How to power data center growth without overloading the grid

Once viewed as too rigid to participate in demand response, data centers are becoming increasingly flexible. Advances in behind-the-meter (BTM) generation, energy storage technologies and workload shifting now allow operators to respond dynamically to grid conditions.

By reframing data centers as flexible grid resources rather than problematic loads, utilities can unlock new capacity, stabilize the grid and support continued expansion of an important pillar of the US innovation economy.

Data centers are breaking traditional utility financial and regulatory models

The rapid growth of data centers in the US has significantly impacted energy infrastructure, leading to soaring electricity costs, especially in areas with high data center activity.³ Utility investments to accommodate large new loads are typically spread across all ratepayers through rate cases, a framework slow to adapt to the scale and speed that data center developers demand. As a result, critics argue that residential customers are subsidizing data center development, fueling rising public opposition and regulatory scrutiny. More than $64 billion in US data center projects have been blocked or delayed amid increasing local and bipartisan resistance.⁴

In response, some utilities have introduced separate large-load tariff structures designed to isolate infrastructure costs associated with data centers and protect other ratepayers. The Electric Power Research Institute’s (EPRI) DCFlex tool tracks adoption nationwide.

Recent analyses, including a Brattle Group review of large-load tariff frameworks in Kansas and Missouri, have been cited as models for balancing data center access to grid resources while limiting the impact to ratepayers.⁵ While these tariff structures offer some relief, they do not solve the core challenge for data centers seeking timely access to power.

A strained grid limits speed to power and drives on-site generation

As grid congestion and interconnection delays persist, data centers are increasingly exploring BTM generation to secure power faster. On-site solutions, including microgrids, generators, renewables and battery storage, reduce reliance on constrained transmission networks, but complicate utility planning by increasing uncertainty around long-term demand and cost recovery.

BTM generation raises concerns about stranded assets, revenue erosion and grid coordination. While data centers will continue to rely on utilities for interconnection and delivery infrastructure, unmanaged growth in on-site generation complicates traditional planning models.

With gas turbine backlogs limiting near-term generation expansion, battery storage has emerged as the most immediate solution. Hybrid grid-plus-on-site approaches are likely to accelerate, increasing the need for flexible strategies that preserve reliability while enabling speed to power.

Demand response is a solution to grid constraints

Even modest, short-duration participation in demand response during these peak periods can ease interconnection bottlenecks and accelerate access to power.

Data center demand response strategies

Traditionally, demand response programs involving large commercial power customers were designed for emergency use, with demand curtailment serving as a measure of last resort to maintain grid stability. While data centers were long viewed as too rigid to participate due to uptime requirements, recent advances are changing that assumption. Modern data centers are inherently flexible, particularly as AI workloads become the dominant driver of incremental power demand.⁹

Many AI-driven tasks, such as model training and fine-tuning, do not require continuous uptime and can be deferred, throttled or shifted during periods of peak grid stress. New demand response technologies now allow operators to dynamically reduce energy use by powering down idle servers or limiting graphics processing unit GPU power draw, enabling meaningful participation without disrupting critical services. For instance, AI platforms like Emerald AI’s Conductor Platform can dynamically adjust data center power consumption to align with grid conditions, achieving up to a 25% reduction during peak stress events.¹⁰

Beyond technology, new commercial and market structures are expanding demand response participation. Virtual power plant (VPP) providers and bring-your-own-capacity models allow data centers to access flexible capacity without relying solely on utility-funded infrastructure.¹¹ Cross-industry pilot programs led by utilities, grid operators and research organizations are also testing flexible data center designs and accelerated interconnection pathways tied to load curtailment commitments.

The DCFlex project, led by EPRI with support from major tech and utility firms, seeks to create flexible data center designs for improved demand response. Its goal is to establish flexibility hubs and reference architectures that enable grid-responsive operations, with demonstration projects running through 2027 to explore renewable backup power and grid integration.¹²

At the same time, regional transmission organizations (RTOs) and independent electric system operators (ISOs) are revising rules, tariffs and market products to better accommodate large, flexible loads. Together, these technology, market and regulatory innovations are repositioning data centers from passive energy consumers to active participants in grid reliability, helping utilities manage near-term constraints while enabling faster access to power.

Data center demand response: challenges and barriers to adoption

Despite growing interest in demand response as a tool to relieve grid pressure and improve speed to power, significant barriers to adoption remain. While the opportunity is clear, participation has historically been limited by operational, financial and cultural constraints on both the utility and data center sides.

The barriers below highlight these challenges:

A path forward: recommendations for utilities and data centers

Addressing these barriers requires a shift toward flexibility, collaboration and technology-driven solutions that balance operational needs with grid reliability. Data centers can adapt far quicker than utilities can expand infrastructure, making flexibility a critical near-term lever for managing growing grid constraints.

The following recommendations summarize steps that both utilities and data centers can take to ease adoption of demand response and grid flexibility programs:

Tony Calabro, Brad Cobb, Brady Klein, Kohil Vyas and Jamie Adams also contributed to this article.