that Equinix is ​​discussing with customers.

“They want to increase the density of their servers, which means they want to put more power-hungry chips in them, or they want to put more servers in them,” he said. He added that it was not something like that.

“This is really more of an evolutionary discussion than anything else,” Gunen said.

Nvidia declined to comment on the chip’s cooling requirements, but announced A new platform for Blackwell GPUs was released earlier this year. The company says this architecture enables organizations to run real-time generative AI on large language models at up to 25 times lower cost and energy consumption than previous technologies.

Liquid cooling will require “reconfiguration,” Gunen explained, adding that new data centers are already being prepared to implement the technology. “Yes, higher density means more power consumption and higher cooling requirements. But the technology is changing, so the way we do things is changing. That’s why there’s a balance between all of this.” she said.

Nevius has about $2 billion in cash on its balance sheet after splitting from Russia’s Yandex. said The company will be one of the first to offer Nvidia’s Blackwell platform to customers in 2025. We also announced plans to Invest over $1 billion above AI infrastructure will be in place in Europe by the middle of next year.

Nebius’ Korolenko said liquid cooling is a “first step” and will initially worsen the cost of ownership until it improves over time.

“There’s a lot of pressure to deliver, but at the same time, if you’re going to scale, you need the ability to make choices that are economical and don’t sacrifice too much. Power efficiency is important for running costs. “It’s a high priority,” Korolenko said.

Even before the surge in demand for AI applications, the European data center industry was struggling to keep up with the growth of the digital sector.

Sicco Boomsma, managing director of ING’s TMT team, said market participants are “very sensitive to electricity” and that while Europe is focused on infrastructure, the U.S. is looking to expand its assets in Europe where electricity is available. He said he is placing more emphasis on

“A huge number of data center operators are also coming in from the US, working together to ensure their data center infrastructure is in line with various EU goals, such as carbon neutrality and efficiency. ”, water use, and maintaining biodiversity.

“This is the kind of race where they want to prove that their knowledge is leading to super-efficient infrastructure,” he said.