Big tech companies want more computing power. Much more. According to their latest quarterly reports, Alphabet (Google's corporate parent), Amazon and Microsoft (the global cloud computing giants) together invested $40 billion between January and March, most of it in data centers equipped to manage growing artificial intelligence (AI) workloads. Last month, Meta, which does not have a cloud business but does run a data-hungry social media empire, stated that its capital spending could reach $40 billion this year as a result of AI-related projects. . This figure is not far from the $50 billion that the oil giant Saudi Aramco plans to spend. Microsoft is likely to spend more.
The comparison with the energy industry, famous for its ease of capital expenditure, is timely not only because of the sums at stake. AI requires massive amounts of processing power. And that processing power requires huge amounts of electricity. On May 2, Bob Blue, CEO of Dominion Energy, one of the largest utilities in the United States, stated that data center developers are now asking him for “several gigawatts” (GW) on a regular basis. Dominion's total installed capacity is 34 GW.
The JPMorgan Chase bank estimates that, in 2022, the cloud arm of Amazon (AWS), Alphabet, Meta and Microsoft will consume 90 terawatt-hours (TWh) of electricity, as much as Colombia. And all of that, for the most part, before ChatGPT sparked the AI revolution in November of that year. The ensuing furor led the International Energy Agency (IEA), an official forecaster, to predict that data centers (including those dedicated to AI and cryptocurrencies, which are also energy-hungry) will consume more than 800 TWh in worldwide in 2026, more than double that of 2022 (see chart). The consulting firm BCG estimates that, in the United States, data processing could triple its share of energy consumption in 2030 and reach 7.5%.
And not just any energy will do. The technological titans want theirs to be clean. In April, its industry association warned Georgia Power, which had managed to fast-track 1.4 GW of new generation powered by fossil fuels citing growing data center demand, that its members would build fewer centers in the state. American southerner if the power company emitted more carbon. Combined with growing demand for increasingly electrified transport, heating and some heavy industry sectors, the energy needs of digital technology are putting enormous pressure on companies that generate and distribute electricity.
The Bloomberg agency estimates that the annual investment needed in the network to completely decarbonize global electricity in 2050 will have to go from about $300 billion in 2022 to $600 billion in 2030 and well over $800 billion in 2050. Risk-averse utility companies, which often undertake grid expansion projects under the watchful eye of cost-conscious regulators, have neither the money nor the desire to do so.
That's where the big tech companies come in. These deep-pocketed giants have already been the biggest force behind the green “power purchase agreements” that helped kick-start the U.S. renewable energy boom by convincing utilities and other investors to They will build wind and solar parks. Now they are entering the field of green energy more directly.
On May 1, Microsoft and Brookfield, one of the world's largest infrastructure investors, announced an agreement to build 10.5 GW of renewable energy capacity in the United States and Europe by 2030. The agreement aims to allow the software giant deliver on its promise to have 100% of its electricity consumption come from zero-carbon sources 100% of the time by 2030. Microsoft and Brookfield haven't disclosed pricing, but adding a gigawatt of wind or solar capacity can cost 1 billion dollars.
One of the problems is that data centers tend to consume energy at a constant rate, even when the sun is not shining or the wind is blowing. Therefore, technology companies are also thinking about ways to make data processing more flexible. In March, Sidewalk Infrastructure Partners, an infrastructure and technology company created jointly by Alphabet, presented a detailed plan to achieve that goal. It consists of a combination of microgrids (able to operate independently, but also to exchange energy with others nearby), batteries and advanced software that allows less urgent tasks, such as training AI models, to be moved to periods of lower demand. Jonathan Winer, one of Sidewalk's founders, expects such data centers to appear first in energy-constrained places like Arizona, California and Massachusetts.
Renewable energies are not the only area of energy interest for big technology companies. In March, AWS paid $650 million for a 960-megawatt (MW) data center in Pennsylvania powered by a nearby nuclear reactor. Microsoft has reached an agreement with Constellation Energy, the largest nuclear operator in the United States, to supply nuclear power to its Virginia data center, as a backup when wind and solar are not available. Both companies have also been studying “small modular reactors,” a promising but unproven nuclear technology.
Google, for its part, is making forays into geothermal energy. The search giant has signed the first-ever corporate deal to develop “enhanced” geothermal energy with Fervo, a startup that has raised $430 million in venture capital. Inspired by the shale industry, this pioneering geothermal energy company has developed horizontal wells, monitored using fiber optic cables. Its center in Nevada produces 24-hour, carbon-free power for the local grid, which Google later acquires. Tim Latimer, head of Fervo, says each drilling rig operated by his company can add 100 MW of energy. The company is building a 400 MW commercial plant in Utah that will begin feeding the grid in 2026. The US Department of Energy estimates that innovations like Fervo's could multiply national geothermal production by 20, to more than 90 GW, by 2050 .
Google and Microsoft have also partnered with Nucor, a giant American operator of electricity-intensive mini-steel factories. In March, the trio announced that they will aggregate demand and jointly offer contracts to clean energy projects, whether they are early-stage commercial projects or entirely new “pioneer” initiatives. The idea is to secure clientele for promoters of promising technologies, such as long-duration energy storage, clean hydrogen, next-generation geothermal energy and nuclear energy.
The AI industry's most exotic power plays come courtesy of Sam Altman, the techno-optimistic CEO of OpenAI, maker of ChatGPT and Microsoft's main modeling partner. In his quest to drive the AI revolution, he has invested in Helion, a nuclear fusion startup, and Exowatt, a startup that develops solar modules that can act as both electricity generators and thermal storage batteries. Altman is now seeking to raise $500 million for Oklo, a company that creates nuclear microreactors that run on spent fuel from larger reactors and could power individual factories, corporate headquarters and, of course, AI server farms. Those bets may seem fanciful. However, 18 months ago the idea that an AI could write articles or paint like a human being also seemed like that.
____________________________________________________________________
© 2024 The Economist Newspaper Limited. All rights reserved
Translation: Juan Gabriel López Guix
