Opinion: AI is here. Get ready for a spike in your electric bill

Opinion by Jessica Kuntz and Lauren Kuntz

(CNN) — For the past two decades, more efficient energy generation has balanced out small increases in America’s energy demand. But the days of steady electricity consumption are over.

McKinsey, Boston Consulting Group and S&P Global all forecast that US power demand will grow between 13% and 15% annually for the rest of the decade. Compared to recent decades, that kind of increase is meteoric — and far outstrips the capacity of US electricity generators.

One of the main drivers behind this exploding demand for power? Artificial intelligence.

Beyond the power to charge your laptop, most of us aren’t used to thinking of computers as particularly energy intensive. But when you type a prompt into ChatGPT (or another large language model), your request is processed in a data center far, far away — and generating that response demands power.

Training each model is also energy intensive: A massive volume of data, scrapped from webpages, Wikipedia, Reddit and transcribed YouTube videos are poured into these AI models. To process this data, hundreds of graphics processing units — electronic circuits capable of performing rapid mathematical calculations — run continually for thousands of hours. And all of that requires electricity — gigawatts upon gigawatts of electricity, on a scale that makes previous data center usage appear quaint.

This is compounded by a national decarbonization effort founded on shifting many sources of energy consumption — vehicles, heating, manufacturing (supercharged by the CHIPS and Science Act and Inflation Reduction Act) — to an electric grid powered by clean energy sources.

So as AI’s demand on the grid surges, and as cars and industry shift to the grid, that means a significant portion of generators in the US grid remain “on” around the clock — and prices surge.

Since energy demand vacillates throughout the day — as well as over the course of the year — the grid needs to provide a flexible supply, ramping up when everyone is home in the evening, but also when temperatures spike and air-conditioning use surges. Most of the energy generation we rely on day to day is inexpensive, around $30 per megawatt.

But when demand exceeds the supply of energy from these base power plants, utilities turn on peaker plants, which are designed to ramp up quickly, but make for very expensive power generation, in the range of $1,000 per megawatt. As consistently higher electricity demand — driven by AI models, electric vehicles and growing manufacturing footprints —keep these high-cost plants “on” more of the time, costs for everyone — households, schools, hospitals — will rise exponentially.

The upshot: Energy costs using existing generation aren’t linear. A 15% increase in electricity demand doesn’t lead to a 15% price increase, as utilities rely increasingly on high-cost peaker plants to meet demand. Exponential cost increases means that a doubling of home energy prices is well within the realm of possibility. Beyond cost, if the grid is working at full capacity just to keep up, we can’t justify taking dirtier sources of energy generation offline — delaying progress toward climate goals.

Within the decade, we’re looking at a grid infrastructure that can’t keep pace with rapidly rising demand, even with all sources running around the clock. This pushes us toward a scenario unimaginable for most Americans — an unreliable energy grid, marked by blackouts and rolling brownouts. Recall the blackouts across Texas in winter 2021 and the 2003 East Coast blackout that impacted 50 million people. But this time, it won’t be a stroke of bad luck linked to extreme weather or a safety incident. Electricity rationing will become a standard part of American life while the grid races to add capacity.

The Biden White House recently made a commendable effort to focus the country’s attention on the growing vulnerability and insufficiency of our national grid, launching the Federal-State Modern Grid Deployment Initiative. The problem diagnosis is correct, but the initiative’s near exclusive focus on augmenting existing infrastructure via grid-enhancing technologies — and absence of major plans to expedite and fund new energy infrastructure — doesn’t match the scale of the challenge.

What’s more, the wait time for approvals for new power generation to be connected to transmission networks recently hit five years, partly because of insufficient transmission of the existing grid. Presently, states through which a power transmission line runs must approve it. This process often includes consultation with property owners in the vicinity and factors in economic, environmental, safety, historical preservation and engineering concerns — making for a deeply cumbersome process.

A truly national grid demands a singular federal oversight body. Such a consolidation of authority is proposed in the Streamlining Interstate Transmission of Electricity (SITE) Act. This, in concert with a nationwide push to build and finance new electricity generation and transmission — on the scale of New Deal era construction — is needed to meet the energy demand of an AI-powered, green economy.

America’s ability to deliver on its aspirations in AI leadership, decarbonization, chips manufacturing and electric vehicles hinges on abundant electricity — but even an AI chatbot can’t provide a workaround to cumbersome bureaucratic processes that imperil grid upgrades. For that, we need policy aimed at getting new power generation built and online — and fast.

The-CNN-Wire
™ & © 2024 Cable News Network, Inc., a Warner Bros. Discovery Company. All rights reserved.