© Mark Ollig
Artificial Intelligence (AI) is transforming industries, yet the energy required to fuel it is often overlooked.
The electrical energy consumed in data centers powering AI platforms is enormous, pushing the limits of the nation’s energy infrastructure.
Data centers are needed to house the expansion of the digital network, which includes cloud computing platforms, data storage and processing, online content delivery, e-commerce, the Internet of Things (IoT), and the expanding use of AI applications.
The building of data centers is only going to increase. Data center energy consumption is on the rise due to AI-powered services from major tech companies like OpenAI (ChatGPT), Meta (Meta AI), Microsoft (Azure & Copilot), Google (Gemini), and Amazon Web Services (AWS), which is raising alarms about the stability of the nation’s electrical grid.
According to the U.S. Department of Energy, 70% of the nation’s transmission lines are over 25 years old, and the average age of large power transformers is over 40 years old.
This growing demand underscores the urgent need for modernization of our aging grid infrastructure and expanded renewable energy use.
AI platforms housed in hyperscale data centers require substantial computational resources and energy, even when in an idle state and not actively processing tasks.
Data centers need to operate continuously to support digital services’ always-on nature, which further increases their energy demands.
AI models generally consume more energy than data retrieval, streaming, and communication applications, which have been the mainstay of data center expansion in the last twenty years.
For example, one Google search consumes an estimated 0.0003 kilowatt-hours (kWh) of energy, while one ChatGPT AI query reportedly uses 10 to 50 times more energy.
As of March, ChatGPT’s user base has surpassed 180.5 million, with an estimated 600 million monthly visits to use its services.
The Electric Power Research Institute’s (EPRI) 2024 report predicts that integrating the large AI language models that Google plans to incorporate into its search engine could increase annual electricity consumption by 22.8 to 29.2 terawatt-hours (TWh).
EPRI estimates that by 2030, data centers could consume up to 9.1% of total U.S. electricity generation due to increasing AI demand.
“Today, fifteen states account for 80% of the national data center load, with data centers estimated to comprise a quarter of Virginia’s electric load in 2023,” states the 2023 U.S. Department of Energy report, “Powering Intelligence: Analyzing Artificial Intelligence and Data Center Energy Consumption.”
According to Statista, a leading provider of market and consumer data, as of March 2024, the U.S. led the world with 5,381 data centers, followed by Germany (521) and the United Kingdom (514).
In 2023, Minnesota had 45 data centers, 37 in the Twin Cities metro area, consuming an estimated 824,316 megawatt-hours (MWh) of electricity, about 1.24% of our state’s total energy consumption.
Meta, formerly known as Facebook, recently announced plans to build an $800 million, 715,000-square-foot data center in Rosemount, Minnesota.
The AI-focused data center will occupy 280 acres of land at UMore Park, purchased from the University of Minnesota.
This data center, Meta’s 19th in the U.S., is planned to open in late summer 2025 and will support platforms such as Facebook, Instagram, Threads, and WhatsApp.
Meta said it is committed to powering the data center with 100% renewable energy.
The National Renewable Energy Laboratory of the U.S. Department of Energy reported in 2023 that Minnesota is among the top 10 states for data center growth.
The 2023 Minnesota Energy Factsheet reported a 33% increase in energy productivity from 2001 to 2023.
In 2023, renewable energy sources made up 33% of Minnesota’s electricity generation, with the total installed capacity reaching 6.8 GW.
Our state consumed 66 terawatt-hours (TWh) of electricity in 2022.
Increased data center construction across the country will establish energy demands that some fear could overload local power grids, especially in densely populated areas.
A 2023 McKinsey & Company report projects total U.S. data center power consumption in 2022 was 17.5 GW and could reach 35 GW by 2030.
On April 28, 2024, Goldman Sachs projected that the growing presence of AI models is expected to boost global data center power demand by 160% by the year 2030. The U.S. EIA’s 2023 Annual Energy Outlook report projected the following U.S. energy sources being used in 2024:
Natural gas: 37%
Nuclear: 18%
Coal: 17%
Wind: 12%
Solar: 9%
Hydropower: 3%
Biomass: 1%
Other: 3%
Although renewable energy usage is rising, coal and natural gas still account for 54% of U.S. electricity generation.
Some may remember when Northern States Power’s (NSP) Reddy Kilowatt told us electricity was “penny-cheap.”
In 2022, the national average residential electricity cost was 14.57 cents per kWh, with commercial businesses paying 12.55 cents and industries 11.66 cents.
In 2022, Minnesota’s average consumer electric rate was 11.98 cents/kWh.
A 2023 Datacenter Dynamics article stated that by 2027, AI-dedicated data centers could consume as much electricity as the Netherlands.
“Advances in energy-efficient computing will be critical… The energy demand of training increasingly large foundation [AI] models is not sustainable,” said the U.S. Department of Energy’s 2023 report, “Advanced Research Directions on AI for Energy,”
Will the US electric grid handle AI’s growth?
Stay tuned.
 |
| Future $800M data center which will open in Rosemount, MN Photo courtesy by META |