Massive energy consumption of AI is raising concerns among ESG investors
In Mount Pleasant, Wisconsin, land is currently in high demand. At the end of July, tech giant Microsoft made another purchase, securing over 70 hectares (about 173 acres) of land in the small community of 26,000 residents, paying nearly 34 million dollars. This deal follows the acquisition of numerous other nearby parcels in Racine County, illustrating the tech giants' arms race for dominance in artificial intelligence.
This is where one of the many new data centers is being built to support Microsoft’s cloud division, Azure, which aims to provide the massive amounts of data and computing power needed by tech firms like OpenAI, the developer of ChatGPT, for training their large language models. However, as the demand for technological infrastructure grows, so does the need for resources at an exponential rate.
The International Energy Agency estimates that the electricity consumption of AI-focused data centers will reach 90 terawatt-hours by 2026, up from six terawatt-hours in 2022. Additionally, consumption in traditional data centers is expected to increase from 345 to 576 terawatt-hours. This trend raises concerns among consumer advocates about the source of this enormous amount of electricity. The tech giants themselves provide an answer that many market participants find alarming: nuclear power.
About one-third of US nuclear power plant operators are currently in talks with cloud companies about supply agreements for electricity to fuel the AI boom. According to insiders, Amazon Web Services is close to a deal with Constellation Energy, which would allow the company to source electricity directly from a nuclear plant on the East Coast. In March, AWS already acquired a nuclear-powered data centre campus in Pennsylvania for 650 million dollars. This trend has sparked interest among stock investors in long-neglected US utilities.
Sustainability investors fear that the tech giants are not only neglecting to drive the expansion of green energy themselves, but are also draining stable resources from the entire grid, which could then be replaced by fossil fuels like natural gas. As per Climate Action Against Disinformation, a coalition of environmental groups, the AI boom could drive global greenhouse gas emissions up by 80%. Consumer initiatives also fear rising electricity prices.
Higher efficiency through innovation
Proponents of the technology argue that ongoing technological innovation could significantly reduce energy demand. Nvidia's current Hopper graphics processors are already 25 to 30 times more efficient than previous chip generations, and upcoming new semiconductors based on the so-called Blackwell architecture are expected to be even more efficient. New software developments also play a role in energy consumption, with language models like those from OpenAI expected to be tailored to more specific functions, thus using less power. Additionally, AI could handle traditional computing tasks more efficiently, creating a positive substitution effect.
But the scale of the AI boom is difficult to predict, despite forecasts like that of the International Energy Agency. Big Tech's capital expenditures exceeded Wall Street expectations in the last quarter by a wide margin. The share of Amazon, Microsoft, Alphabet, Apple, Meta Platforms, Nvidia, and Tesla in S&P 500 capital expenditures was 18% in 2023, up from 5% a decade ago. Analysts emphasise that the share could be even higher this year, based on management teams' plans.
Aggressive managers
Bank of America predicted as early as this spring that major cloud service providers’ capital expenditures would reach 180 billion dollars in 2024. Some voices on Wall Street are now even suggesting that the bank may need to revise this forecast upward. In addition to publicly traded tech giants, private AI-focused cloud providers like CoreWeave have been active recently, securing large-scale financing from private equity firms such as Blackstone.
In any case, managers in Silicon Valley are being aggressive. In a market situation like the current one, „investing too little is a dramatically bigger risk than overinvesting“, said Alphabet CEO Sundar Pichai recently. Amazon CEO Andy Jassy has stated that he wants „even more capacity“ despite already large investments, while Microsoft CFO Amy Hood emphasises that the rapidly increasing spending on AI and cloud infrastructure should „support monetization over the next 15 years and beyond.“
Intensified water shortages
It’s not just the explosion in electricity consumption that’s worrying environmentalists. The tech giants also need significantly more water to cool their data campuses. Data centres are „the tenth-largest water consumer in the US, and ChatGPT consumes a litre of water for every 40 queries“, comments ESG strategist Alex Kusen of Deka.
The tech sector is thus contributing to global water demand outstripping current population growth by a factor of 1.7. Water is becoming the most valuable and scarce resource of all time. Sustainability investors now hope that the tech giants’ thirst for water and hunger for electricity can indeed be met through innovation.