- The artificial intelligence boom is putting a strain on America's power grid.
- AI queries can use 30 times more energy than traditional searches, dramatically increasing electricity demands.
- “The costs of that generation of AI architecture are spiraling out of control,” said Ted Mortenson, a strategist at Baird.
First it was electric cars. Then Bitcoin. Now it's AI.
These three trends have raised ongoing concerns about the magnitude of the power-hungry new technologies that are pushing America's shaky power grid to its limits.
The AI boom, still in its early stages, appears to be one of the biggest stressors for the nation's power grid.
That's because big tech companies are spending hundreds of billions of dollars on power-hungry AI-enabled GPU chips that are installed in giant data centers that require cutting-edge cooling technology to dissipate the heat generated by the computers.
AI research company Hugging Face estimates that generative AI search queries can consume 30 times more energy than a traditional Google search.
And with hundreds of millions of users already using AI tools like ChatGPT, the power demand for AI technology will only grow.
Bank of America outlined the challenges the power grid faces as it deals with a surge in demand from AI data centers.
“Driven by manufacturing, data centers, artificial intelligence and electrification, we expect to add massive demand to an already strained power grid. Intermittent wind and solar power will not be able to provide the needed electricity, and tight supplies could lead to higher prices, bottlenecks and outages,” Bank of America said in a recent note.
Some startling statistics about the U.S. power grid cited by Bank of America include:
- “The U.S. power grid produces 1,250 gigawatts (GW) of electricity from 9,200 generating units. The U.S. power grid has been called the “world's greatest machine,” with 600,000 miles of power lines that circle the Earth 24 times. The average lifespan of transformers, transmission lines, and other grid equipment is 40 to 50 years.
- “Demand is increasing for the first time in 10 years. “Over the past decade, electricity demand has grown by just 0.4% per year. Growth over the next decade is expected to be 2.1% to 2.8%. Projected demand of 70 GW by 2030 is like adding the size of Michigan to the power grid every year.”
- “Supplies are tight and it’s hard to add more. “No large utility projects are scheduled through 2026, and 160 GW of fossil fuel supply has been shut down over the past decade. Regulatory, permitting and political obstacles often thwart new energy and extractive efforts. Our colleagues expect only 55-60 GW of capacity to be added in the near future.”
- “Wind and solar are struggling to make up the gap. They only operate 24-40% of the time and generate much less power than their rated capacity numbers would suggest. Adding batteries adds an extra burden: battery storage is 141 times more expensive than liquefied natural gas and it takes 50KWh of energy to generate 1KWh of battery storage.”
Ted Mortenson, a managing director and technology strategist at Baird, told Business Insider last month how AI's power demands are a big problem.
“The costs of generative AI architectures are spiraling out of control. Oracle basically said on a conference call that they're building 70-megawatt data centers and will expand to 200 megawatts, which is the size of a city. So they're very power hungry,” Mortenson said. In a March earnings call, Oracle announced it would invest $10 billion in expanding its data center capacity to meet the massive demand for generative AI.
Amazon realized this earlier this year, leading to its decision to buy a nuclear power plant in Pennsylvania for $650 million.
According to a recent report from the Wall Street Journal, Amazon's cloud division is close to a deal with Constellation Energy to get electricity directly from nuclear plants on the East Coast.
The surge in demand has sparked a revival in utility stocks, with the sector surging 8% so far this year and Goldman Sachs seeing the rally continuing.
“While investor interest in the theme of the AI revolution is not new, we believe downstream opportunities in utilities, renewable energy generation and industrial, where investments and products will be needed to support this growth, are undervalued,” Goldman Sachs said in a note earlier this year.
The bank highlights four utility stocks to buy, including Xcel Energy, NextEra Energy, Southern Company and Sempra.
“U.S. electricity demand is likely to experience growth not seen in a generation. Never since the beginning of this century has U.S. electricity demand grown by 2.4% over an eight-year period, and over the past two decades, annual U.S. electricity generation has grown by an average of less than 0.5%,” Goldman Sachs said.

