AI Is Here. Get Ready For A Spike In Your Electric Bill

Over the last two decades, more efficient energy output has offset slight increases in America’s energy demand. But the days of constant electricity use have passed.

McKinsey, Boston Consulting Group, and S&P Global predict that US power demand will increase between 13% and 15% per year for the rest of the decade. Compared to former decades, this level of growth is meteoric, greatly exceeding the capacity of US electrical generators.

One of the primary drivers of this skyrocketing demand for power? Artificial Intelligence.

Aside from the power required to charge your laptop, most do not consider computers very energy-intensive. However, when you enter a prompt into ChatGPT (or any large language model), your request is processed in a faraway data center, and generating that response requires power.

Training each model also requires a lot of energy: These AI models receive large amounts of data scraped from websites, Wikipedia, Reddit, and transcribed YouTube videos. To process this data, hundreds of graphics processing units—electronic circuits capable of conducting rapid mathematical calculations—run continuously for thousands of hours each. And all of this necessitates electricity — gigawatts upon gigawatts of electricity, on a scale that makes prior data center usage seem archaic.

AI | Emarsys Image

AI Is Here. Get Ready For A Spike In Your Electric Bill

This is exacerbated by a national decarbonization program aimed at moving numerous energy-consuming sources—automobiles, heating, and manufacturing (bolstered by the CHIPS and Science Act and the Inflation Reduction Act)—to an electric grid powered by clean energy sources.

So, as AI’s demand on the grid grows and automobiles and industry transition to the grid, a major number of generators in the US grid remain “on” around the clock, causing costs to rise.

Because energy demand varies throughout the day and year, the grid must provide a flexible supply, ramping up when everyone is home in the evening and when temperatures rise and air-conditioning usage increases. Most of the energy we use daily is inexpensive, costing roughly $30 per megawatt.

When demand exceeds the electricity supply from these base power plants, utilities turn on peaker facilities, designed to ramp up quickly but generate very expensive power at $1,000 per megawatt. As rising electricity consumption — driven by AI models, electric vehicles, and expanding manufacturing footprints — keeps these high-cost plants “on” more of the time, expenses for everyone — families, schools, and hospitals — will rise dramatically.

As a result, energy costs with existing generation are non-linear. A 15% increase in power demand does not result in a 15% price increase, as utilities increasingly rely on expensive peaker facilities to meet demand. Because of exponential cost increases, household energy bills may double. Beyond cost, if the system is operating at maximum capacity to keep up, we cannot justify shutting down dirtier types of electricity generation, so we are postponing progress toward climate targets.

Within a decade, we’ll have a grid infrastructure that can’t keep up with fast-expanding demand, even if all sources are operational 24 hours a day. This propels us toward an inconceivable prospect for most Americans: an unreliable electrical grid characterized by blackouts and rolling brownouts. Consider the blackouts in Texas in the winter of 2021 and the 2003 East Coast blackout, which affected 50 million people. But this time, it won’t be a stroke of bad luck caused by extreme weather or a safety issue. Electricity rationing will become a common feature of American life as the system expands.

The Biden White House recently launched the Federal-State Modern Grid Deployment Initiative, a commendable effort to draw attention to the country’s rising vulnerability and insufficiency in our national grid. The problem diagnosis is true, but the initiative’s near-exclusive focus on complementing current infrastructure with grid-enhancing technology and the lack of major plans to expedite and fund new energy infrastructure falls short of the enormity of the challenge.

AI | Net Image

AI Is Here. Get Ready For A Spike In Your Electric Bill

Furthermore, the wait time for approvals for additional power generation to be connected to transmission networks has lately reached five years, partly due to insufficient transmission capacity on the current grid. Currently, the states that pass a power transmission line must authorize it. This process frequently involves communication with nearby property owners and considers economic, environmental, safety, historic preservation, and engineering concerns, making it time-consuming.

A fully national grid requires a single central supervisory authority. The Streamlining Interstate Transmission of Electricity (SITE) Act proposes consolidating authority. This, combined with a nationwide push to create and fund new electrical generation and transmission on the scale of New Deal-era projects, is required to meet the energy demand of an AI-powered, green economy.

America’s capacity to meet its goals in AI leadership, decarbonization, chip manufacturing, and electric vehicles is dependent on ample electricity. Yet even an AI chatbot cannot provide a workaround for tedious bureaucratic processes that jeopardize grid upgrades. To do this, we need policies that encourage the construction and operation of new power plants as soon as possible.

SOURCE – CNN