‘Gold rush’

The artificial-intelligence industry is going through an enormous investment boom. The four major tech firms backing AI development—Microsoft, Google, Meta, and Amazon—have announced AI spending of more than $320 billion for this year alone. That’s more than double the $151 billion they invested in AI just two years ago.

The surge in investment also means a dramatic increase in energy use worldwide. The semiconductor chips used to train AI models and fill AI data centers are especially energy-hungry: They use more energy and give off more heat than most common chips. Today, data centers globally already use as much energy each year as Italy—which has the eighth-largest GDP of any country.

And that dramatic increase is transforming the energy industry. Chevron and Exxon, long among the biggest oil producers globally, recently unveiled plans to build dedicated power plants for AI data centers—a major shift for companies whose main business for a century has been drilling and refining crude oil.

But the surge in energy consumption by AI is also driving an unexpected and worrying rise in planet-warming, greenhouse-gas emissions. Meta’s emissions grew by 66 percent from 2021 to 2023; Google’s emissions rose 50 percent over the past 5 years, mostly because of AI. In West Virginia, coal plants scheduled for closure are now being kept open to supply energy to data centers in Virginia, the global epicenter of AI data hubs. Meanwhile, researchers in California have estimated that the excess pollution from AI’s energy use could lead to an extra 1,300 deaths and billions of dollars in healthcare costs this decade in the U.S. alone.

And AI’s demand for energy looks certain to grow sharply for the foreseeable future: Many economists expect AI energy usage to double by 2030, pushing overall energy consumption in just the U.S. to rise by as much as 20 percent. What’s this mean for the world’s energy use?

Nicholas Kumleben is the energy director at the consulting firm Greenmantle, where he leads research on energy, commodities, and climate change. Kumleben says the core of the energy industry’s transformation is the unexpected bonanza for natural gas. Oil—common in powering cars and home heating—is too dirty and too expensive for generating power at scale, while renewable sources can’t provide the uninterrupted electricity supply that data centers need.

As a result, the world’s biggest tech firms are making massive deals with energy companies to build dedicated data-center power plants that run on natural gas—and these contracts include guaranteed purchases of natural gas for decades to come.

But the climate consequences of all this added consumption of natural gas are extremely uncertain. Tech companies want clean power, and they’ve made commitments to reducing their net greenhouse-gas emissions to zero over time. These long-term power agreements will allow energy firms to develop and deploy carbon-capture technology that could someday halt emissions from these power plants—but they also mean extending and expanding the use of a fossil fuel that now often leaks harmful gases into the atmosphere during its extraction …

Michael Bluhm: Just how much energy do AI data centers use?