AI Is Fueling a Data Center Electricity Surge. But the Grid Isn’t Ready
The rapid rise of artificial intelligence, fueled by large language models like ChatGPT, is driving an unprecedented surge in electricity demand from data centers across the United States. This growing energy appetite is placing increasing strain on the nation’s aging electric grid, leading to declining power quality, heightened risks of outages, and new challenges for residential infrastructure. As AI adoption accelerates, the need for a smarter, more resilient grid has become urgent. This article examines the scale of the impact, the vulnerabilities it exposes, and the steps utilities, regulators, and technology providers must take to ensure the grid can keep pace with the AI revolution.
When OpenAI launched ChatGPT on November 30, 2022, it ignited a global artificial intelligence (AI) race. The chatbot’s ability to generate human-like responses and tackle complex tasks marked a major leap in AI. It pushed competitors like Google, Meta, and Anthropic to accelerate their own large language models (LLMs).
But the AI boom is putting significant pressure on physical infrastructure. AI is driving a surge in electricity demand as companies build out new data centers to support increasingly intensive computing workloads.
This article focuses on the United States, examining the projected surge in electricity demand from AI-driven data centers, its impact on the electric grid, and why utilities and regulators must urgently modernize the grid to keep pace with this transformation.
How AI is Reshaping U.S. Electricity Demand
AI has quickly become one of the most consequential innovations of the 21st century. According to the International Energy Agency (IEA), large language models such as ChatGPT have reached a 40 percent adoption rate in the United States within two years. That pace outstrips the early growth of both the internet and personal computers (Graph 1).
Graph 1: Generative AI Adoption Is Outpacing Previous Workplace Technologies
Within two years of commercial release, Generative AI has reached 40 percent adoption in the United States
This rapid uptake has a significant energy cost. Training and running these models require enormous computing power, increasing pressure on data centers and pushing electricity use to record levels. A major factor behind this trend is the transition from traditional CPU-based servers to GPU-powered systems. GPUs, while less energy-efficient than specialized chips, dominate AI workloads today due to their strength in parallel processing.
One estimate finds that a single ChatGPT query uses nearly 10 times the electricity of a typical Google search. That is roughly equivalent to running a low-wattage LED light bulb for an hour (Graph 2).
Graph 2: AI-Powered System Requires more Energy than a Standard Google Search
As GPU infrastructure expands, data center energy use in the United States has climbed sharply. It rose from 176 terawatt-hours in 2018, or about 1.9 percent of national electricity consumption, to more than 4.4 percent in 2023. The Department of Energy projects that share could grow significantly, reaching as much as 12 percent in the coming years (Graph 3).
Graph 3: U.S. Data Center Electricity Use Is Surging with AI Growth
Electricity demand from data centers could reach up to 12 percent of total U.S. electricity consumption by 2028
To put the growth in context, the projected increase in U.S. data center electricity consumption between 2023 and 2028 exceeds the entire electricity use of the United Kingdom in 2023, which served approximately 28 million households.
As electricity demand accelerates, the stability of the U.S. grid is becoming a central concern for utilities, regulators, and the data center industry.
The Grid Under Strain
Most Americans run their appliances without giving much thought to the electricity behind them. But in parts of the country, that power supply is becoming less reliable.
As AI data centers expand, they are not only increasing electricity demand but also disrupting how power flows through local grids. In areas with dense clusters of energy-hungry servers, voltage fluctuations are becoming more frequent.
According to Bloomberg analysis, sensor data from more than 700,000 homes shows a clear pattern of declining power quality near fast-growing AI hubs. The most severe distortions were recorded within 20 miles of major data center clusters (Graph 4).
Graph 4: AI Data Center Hubs Are Linked to Declining Power Quality
Sensor data reveals severe power distortions within 20 miles of fast-growing AI infrastructure in the U.S.
Bloomberg’s analysis of sensor data found that in the average U.S. county, about 1.7 percent of sensors recorded at least one monthly reading above the 8 percent threshold for bad harmonics. In Loudoun County, where most of Northern Virginia’s data centers are located, that figure was more than four times higher.
In neighboring Prince William County, where significant new data center capacity was added recently, about 6 percent of the 1,100 residential sensors showed excessive harmonic distortion. Nearly all of those readings came from homes located within seven miles of major data center sites. Two dozen sensors recorded double-digit levels, with some reaching as high as 12.9 percent.
Beyond power quality concerns, data centers also pose a growing risk to grid stability. A standard safety feature in the data center industry is to disconnect from the grid and switch to on-site generators when voltage or frequency levels fall outside permitted thresholds. This protects sensitive equipment from damage. But when multiple facilities disconnect at once, it creates a sudden drop in demand, releasing a surge of excess electricity into the system. That imbalance can destabilize the grid and increase the risk of outages.
The threat is expected to grow as more data centers come online. The North American Electric Reliability Corporation, the federal agency responsible for grid reliability, warned in its 2024 Long-Term Reliability Assessment that the rapid growth of large data centers presents an emerging risk to grid stability. The concern centers on the way these facilities behave during faults, particularly the automatic disconnection of large loads, which can create serious operational challenges.
As data center development accelerates, so does the urgency to ensure the grid can adapt. Utilities are beginning to raise alarms. Many regions now face a dual challenge: meeting a historic rise in electricity demand while stabilizing the infrastructure that delivers it.
Modernizing the Grid
Meeting the challenge of soaring electricity demand from data centers and declining grid reliability requires a smarter and more resilient electric grid that can integrate more renewable energy. It must be able to handle rapid shifts in load, detect stress in real time, and respond before disruptions spread.
Yet much of the U.S. system, particularly at the distribution level, still runs on aging infrastructure. While investment in renewable energy has surged, spending on grid upgrades has lagged. In a 2023 report titled Electricity Grids and Secure Energy Transitions, the International Energy Agency found that more than 90 percent of power supply interruptions in the United States originate in the distribution network.
U.S. utilities have started to invest in the grid with a new sense of urgency. Among 2025 announcements:
Los sistemas de distribución, que suministran energía directamente a los usuarios finales, han sido el principal impulsor del gasto de capital de las empresas de servicios públicos en las últimas dos décadas. Entre 2003 y 2023, la inversión en infraestructura de distribución aumentó en 31 400 millones de dólares, lo que representa un aumento del 160 por ciento. Más de una quinta parte de ese crecimiento se produjo en un solo año. Entre 2022 y 2023, el gasto aumentó en 6 500 millones de dólares hasta un total de 50 900 millones de dólares, debido en gran medida a la sustitución y mejora de equipos antiguos (gráfico 7).
Gráfico 7: Adiciones anuales de capital en EE. UU. por sector
Miles de millones de dólares estadounidenses, 2003-2023
Las empresas de tecnología también están interviniendo. Empresas como Corinex con tecnologías de mejora de la red están trabajando con las empresas de servicios públicos para digitalizar la red de distribución de baja tensión mediante la tecnología de banda ancha a través de la línea eléctrica. Estos sistemas proporcionan información en tiempo real datos sobre voltaje, armónicos y flujos de energía , lo que ayuda a los operadores a detectar el estrés en forma temprana y a responder con mayor eficacia. Al mejorar la visibilidad y la flexibilidad a nivel local, nuestras herramientas ayudan a las empresas de servicios públicos a gestionar los impactos más amplios de las grandes cargas ascendentes, como los centros de datos.
Sin embargo, actualizar la infraestructura es solo una parte de la solución. Los planificadores y reguladores de la red están explorando nuevos estándares de interconexión para cargas grandes, centrándose en la ubicación, la coordinación en caso de fallos y una mejor comunicación entre los centros de datos y las empresas de servicios públicos.
Conclusión
La IA ya no es solo un avance tecnológico. Su auge está imponiendo demandas reales y crecientes a la red eléctrica de EE. UU. Los centros de datos, que antes se consideraban simplemente instalaciones para procesar información, ahora son consumidores de energía a gran escala que están transformando la forma en que se produce, distribuye y administra la electricidad.
Las empresas de servicios públicos, los reguladores y los proveedores de tecnología están empezando a responder, pero el ritmo de adopción de la IA se acelera más rápido que la capacidad de adaptación de la red. Sin una planificación coordinada y una inversión sostenida, la presión sobre los sistemas de energía solo aumentará.
El futuro de la IA depende de la confiabilidad de los sistemas que la respaldan. Construir ese futuro requerirá una red tan avanzada y adaptable como la tecnología que se espera que alimente.
Acerca del autor
Colin Tang es el director sénior de inversiones de Corinex, donde aprovecha su amplia experiencia en finanzas para impulsar la estrategia de inversión y el rendimiento de la cartera de la empresa. Con un historial comprobado de identificación y capitalización de oportunidades de inversión, Colin desempeña un papel crucial a la hora de respaldar los objetivos financieros y el crecimiento de Corinex.
Póngase en contacto con nosotros para solicitar acceso a los datos y análisis mencionados en nuestro artículo, o descubra cómo las soluciones de Corinex permiten la digitalización de la red y abordan las restricciones de la red.
¡Mantente a la vanguardia en el juego de la energía! Suscríbase a nuestro boletín CorinExt y reciba las últimas tendencias en visibilidad y flexibilidad de la red, actualizaciones de políticas, consejos de expertos y oportunidades de talleres, directamente en su bandeja de entrada.
Suscríbase a CoriNext para obtener las últimas noticias sobre la visibilidad de la red, la flexibilidad, los conocimientos de los expertos, los talleres y el futuro de la modernización de la red.
