More AI requires more energy: How the US power grid is affected

Al Seeb/Getty Images

First there were electric cars. And then it's Bitcoin. Now it's AI. All three trends have fueled ongoing concerns about the energy hunger of new technologies. Because they're pushing the shaky US power grid to its limits.

The rise of AI, which is still in its infancy, could be the biggest strain on the nation's power infrastructure.

That's because big tech companies are putting hundreds of billions of dollars into massive data centers housing power-hungry AI-enabled GPU chips that require sophisticated cooling technologies to dissipate the heat generated by computers.

Read it too

AI consulting is the new money printing machine for consulting firms like BCG, Bain and Deloitte.

More AI requires more energy

AI research firm Hugging Face estimates that generative AI searches use 30 times more energy than a traditional Google search. With hundreds of millions of users already interacting with AI tools like Chat GPT, the power demand for AI technologies will increase.

Bank of America examines the challenges facing the power grid as demand for AI data centers grows. “Increasing manufacturing, data centers, artificial intelligence and electrification will put massive pressure on an already strained power grid. Intermittent wind and solar systems cannot provide the energy needed, and tight supplies will lead to higher prices, shortages and outages,” Bank of America said in a recent note.

Here are some eye-opening statistics about the US power grid, from Bank of America:

1. “The US grid generates 1,250 gigawatts (GW) of electricity from 9,200 power plants. Sometimes called “the world's largest machine,” America's power grid is 600,000 miles long. [circa 965.606 Kilometer] Transmission lines are enough to circle the Earth 24 times. The average age of transformers, transmission lines and other grid equipment is 40 to 50 years.
2. “Demand is increasing for the first time in a decade. Over the past ten years, electricity demand has grown by only 0.4 percent per year. The growth rate is expected to be between 2.1 percent and 2.8 percent over the next decade. The expected future demand of 70 GW by 2030 is equivalent to adding the state of Michigan to the grid every year.
3. “Supply is scarce and difficult to expand. No major utility projects are expected before 2026, and 160 GW of fossil fuels have been phased out over the past decade. Regulatory, permitting and political barriers often hinder new energy and mining initiatives. Our colleagues assume that only 55 to 60 GW of capacity will be added in the future.
4. “Wind and solar power have a tough time making up the difference. They run only 24 to 40 percent of the time and produce far less than the 'rated capacity' numbers indicate. Adding batteries adds additional burden: battery storage is 141 times more expensive than LNG, and for every KWh of battery storage, 50 KWh of energy is required to produce it.

There is a lot of investment in AI

Baird CEO and technology strategist Ted Mortensen explained to Business Insider last month just how big a problem AI's energy needs are. “The cost of this general AI architecture is completely out of control. Oracle said in its conference call that they're building 70 megawatt data centers now and going to 200 megawatts. It's as big as a city. So they're very power hungry,” Mortensen said. Oracle said in its earnings release in March. It announced a $10 billion (9.24 billion euros) investment to expand data center capacity to meet the huge demand for AI it creates.

Read it too

AI startup attracts more than 10 million people in first year of founding – with this pitch deck

Amazon learned this fact earlier this year with its decision to buy a nuclear power plant in Pennsylvania for $650 million (€600 million). According to a recent Wall Street Journal report, Amazon's cloud unit is close to a deal with Constellation Energy for electricity supplied directly from a nuclear plant on the East Coast.

This boom in demand has led to a revival in utility stocks, which are up eight percent year-to-date. Goldman Sachs believes this trend will continue. “Investor interest in the AI ​​revolution is not new, but the downstream investment opportunities of utilities, renewable energy and industrial companies, the investments and products needed to support this growth, are underappreciated,” says Goldman Sachs. A report earlier this year.

The bank highlighted four top utility stocks to buy. These include Xcel Energy, Next Era Energy, Southern Co. and Sempra included. “US electricity demand is likely to experience growth not seen in a generation. “Since the turn of the century, U.S. electricity demand has not increased by 2.4 percent in eight years, while annual U.S. electricity production has grown by an average of less than 0.5 percent over the past 20 years,” Goldman Sachs said.

Read it too

Despite oil giant Aramco's stake sale: Major financial problems at Saudi mega project Neom