Strained power grids: the hidden cost of the AI ​​boom

Artificial intelligence is as hot a topic as one can imagine. From better search results to more efficient power management, AI can do it. However, the power demand from AI data centers has become a hot topic in its own right. Squaring the AI ​​circle is getting harder and harder.

In 2023, Dutch researcher Alex de Vries calculated the potential global energy consumption of artificial intelligence, which put the technology in the energy spotlight. “You’d be talking about the size of a country like the Netherlands in terms of electricity consumption. You’re talking about half a percent of our total global electricity consumption,” de Vries. said BBC at the time.

Now, it turns out that de Vries may have been too optimistic. The Wall Street Journal recently reported that in Ohio, a single power company faces “three New York cities with data centers asking to connect to the grid” after 2028, when electricity demand in the region will double — because of the same data centers.

The rush to deploy AI for multiple purposes is in some ways like the first US shale boom, when companies drilled just because they could, demand be damned. To be fair, AI is being touted as a solution to almost any efficiency problem a business may have, regardless of industry. When it comes to energy, however, there is an irony to the AI ​​story. Artificial intelligence can help make energy consumption more efficient while depleting the available supply.

Related: U.S. oil demand shines as China eases

In August of this year, a researcher from the University of Texas at Austin wrote in a piece to The Conversation that his team developed an AI system that could change a building’s energy consumption during times when there was more wind and solar power on the grid.

“The system can learn on a set of buildings and occupants and can be used in buildings with different controls and energy use patterns,” wrote Zoltan Nagy, whose work received funding from the Electric Power Research Institute and the another NGO, Climate Change AI. He explained that the AI ​​system looks for the best times to charge the home’s batteries, allowing the household to continue using electricity as needed, regardless of the state of the grid.

This certainly sounds like something quite useful if it is not limited to households with home batteries. Most homes, however, don’t have storage, so they’re effectively competing with AI developers like Nagy’s team and server vendors for a limited amount of electricity.

This is becoming a real headache for power utilities, the WSJ wrote, citing the case of Phoenix, which is experiencing a boom in production and has also become an AI data center. The city will run out of transmission capacity by 2030, the Arizona Public Service has calculated, making it urgent to build and upgrade about 800 miles of pipeline over the next ten years.

Utilities’ expectations of an increase in electricity demand served pressed their higher inventories, bullish forecasts for building new power generation capacity – particularly gas and nuclear – and prompting some to issue warnings about capacity growth. The latest comes from Fitch Ratings, which said utilities may be overestimating future data center demand “given the inconsistent ways the industry calculates future demand,” according to the Wall Street Journal.

The current surge in demand from data centers handling artificial intelligence is far from inconsistent, or so it seems. According to the WSJ report, data center developers are still looking for sources of electricity as competition between them heats up as electric vehicle charging providers and new manufacturing facilities are built with funding under the Inflation Reduction Act.

This is perhaps the greatest irony when it comes to artificial intelligence and the energy transition. Proponents of the premium argue it would help the transition by adjusting electricity demand to supply variation as the grid becomes more reliant on wind and solar.

However, AI data centers themselves demand so much electricity that wind and solar can’t handle the load, so developers are looking to lock in future supply from grassroots sources like natural gas and nuclear. Microsoft did headlines Last week, with news that it struck a deal to reopen the Three Mile Island nuclear power plant to power its data centers.

Artificial intelligence was listed as one great reason for the increase in electricity demand in the United States this year — and the subsequent increase in the supply of electricity from gas and coal. The other major reason was population growth.

For now, that population appears to be protected from the negative effects of this growing imbalance between electricity supply and demand – at least those with long-term, fixed-price contracts. However, if the increase in demand that the electricity companies are predicting, based on what they are currently witnessing, materializes, it would likely affect everyone who uses electricity sooner or later. And it could ensure the long-term survival of both gas and coal generation, despite efforts to phase them out in favor of wind and solar power.

By Irina Slav for Oilprice.com

More top reads from Oilprice.com