Tweaks to US Data Centers Could Unlock 76 GW of New Power Capacity

The rapid expansion of AI has sent tech companies, data center developers, and power utilities into a frenzy over the potential for soaring electricity demand in the U.S. Yet, a recent study offers a glimmer of hope, suggesting that a small tweak in how data centers manage their electricity consumption could significantly ease the strain on the grid.
The study proposes that if data centers and other major electricity consumers were to limit their power draw to 90% of their maximum for just a couple of hours at a time—totaling about one day a year—they could free up an impressive 76 gigawatts of capacity across the U.S. To put this into perspective, that's more than the total electricity used by data centers worldwide, according to Goldman Sachs, and it represents about 10% of the peak demand in the U.S. The more data centers curtail their usage, the more capacity they could unlock.
This concept of demand response isn't new. For years, utilities have incentivized large electricity users such as shopping malls, universities, and factories to reduce their power usage during peak times, like sweltering summer afternoons. In exchange for turning down the AC or powering off energy-intensive machines for a few hours, these users receive a bill credit. Data centers, however, have traditionally prioritized maintaining uptime and performance for their clients, opting out of such programs. But the study highlights that data centers could be prime candidates for demand response due to their inherent flexibility.
Strategies for Data Centers to Reduce Power Usage
The study outlines several ways data centers can adjust their power consumption:
- Temporal Flexibility: By shifting computing tasks to periods of lower demand. For instance, AI model training, which doesn't need to happen in real-time, could be rescheduled to avoid peak hours.
- Spatial Flexibility: Companies can move computational tasks to data centers in regions with lower demand. Additionally, operators can consolidate server loads and temporarily shut down some servers.
- Alternative Power Sources: For tasks that can't be delayed or relocated, data centers can rely on backup power sources like batteries, which can provide several hours of power quickly and efficiently.
Some companies are already experimenting with these strategies. Google, for example, uses its carbon-aware computing platform to facilitate demand response, a tool initially developed to reduce carbon emissions. Enel X collaborates with data centers to use batteries in uninterruptible power supplies (UPS) to support grid stability. Meanwhile, PG&E offers expedited grid connections to data centers that agree to participate in demand response programs.
While these adjustments won't entirely eliminate the need for new power sources, they could transform a potentially dire situation—where half of new AI servers might be underpowered—into one that's much more manageable.
Related article
Bain forecasts US$100 billion SaaS market in agentic AI automation
Bain & Company has estimated a $100 billion market in the U.S. for SaaS companies leveraging agentic AI. The firm said this market stems from automating coordination tasks within enterprise systems.This estimate comes from the second installment in B
AI Search Mandatory Policy Fuels Exodus, DuckDuckGo Sees User Surge
Following Google's 2026 I/O conference announcement of a full AI overhaul of its search engine, many users started looking for more controllable alternatives because there was no simple "one-click disable" for AI features. The privacy-focused search
Xiaohongshu Restructures: Conan Named President, Creates AI Primary Department Dots and Overseas Division Rednote
On April 30, Xiaohongshu sent an internal memo to all employees announcing the launch of a new organizational restructuring. The core of this change involves fully integrating three business lines—community, e-commerce, and commercialization—along wi
Related Special Topic Recommendations
Comments (40)
0/500
Ça donne enfin une perspective positive au débat énergétique autour de l'IA ! Plutôt que de toujours dire 'trop de consommation', on propose une solution technique simple. C'est rafraîchissant. 76 GW, ça représente quoi concrètement ? L'équivalent de combien de centrales ? L'article donne-t-il plus de détails ?
Die Diskussion um Rechenzentren und Stromverbrauch wird immer relevanter, gerade mit dem KI-Boom. Der vorgeschlagene kleine 'Tweak' klingt fast zu schön, um wahr zu sein - 76 GW sind eine irre Menge! Würde mich interessieren, wie genau diese Anpassungen in der Praxis aussehen. Gibt es schon Pilotprojekte?
C'est fou comme une petite modif peut libérer autant d'énergie ! 76 GW, ça donne envie de rêver à des data centers plus verts, mais est-ce que ça suivra vraiment ? 🤔
This article's take on tweaking data centers to unlock 76 GW is wild! 😮 It's like finding extra juice in a squeezed lemon. Makes me wonder how much AI's gonna reshape our power grids. Anyone else think this could spark a green energy race?
The idea of unlocking 76 GW with just tweaks is wild! 😮 Makes me wonder how much more AI can push the grid before we hit a wall. Efficiency is key, but are we ready for the energy demands of super-smart AI?

The rapid expansion of AI has sent tech companies, data center developers, and power utilities into a frenzy over the potential for soaring electricity demand in the U.S. Yet, a recent study offers a glimmer of hope, suggesting that a small tweak in how data centers manage their electricity consumption could significantly ease the strain on the grid.
The study proposes that if data centers and other major electricity consumers were to limit their power draw to 90% of their maximum for just a couple of hours at a time—totaling about one day a year—they could free up an impressive 76 gigawatts of capacity across the U.S. To put this into perspective, that's more than the total electricity used by data centers worldwide, according to Goldman Sachs, and it represents about 10% of the peak demand in the U.S. The more data centers curtail their usage, the more capacity they could unlock.
This concept of demand response isn't new. For years, utilities have incentivized large electricity users such as shopping malls, universities, and factories to reduce their power usage during peak times, like sweltering summer afternoons. In exchange for turning down the AC or powering off energy-intensive machines for a few hours, these users receive a bill credit. Data centers, however, have traditionally prioritized maintaining uptime and performance for their clients, opting out of such programs. But the study highlights that data centers could be prime candidates for demand response due to their inherent flexibility.
Strategies for Data Centers to Reduce Power Usage
The study outlines several ways data centers can adjust their power consumption:
- Temporal Flexibility: By shifting computing tasks to periods of lower demand. For instance, AI model training, which doesn't need to happen in real-time, could be rescheduled to avoid peak hours.
- Spatial Flexibility: Companies can move computational tasks to data centers in regions with lower demand. Additionally, operators can consolidate server loads and temporarily shut down some servers.
- Alternative Power Sources: For tasks that can't be delayed or relocated, data centers can rely on backup power sources like batteries, which can provide several hours of power quickly and efficiently.
Some companies are already experimenting with these strategies. Google, for example, uses its carbon-aware computing platform to facilitate demand response, a tool initially developed to reduce carbon emissions. Enel X collaborates with data centers to use batteries in uninterruptible power supplies (UPS) to support grid stability. Meanwhile, PG&E offers expedited grid connections to data centers that agree to participate in demand response programs.
While these adjustments won't entirely eliminate the need for new power sources, they could transform a potentially dire situation—where half of new AI servers might be underpowered—into one that's much more manageable.
AI Search Mandatory Policy Fuels Exodus, DuckDuckGo Sees User Surge
Following Google's 2026 I/O conference announcement of a full AI overhaul of its search engine, many users started looking for more controllable alternatives because there was no simple "one-click disable" for AI features. The privacy-focused search
Xiaohongshu Restructures: Conan Named President, Creates AI Primary Department Dots and Overseas Division Rednote
On April 30, Xiaohongshu sent an internal memo to all employees announcing the launch of a new organizational restructuring. The core of this change involves fully integrating three business lines—community, e-commerce, and commercialization—along wi
Ça donne enfin une perspective positive au débat énergétique autour de l'IA ! Plutôt que de toujours dire 'trop de consommation', on propose une solution technique simple. C'est rafraîchissant. 76 GW, ça représente quoi concrètement ? L'équivalent de combien de centrales ? L'article donne-t-il plus de détails ?
Die Diskussion um Rechenzentren und Stromverbrauch wird immer relevanter, gerade mit dem KI-Boom. Der vorgeschlagene kleine 'Tweak' klingt fast zu schön, um wahr zu sein - 76 GW sind eine irre Menge! Würde mich interessieren, wie genau diese Anpassungen in der Praxis aussehen. Gibt es schon Pilotprojekte?
C'est fou comme une petite modif peut libérer autant d'énergie ! 76 GW, ça donne envie de rêver à des data centers plus verts, mais est-ce que ça suivra vraiment ? 🤔
This article's take on tweaking data centers to unlock 76 GW is wild! 😮 It's like finding extra juice in a squeezed lemon. Makes me wonder how much AI's gonna reshape our power grids. Anyone else think this could spark a green energy race?
The idea of unlocking 76 GW with just tweaks is wild! 😮 Makes me wonder how much more AI can push the grid before we hit a wall. Efficiency is key, but are we ready for the energy demands of super-smart AI?





Home






