Is AI and Data Center Growth Fueling an Energy Crisis?
The pace of innovation in artificial intelligence (AI) since ChatGPT’s launch in 2022 has been nothing short of staggering.
In terms of user adoption, ChatGPT is the fastest-growing app in history. As of mid-2025, it had about 800 million weekly active users and handled more than 2.5 billion prompts per day globally.
And ChatGPT is hardly the only AI-powered chatbot or tool.
The top 5 tech companies in the US are investing hundreds of billions of dollars in AI technology and data center growth.
Meta, Amazon, Microsoft, Google, and Apple all have plans to build new data centers to fuel AI-powered growth and integration into virtually all of their existing products, like Microsoft CoPilot and AI Overviews in Google Search.
AI-powered startups like Anthropic (Claude), Perplexity, and Frontier AI (Mistral) are also growing rapidly.
Many consumers perceive large language models (LLMs) as frictionless and mostly free technologies that make it easier to do their jobs or generate memes.
However, everything comes at a price, even if it’s not visible in plain sight.
The “big compute” that powers LLMs requires massive resource consumption that already impacts grid reliability, business and consumer energy bills, drinking water, and the environment.
And it’s going to get worse before it gets better.
AI Drives a Surge in Data Center Electricity Demand
According to a recent IEA report, global data centers consumed around 415 TWh of electricity in 2024 (about 1.5% of total global demand).
Consumption is projected to more than double by 2030 to around 945 TWh, with AI identified as the primary driver of this growth.
A September 2025 study found that “The rapid expansion of large‐scale AI data centers is imposing unprecedented demands on electric power grids. With immense electricity consumption subject to large and fast fluctuations, these facilities introduce emerging impacts and operational challenges for power grids.”
How Does AI Data Center Demand Threaten Power Grid Reliability?
In March 2025, Reuters published an exposé about a “near miss” event in Data Center Alley, located close to Washington, DC suburb Fairfax, VA, that almost caused a widespread power outage.
Data Center Alley is home to multiple facilities serving Microsoft, Google, Amazon, and many other companies. According to local officials, “about 70% of the world's internet traffic flows through the area.”
It covers a 30-square mile area with 200+ data centers that currently consumes roughly the same amount of electricity as Boston.
In Summer 2024, 60 data centers dropped off the utility grid suddenly and switched to backup generators.
The drop was triggered by the industry equivalents of uninterruptible power supplies (UPS) designed to protect sensitive electronics and data storage devices, which can easily be permanently damaged or destroyed by fluctuations in voltage.
By disconnecting from the utility grid so suddenly, the data centers triggered “a huge surge in excess electricity, according to federal regulators and utility executives.”
Grid operator PJM Interconnection and local utility Dominion Energy were forced to quickly cut back power plant output to “protect grid infrastructure and avoid a worst-case scenario of cascading power outages across the region.”
John Moura, Director of Reliability Assessment and System Analysis for NERC, told Reuters, “As these data centers get bigger and consume more energy, the grid is not designed to withstand the loss of 1,500-megawatt data centers. At some level, it becomes too large to withstand unless more grid resources are added.
Also interviewed for the article, Alison Silverstein, a former senior adviser to the chairman of the U.S. Federal Energy Regulatory Commission, said, “What this event tells us is that the behavior of data centers has the potential to cause cascading power outages for an entire region.”
As the electricity demands of AI grow, there are many new projects in development to add more capacity to the utility grid or dedicated entirely to providing power for the data center.
Project | Company/Operator | Location | Power Capacity (MW) | Estimated Cost (USD) | Status |
Stargate (Phase 1) | OpenAI / Oracle / SoftBank (JV) | Abilene, TX | ≈200 MW (Phase 1); campus up to ~1.2 GW | ≈$1.1B (Phase 1); long-term campus >$500B projected | Under construction / announced |
Quantum Loophole | Quantum Loophole | Frederick County, MD | ≈1.8 GW (planned) | Unavailable | Under development; facing delays (governance issues) |
Vantage 'Mega‑Campus' | Vantage Data Centers | Shackelford County, TX | Up to ~1.4 GW (reported) | ≈$25B (multi‑site expansion) | Announced / planned |
CoreWeave AI Data Center | CoreWeave | Lancaster, PA | 100 MW initial; expandable to 300 MW | ≈$6B | Announced / planned |
Utah AI Data Center | CIM Group / Novva | West Jordan, UT | ≈175 MW | ≈$2B | Under construction |
Soluna 'Project Kati' | Soluna | Willacy County, TX | 166 MW planned; 35 MW Phase 1 funded | ≈$20M (Phase 1) | Announced (Phase 1 under development) |
xAI 'Colossus' (Grok) | xAI | Memphis, TN | 150 MW allocated (Phase 1) | Unavailable | Operating / expanding |
Vantage – San Antonio Development | Vantage Data Centers | San Antonio, TX | ≈96 MW | ≈$277M | Announced / planned |
Microsoft 'Fairwater' AI Data Center | Microsoft | Pleasantville, WI | Unavailable | ≈$3.3B (initial facility) | Announced; opening targeted 2026 |
Microsoft – Wisconsin Expansion | Microsoft | Mount Pleasant, WI | Unavailable | ≈$7B (announced 2025) | Announced / planned |
(Sources: See Citations)
Unfortunately, power plants take years to construct, and in the meantime, data center power demand is putting severe strain on aging electrical grid infrastructure, which contributes to blackouts and higher electricity bills.
How Does AI Resource Consumption Impact Your Electricity Bills
Residential electricity prices (cents per kWh) vary radically by location, particularly from state to state.
Top 5 Highest/Lowest Residential Electricity Rates by State (June 2025)
Rank | Category | State | Average Rate (¢/kWh) |
1 | Highest | California | 33.52 |
2 | Highest | Massachusetts | 30.37 |
3 | Highest | Maine | 28.14 |
4 | Highest | Rhode Island | 26.84 |
5 | Highest | New York | 26.53 |
1 | Lowest | Nevada | 11.42 |
2 | Lowest | Idaho | 12.07 |
3 | Lowest | Louisiana | 12.64 |
4 | Lowest | Washington | 12.98 |
5 | Lowest | Texas | 15.23 |
(Source: Energy Information Administration)
California is the most expensive state for household electricity at 33.52¢/kWh, while Nevada is the cheapest at 11.42¢/kWh — almost 3 times less.
Similarly, how impacted your electricity bills are likely to be by growing data center demand will vary considerably based on where you live.
Top 10 U.S. States by Data Center Share of Electricity Use (2023)
Rank | State | Data Center Consumption (2023, MWh) | Percentage of Total Consumption |
1 | Virginia | 33,851,122 | 25.59% |
2 | North Dakota | 3,915,720 | 15.42% |
3 | Nebraska | 3,959,520 | 11.7% |
4 | Iowa | 6,193,320 | 11.43% |
5 | Oregon | 6,413,663 | 11.39% |
6 | Wyoming | 1,857,120 | 11.26% |
7 | Nevada | 3,416,707 | 8.69% |
8 | Utah | 2,562,037 | 7.68% |
9 | Arizona | 6,253,268 | 7.43% |
10 | Washington | 5,171,612 | 5.69% |
(Source: EPRI 2023, via Quartz)
As of 2023, the state with the largest share of electricity used by data centers was Virginia, home of Data Center Alley, which accounted for 26% of the state's total annual electricity consumption.
It stands to reason that the states with the highest share of data center electricity consumption would be at the greatest risk of price increases and power outages, but it’s not quite that simple.
As illustrated in the map above, populous states like California and Texas are at high risk of increased electricity rates and power outages due to AI-fueled data center growth.
Utility Dive reports that in July 2025, the Electric Reliability Council of Texas (ERCOT) called the “disorganized integration” of large loads, like data centers, the biggest growing reliability risk facing the Lone Star State’s electric grid.”
ERCOT is one of the largest electricity providers in the country, serving over 26 million customers and powering over 90% of the state.
In California, where one in five households served by the state’s largest investor-owned utilities are behind on their electricity bills, /kWh rates have skyrocketed in recent years.
According to CalMatters, “residential rates spiked 63% for Pacific Gas & Electric customers, 52% for Southern California Edison customers, and 13% for San Diego Gas & Electric customers between 2021 and 2024 and are now the highest in the US (on average).
While a conclusive link between data center growth and the rising electricity rate is difficult to prove conclusively, many California legislators believe the correlation between surging demand from AI and soaring household power bills is clear.
Assemblymember Rebecca Bauer-Kahan of San Ramon introduced a measure that “would require data centers and the developers of large AI models to publicly share how much energy they use.”
If passed, the measure could have global implications, because right now, nobody outside of the tech corporations that own the LLMs knows how much electricity it takes to train the models or to operate them.
How Much Power Does AI Consume?
Up until now, the actual energy consumption of training and operating general-use large language models has not been disclosed by OpenAI, Google, or any of the other GenAI giants.
Source | Scope | Estimate | Notes |
Wired (2025) | Inference per ChatGPT query | ~0.34 Wh per query | Based on Altman’s statement; scales to very large totals if multiplied by billions of queries. |
MIT Tech Review / MIT News (2025) | General context on training + inference energy | Inference could dominate long-term; training of GPT-3 estimated ~1,287 MWh (older figure) | Highlights lack of transparency for GPT-4/5; cites academic estimates for smaller models. |
Wired (2024/2025) | Global AI electricity use | ~20% of total DC demand in 2024. Could reach ~82 TWh annually | Estimate based on hardware production, GPU shipments, and usage assumptions. |
Digiconomist / de Vries (2023–2024 studies) | AI workloads globally | Tens of TWh annually (range 20–80 TWh depending on adoption) | One of the first widely cited attempts to quantify AI electricity consumption globally. |
Numerous reputable media sources, like Wired and MIT Technology Review, have attempted to infer the resource consumption and costs of training and operating LLMs by reverse-engineering the process, but the tech giants have so far refused to confirm or deny the reporting.
AI Data Center Growth: Environmental and Quality of Life Impacts
Unfortunately, rising electricity costs and decreased power grid resilience aren’t the only negative impacts of AI’s voracious growth and consumption of resources.
Skyrocketing data center demand also impacts the water we drink and the air we breathe.
AI Data Center Impacts on Human Health and Quality of Life
The Unpaid Toll: Quantifying the Public Health Impact of AI is a research paper published by Caltech and UC Riverside (UCR) scientists in December 2024.
It paints an alarming picture of the potential impact of AI demand and data center growth on human life.
Here are some direct pull quotes from the report:
Premature Deaths & Asthma Burden
“In 2030, the scope-2 pollutants of U.S. data centers alone could cause, among others, approximately 600,000 asthma symptom cases and 1,300 premature deaths, exceeding 1/3 of asthma deaths in the U.S. each year.”
Economic Toll
“Under McKinsey’s projection with a medium-growth scenario, the U.S. data centers in 2030 could contribute to nearly 1,300 deaths annually, resulting in a public health burden of more than $20 billion which could even exceed that of on-road emissions of California.”
Car Travel Comparison
Training an AI model of the Llama-3.1 scale can produce an amount of air pollutants equivalent to driving a passenger car for more than 10,000 LA-NYC round trips, resulting in a health cost that even exceeds 120% of the training electricity cost.”
Disproportionate Impact On Poor Communities
“Low-income counties … could experience per-household health burdens equivalent to nearly 8 months of electricity bills and more than 200× compared to that in other counties.”
Backup Generator Pollution
“If the backup generators in northern Virginia (Data Center Alley) emit air pollutants at the maximum permitted level, the total public health cost could reach $2.2–3.0 billion per year … impacting residents in multiple surrounding states and as far as Florida.”
If the Unpaid Toll and other reputable studies are correct, AI data center growth may come at a significant cost to human life, not to mention human labor through potential job losses.
AI Data Center Impacts on Carbon Emissions and Climate Change
The One Big Beautiful Act phased out most federal incentives for residential and utility-scale renewable electricity generation projects on an accelerated schedule, ending many six years earlier than initially planned.
Many utility-scale wind and solar farms designed to ramp up electricity generation from clean renewable energy sources now face much higher costs, and even projects that have already started planning and construction are shutting down.
With far fewer terawatts of generation capacity from renewables now expected to go online in the foreseeable future, electricity generated by burning fossil fuels like oil and coal is bound to increase.
Much public policy in the US has recently sought to dispel fears and data about the impacts of carbon emissions and global climate change.
However, many other countries, government organizations, academic institutions, and scientists disagree.
The most recent Intergovernmental Panel on Climate Change (IPCC) report found:
It is now established (very high confidence) that some weather and climate extremes have become more frequent and/or intense due to human influence (greenhouse gas emissions), especially temperature extremes, heatwaves, heavy precipitation, and coastal flooding.
Global average temperature has increased, and many changes are already irreversible over centuries to millennia (e.g. sea level rise).
Without rapid, deep, and sustained reductions in greenhouse gas emissions, the risks of more severe, widespread, and consequential extreme weather (e.g. more frequent heatwaves, drought, heavy rainfall, tropical cyclones) increase sharply.
Final Thoughts
It remains to be seen whether AI will be humankind’s savior, destroyer, or somewhere in between
AI evangelists and AI doomers have very different forecasts of the future, and many prominent CEOs and scientists, including Sam Altman, Daniel Hinton, Yann LeCun, and Elon Musk, alternate between utopian and dystopian scenarios regularly.
One thing is certain, unless we’re in a massive AI bubble, energy demand for data center applications is almost certain to continue increasing at a dizzying pace.
Newer, more efficient AI models that are less energy intensive may emerge and efforts to increase electricity generation capacity are under way…
However, in the short to mid-term, grid instability and power outages are widely expected to increase.
One of the best ways to protect yourself and your family from rising electricity costs and energy insecurity is to invest in a whole-home backup system like EcoFlow DELTA Pro Ultra.
With multiple charging options, including grid and solar power, it can keep your house up and running almost indefinitely during outages.
Built-in smart-energy management tools and optional accessories can also help you reduce your electricity consumption and dependence on the grid.
EcoFlow has a wide variety of off-grid portable power and residential home backup systems.
Check out our selection today.
Resources Cited
AI. “Energy Demand from AI – Energy and AI – Analysis - IEA.” IEA, 2025. https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai?
Chen, Xin; Xiaoyang Wang; Ana Colacelli; Matt Lee; and Le Xie. “Electricity Demand and Grid Impacts of AI Data Centers: Challenges and Prospects.” arXiv:2509.07218, 10 Sep 2025. https://arxiv.org/pdf/2509.07218?
Nguyen, Britney. “10 States Where Data Centers Use the Most Electricity.” Quartz, December 2, 2024. https://qz.com/us-states-data-centers-electricity-use-share-ai-tech-1851708857?
Eia.gov. “Electric Power Monthly - U.S. Energy Information Administration (EIA),” 2025. https://www.eia.gov/electricity/monthly/epm_table_grapher.php?
McLaughlin, Tim. “Big Tech’s Data Center Boom Poses New Risk to US Grid Operators.” Reuters, March 19, 2025. https://www.reuters.com/technology/big-techs-data-center-boom-poses-new-risk-us-grid-operators-2025-03-19/?utm_source=chatgpt.com.
Thompson, Ramona. 2019. “Data Center Alley: Computing Power of 10 Million Servers at Plug-And-Play Price - Loudoun County Economic Development, VA.” Loudoun County Economic Development, VA. August 15, 2019. https://biz.loudoun.gov/computing-power/.
DatacenterDynamics. “xAI’s Colossus Memphis Data Center Gets 150MW Power Allocation from TVA.” November 2024. https://www.datacenterdynamics.com/en/news/xai-colossus-memphis-power-tva/.
DatacenterDynamics. “CoreWeave Plans $6bn Data Center in Lancaster, Pennsylvania.” July 2025. https://www.datacenterdynamics.com/en/news/coreweave-plans-6bn-data-center-in-lancaster-pennsylvania/.
NextBigFuture. Wang, Brian. “OpenAI Stargate Phase 1 Construction of 200 Megawatts and 980,000 Square Feet.” May 2025. https://www.nextbigfuture.com/2025/05/openai-stargate-phase-1-construction-of-200-megawatts-and-980000-square-feet.html.
R&D World. “Stargate’s $500B Bet Could Force Data Center and 1.2 GW Grid Rethink.” June 2025. https://www.rdworldonline.com/stargates-500b-bet-could-force-data-center-and-1-2-gw-grid-rethink/.
DatacenterDynamics. “Quantum Loophole Faces Governance Uncertainty as TPG Acquisition Collapses.” February 2025. https://www.datacenterdynamics.com/en/news/quantum-loophole-faces-governance-uncertainty-as-tpg-acquisition-collapses/.
DatacenterDynamics. “Vantage Data Centers Confirms $25B Expansion in Texas.” April 2025. https://www.datacenterdynamics.com/en/news/vantage-data-centers-confirms-25b-expansion-in-texas/.
Novva Data Centers. “CIM Group and Novva Data Centers Announce $2 Billion Expansion in Utah.” Press Release, October 2024. https://novva.com/press/cim-group-and-novva-data-centers-announce-2-billion-expansion-in-utah.
DatacenterDynamics. “Soluna Secures $20M for First Phase of Project Kati Development in Texas.” July 2024. https://www.datacenterdynamics.com/en/news/soluna-secures-20m-for-first-phase-of-project-kati-development-in-texas/.
Datacentre Magazine. “Where Is Soluna’s Biggest Sustainable Data Centre to Date?” September 2024. https://datacentremagazine.com/news/where-is-solunas-biggest-sustainable-data-centre-to-date.
Reuters. Shepardson, David. “Microsoft Boosts Wisconsin Data Center Spending to $7 Billion.” September 18, 2025. https://www.reuters.com/business/microsoft-boosts-wisconsin-data-center-spending-7-billion-2025-09-18/.
Wall Street Journal. “Inside Microsoft’s Plans for the Most Advanced AI Data Center in the World.” September 2025. https://www.wsj.com/articles/inside-microsofts-plans-for-the-most-advanced-ai-data-center-in-the-world-ee50bd4c.
Walton, Robert. “Data Center Activity ‘Exploded’ in Texas, Spiking Reliability Risks: Monitor.” Utility Dive, July 14, 2025. https://www.utilitydive.com/news/data-center-activity-has-exploded-in-ercot-spiking-grid-reliability-risk/752780/?
Johnson, Khari. “Crackdown on Power-Guzzling Data Centers May Soon Come Online in California.” CalMatters, February 18, 2025. https://calmatters.org/economy/technology/2025/02/data-center-crackdown-to-protect-california-electricity-rates/?
https://www.publicadvocates.cpuc.ca.gov/-/media/cal-advocates-website/files/press-room/reports-and-analyses/240119-caladvocates-q4-2023-quarterly-rate-report.pdf
Caretto, Theo. “If Done Correctly, Californians Can Save on Utility Bills by Creating Income-Based Charges.” CalMatters, February 7, 2024. https://calmatters.org/commentary/2024/02/utility-bills-reform-income-based/.
Digitaldemocracy.org. “AB 222: Data Centers: Power Usage Effectiveness: Cost Shifts.” 2025. https://calmatters.digitaldemocracy.org/bills/ca_202520260ab222.
Taft, Molly. “How Much Energy Does AI Use? The People Who Know Aren’t Saying.” WIRED, June 19, 2025. https://www.wired.com/story/ai-carbon-emissions-energy-unknown-mystery-research/
Zewe, Adam. “Explained: Generative AI’s Environmental Impact.” MIT News | Massachusetts Institute of Technology, January 2025. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117?
Taft, Molly. “AI Is Eating Data Center Power Demand—and It’s Only Getting Worse.” WIRED, May 22, 2025. https://www.wired.com/story/new-research-energy-electricity-artificial-intelligence-ai/?
Digiconomist. 2025. “Research - Digiconomist.” Digiconomist. May 24, 2025. https://digiconomist.net/research/.
Han, Yuelin, Zhifeng Wu, Pengfei Li, Adam Wierman, and Shaolei Ren. “The Unpaid Toll: Quantifying the Public Health Impact of AI.” arXiv.org, 2024. https://arxiv.org/abs/2412.06288?
Vlamis, Kelsey. “Anthropic CEO Says Warning World about AI Replacing Jobs Is Necessary.” Business Insider, September 18, 2025. https://www.businessinsider.com/anthropic-ceo-warning-world-ai-replacing-jobs-necessary-2025-9?utm_source=chatgpt.com.
“AR6 Synthesis Report: Climate Change 2023.” 2023. Ipcc.ch. IPCC. 2023. https://www.ipcc.ch/report/ar6/syr/.