Artificial intelligence is one of the most advanced technologies in the world. Fast growing AI technologies, including Google, Microsoft, Meta, Apple and others, are deploying generative AI models. It is also one of the most energy-intensive technologies, and that is raising alarms that AI data centers’ thirst for electricity will increase planet-warming emissions and overload the power grid.
According to one analysis, Google's artificial intelligence consumes ten times more electricity to process a result than a normal Google search. Data centers, or large buildings filled with computer servers, already accounted for approximately 4% of US energy consumption in 2022, Consumption is expected to reach 6% by 2026an increase driven in part by the boom in the use of AI.
And there is a push to build more in California and across the country. Pacific Gas & Electric disclosed in June that it had received 26 applications for new data centers that would use 3.5 gigawatts in total, according to Times reporter Melody Petersen. That amount of energy could power nearly 5 million homes.
Meeting that demand will put pressure on America’s aging power grid, and with 60% of electricity still coming from fossil fuels, it will increase planet-warming carbon emissions. In fact, Google’s carbon emissions rose nearly 50% compared to 2019, which the company attributed to its data center energy consumption and supply chain emissions. Microsoft reported a nearly 30% increase in carbon emissions since 2020 due to data center construction. The International Energy Agency estimates that in 2021, Amazon, Microsoft, Google, and Meta collectively used 72 trillion watt hours — more than double the amount they used in 2017 — and that number is expected to continue to rise.
Data centers operate approximately 100,000 servers on average and often must be located near power plants. There is concern that these facilities could overload local power supplies, causing rolling blackouts. California is particularly vulnerable; the state is located 49th outside of 50 in the ability to avoid blackouts by having more electricity than homes and businesses need during peak times. In addition to being a threat to the power grid, servers produce heat and data centers use a lot of water to cool them.
Given the environmental impacts, there has been growing resistance from communities to plans for data center expansion. Lawmakers in Washington, Virginia, Georgia and other states have pushed for studies on data center energy use and the effects on grid reliability, ProPublica reported. It’s a good start. We need to make sure this rapidly expanding technology doesn’t undermine climate goals.
Jesse Dodge, a senior research scientist at the Allen Institute for AI, said that before this recent wave of consumer AI, AI models were primarily used by researchers for academic purposes and were prioritized for their efficiency and sustainability. Big tech companies developing AI for the market prioritize larger, more power-hungry models.
While tech companies aren’t required to specify how they trained newer models, disclosing this data could help reduce energy demands. Often, if a company is updating a model, it trains it from scratch rather than expanding on a model it has already completed. If companies published these models, that could help eliminate the repetition seen throughout generative AI, which would reduce wasted energy and water, Dodge said.
Shaolei Ren, an assistant professor of electrical and computer engineering at UC Riverside, has been researching ways big tech companies can be responsible when creating new AI models, and he believes these companies are more than capable of operating sustainably.
“In theory, they could physically get to zero carbon if they were to run workloads all over the world. There are data centers all over the world. Because California has solar, they can put the workload somewhere else. [during the day] “They can move the computing to Europe here and at night. They could do it, but they don't because there is too much risk,” Ren said.
Big tech companies have the resources to curb their energy demand, but so far they have chosen not to. For example, users cannot opt out of Google’s AI-generated search, forcing them to get the energy-intensive results even if they don’t want them. That needs to change. AI is only going to get bigger, and the companies behind this boom have a responsibility to ensure their technology doesn’t slow progress in the fight against climate change.
