Momentum Growing To Make Big Tech Data Centers More Flexible

New efforts are underway to reshape the way data centers and utilities operate — and unlock huge amounts of unused power in the process. 

Unprecedented electricity demand from data centers is straining the limits of regional power grids, slowing the pace of new development and driving concerns about rising power bills for consumers.

The power pinch has pushed data center development into new regions as providers chase available grid capacity, and it has led to the rise of campuses that use on-site or behind-the-meter power generation instead of relying exclusively on the grid. 

But there is also a third path.

There is a growing push for data centers to utilize on-site generation or energy storage to become “flexible loads,” capable of reducing their power consumption or disconnecting completely from regional grids and operating under their own power at the behest of a utility.

If widely adopted, the shift could deliver significant benefits to both data center developers and the power sector, industry leaders said this month at Bisnow’s DICE: Power Capacity, Energy & Sustainability event, held at the Sheraton Fort Worth Downtown Hotel.

For developers, it could unlock access to power on far shorter timelines.

For utilities and ratepayers, greater flexibility could ease reliability concerns and help contain the potential for rising costs associated with the ongoing artificial intelligence-driven data center boom. 

Data centers becoming more active grid participants would represent a fundamental change in the way that both data centers and utilities operate. Because of this, the much-discussed idea of flexible data centers has seen little real-world adoption.

But that may be about to change. A pair of industry-led initiatives are steering substantial resources toward bringing flexible data centers to fruition.

Those behind the initiatives are aiming to get buy-in from a wide array of stakeholders, including grid operators and state governments, who would need to be active participants in the strategic shift. 

“What does this do for the utility and for ratepayers? This turns an AI factory from a grid burden to a grid asset,” David Chernis, director of flexible compute platforms at CPower Energy, said at DICE: Power. “And with just 1% flexibility of these massive AI factory loads, we could free up another 80 or 90 gigawatts of capacity to spread around to all these projects we’re trying to build.”

Chernis and other advocates for flexible data centers say the solution to the data center power pinch isn’t adding capacity to overmatched power grids — it’s being smarter about using the capacity that’s already available.

They argue that there is a fundamental inefficiency underpinning the energy crisis: The grid is designed to ensure demand doesn't exceed supply during short periods of peak use, but, the vast majority of the time, only a small percentage of available power is actually being used.

Typically, electricity demand only comes close to exceeding supply a few days a year, often during periods of extreme heat or extreme cold. 

By establishing a dynamic system in which a grid operator could curtail data centers’ power consumption as needed during the handful of days each year when demand is the highest, it would effectively increase the capacity of the power grid. Power providers would be able to sell most of the grid capacity that’s available on a typical day, not just the amount available on the most extreme demand days each year. 

This means faster speed to power for data center developers, who would no longer have to wait years for grid connections while utilities build the generation or transmission infrastructure needed to accommodate the most extreme demand scenarios. 

“We help engineer and enable AI factories and new data centers to be flexible so the utility can accept everybody's interconnection request in a timely and fast manner,” Chernis said. “You’re coming to the interconnection with a way to get off the grid for a handful of hours a year where the grid cannot provide enough capacity.”

Flexible data centers would also benefit utilities and regional system operators. 

For the first time, they would have a demand-side lever to help balance supply and demand across regional transmission systems. Instead of just being able to control electricity generation, they could also control consumption, potentially improving grid reliability. 

Flexibility could also mean lower costs for utilities and their customers. Power providers would be able to generate more revenue from their existing infrastructure. At the same time, demand-side flexibility lowers the need for costly infrastructure upgrades needed to accommodate periods of peak demand.

A report published this year from Duke University found that flexible data center power consumption would allow the U.S. to avoid up to $150B in power plant, fuel and transmission costs over the next decade.

“The utilities, in the past, built for that high point of demand,” Jeff Moerdler, chair of the data center and digital infrastructure practice at Haynes Boone, said at DICE: Power. “Now, you would need less capacity on the grid to serve the network. ”

There has been limited adoption of data center flexibility. Some regional power systems already have a form of “demand response,” under which participating data enters curtail power consumption at the grid operator’s behest to avoid blackouts or other catastrophic events. 

Yet there has been little progress when it comes to making demand-responsive data centers not just a safety valve to avoid grid emergencies but an integral part of how the facilities acquire power and how regional grids are managed. 

Flexibility has become viable for a growing share of the data center sector over the past 24 months.

More facilities than ever before are being built with on-site generation beyond backup generators. At the same time, new power generation and storage technologies — particularly battery energy storage systems — have become more economically feasible.

Now, the main barrier to adoption is human rather than technical, industry leaders say. Data center operators and tenants are inherently risk-averse, and participation in a program that would hand certain power management decisions to a third party would run afoul of many tenant contracts. 

Additionally, many different stakeholders within the country's various grid systems would need to be convinced to change how they operate in order for demand-responsive data centers to gain widespread implementation.

In a balkanized U.S. power landscape, making any kind of structural change to how power is bought, sold and delivered is extremely difficult. There needs to be buy-in not just from data center operators and utilities but from regional grid operators, federal regulators, state governments and a range of environmental and consumer watchdogs.  

But with these solutions more technologically plausible — and more necessary — companies connected to the data center sector are ramping up efforts to bridge the bureaucratic and political barriers preventing large-load flexibility. 

A new organization called Utilize launched March 10 with backing from a coalition of technology and energy giants including Google and Tesla. It plans to lobby state governments and utilities to adopt policies and technologies that increase utilization of existing grid capacity by allowing data centers and other large power users to become flexible loads. 

“For decades, we’ve built the grid to meet peak demand, even though large portions of it sit unused for most hours of the year,” Utilize Executive Director Ian Magruder said in a statement. “It’s like building an airplane that only flies with full passengers a few times a year. That excess capacity is hiding in plain sight, and new technologies give us the opportunity to unlock it.”

A week later, power research nonprofit EPRI unveiled what it is calling Flex Mosaic: a standardized framework for implementing flexible-load data centers across U.S. power systems. This stems from an initiative called DCFlex, launched in 2024 at the behest of the Biden administration.

EPRI spearheaded DCFlex in collaboration with data center providers like Compass Datacenters and QTS, tech companies like Google and Meta, power providers like Southern Company and grid operators like the Electric Reliability Council of Texas. It aims to develop models and proofs of concept for the technologies, policies and operating practices needed across these sectors to make grid-responsive data centers viable. 

In creating the Flex Mosaic framework, the coalition’s leadership hopes to establish a common set of definitions and shared knowledge base to help bridge the divide between the stakeholders that need to collaborate if flexible-load data centers are going to become commonplace. 

“This framework allows everyone — utilities, regulators, and large-load developers — to have common language about flexibility and to trust what that language means,” EPRI CEO Arshad Mansoor said in a statement. “That shared understanding is essential to moving faster while maintaining reliability.”