How Big Tech Wants to Satisfy AI’s Energy Thirst

The significant increase in the number of data centers around the world shows no signs of slowing down, forcing Big Tech to consider how best to support the artificial intelligence revolution.

Some of the options on the table include going nuclear, liquid cooling in data centers, and quantum computing.

But critics say big tech companies must recognize the costs of power generation as efficiency gains in electricity use slow. AI Boom across the entire supply chain and abandon the “move fast and solve the problem” narrative.

“The real environmental costs are currently quite hidden – they are simply subsidized by the fact that technology companies have to buy and agree to the products,” says Somya Joshi, Head of Global Agenda, Climate and Systems in Stockholm. The Environmental Institute (SEI) told CNBC via video call.

According to the International Energy Agency, the wave of data center investments is expected to accelerate in the coming years, largely driven by digitalization and the use of generative AI.

It is this prospect that has raised concerns about AI’s often overlooked but very important environmental impact as well as a surge in electricity demand.

Data centers, consuming increasingly large amounts of energy, represent the core infrastructure for modern cloud computing and AI applications.

Giampiero Frisio, president of electrification at Swiss multinational ABB, said the engineering group’s data center business has enjoyed significant growth in recent years and the segment is expected to grow by more than 24% in 2024.

Frisio said ABB is well-positioned amid the AI ​​demand boom to supply mid-sized and large enterprise industry players with all the components they need to run their data centers.

“I think the best way to act now is to increase energy efficiency. This is the best way, because we have technologies, for example, medium-voltage HiPerGuard UPS. We can do it and we can do it tomorrow morning,” Frisio said. . CNBC via video call.

The HiPerGuard UPS refers to ABB’s industry-first medium-voltage uninterruptible power supply, which is said to be capable of providing continuous power to large facilities.

“The second is to move to liquid cooling. There’s no doubt about it. Again, it’s more energy efficient. Why? You see a single rack, a black box that looks like a closet with all the servers in it? The power density inside will be four to six times what it was before,” Frisio said.

He added, “We are talking about 5 to 10 years later, and it is a nuclear modular system.”

US tech giants Microsoft, Google and Amazon have signed billions of dollars worth of nuclear energy contracts in recent months as they seek to bring additional energy capacity online to train and run the large-scale generative AI models behind today’s applications. .

The surge in demand for generative AI coincides with efforts to find more efficient cooling solutions in data centers, especially liquid cooling (a process that uses water to lower the temperature of servers and other electronic equipment).

French power equipment manufacturer Schneider Electric recently completed an $850 million deal to acquire a controlling stake in Motivair Corp, a US-based company specializing in liquid cooling for high-performance computing.

At the time, Schneider Electric’s CEO told CNBC that the all-cash deal, designed to boost services to data centers, was “rich but not overly expensive” and “fits nicely” with the company’s strategy.

Along with nuclear energy and liquid cooling technology, some technology companies have suggested that developments within AI could help decarbonize data centers.

For example, former Google CEO Eric Schmidt said last month that “we’re not going to meet our climate goals anyway,” so investing in AI could play a pivotal role in solving our biggest environmental problems. .

SEI’s Joshi emphatically rejects this view.

“These claims are not new. They are very consistent with the kind of ‘panacea’, ‘technology will save us’ rhetoric,” Joshi said.

“It is inherently contradictory to say that we will somehow solve the problems we currently face by operating within certain finite planetary boundaries but exceeding them and continuing the same extractive narrative,” she added.

“I think every great technology we discover has a winter in the summer, but don’t pay attention to it until winter comes,” Raj Hazra, CEO of Quantinuum, the world’s largest integrated quantum computing company, told CNBC via video call. said. .

“This is how we talk about generative AI, the infrastructure needed to support it, and the large data centers that need to be built.”

Hazra said optimism about the generative AI boom is already straining the costs of operating the technology.

“As good as AI is, there are two challenges: Is it sustainable from a resource perspective? Is it responsible?” Hazra said. “The reason I mention this context is because it is very important for both.”

Quantum computing refers to a field of computer science that uses the laws of quantum mechanics to solve very complex problems.

Hazra said a “who’s who” of companies and strategic investors have recently expressed interest in Quantinuum, and the company recently raised about $300 million in an equity funding round. Honeywell, the company’s majority shareholder, said the financing values ​​Quantinuum at $5 billion.

“One of the things that has become very clear is that it is no longer right to say there is a solution to the problem. We have to say there is a sustainable solution to the problem,” Hazra said.

The CEO said one of quantum’s greatest contributions to society is making AI sustainable and responsible.

“I expect that in the next three to five years, you’ll see people saying, what is the computing infrastructure to run my business? It’s going to be a combination of high-performance computing, AI and quantum,” he added.