Solving the grid’s AI power struggle with virtual power plants

Grid operators and utilities need to explore new strategies to meet the imminently growing power demand not only quickly but cost-effectively. One solution? Virtual power plants (VPPs), writes Hannah Bascom, chief market innovation officer at Uplight.

Everywhere you turn, it’s clear that AI is the topic du jour in technology and business circles. The opportunities that it offers are staggering and too enticing to turn down: PwC recently predicted that AI could contribute up to $15.7 trillion to the global economy before the end of the decade. Now, almost two and a half years after ChatGPT’s public launch, which catapulted AI into the public eye, the question isn’t whether AI will be adopted broadly; it’s how we will power it.

AI, and the data center infrastructure that supports it, is energy hungry. Even with the recently reported breakthroughs from the team behind Chinese startup DeepSeek—whose open-source AI model seems to be capable of outperforming current industry standards with a small fraction of the chips, hardware and energy that other hyperscalers require—we don’t yet have enough data center infrastructure to support AI’s computational power, storage and networking requirements. It’s estimated that AI and data center energy needs will drive a 67GW increase in energy demand over the next five years. Building the energy infrastructure to supply this demand would require a massive capital investment, and, especially given supply chain delays on some critical infrastructure, it will take years to come online.

But we don’t have years! Grid operators and utilities must explore new strategies to meet this imminently growing demand not only quickly but cost-effectively. One solution? Virtual power plants.

From the Power Playbook:
DeepSeek: What tumbling energy stocks say about AI’s power consumption

The value of virtual power plants

The largest untapped resource at utilities’ disposal is one that already exists: their customers and the increasingly electrified devices in their homes and businesses, including solar panels, energy storage systems, electric vehicles (EVs) and chargers, heat pumps and more. By enrolling and aggregating these distributed energy resources (DERs) in flexibility management programmes, utilities can not only unlock grid flexibility and resiliency but also leverage the megawatts these customers and devices bring with them as a Demand Stack that delivers firm, reliable, and clean capacity.

Many utilities already have demand stack building blocks in place: a combination of energy efficiency programmes that steadily reduce daily consumption, time-varying rates that actively shape daily load patterns, and demand response programmes that provide targeted relief during critical periods. Now, consider adding virtual power plants to that mix.

VPPs are the natural evolution of utility demand-side management initiatives, combining multiple technologies, customer segments, and utility programmes into a unified portfolio that operates like a conventional supply resource.

Because VPPs do not require the capital and infrastructure investments needed for conventional supply resources, they can be deployed rapidly and at substantially lower costs than supply resources. VPPs have enormous potential: the Department of Energy estimates that tripling the scale of VPPs by 2030 could help meet rapid electricity demand growth and help save on the order of $10 billion in annual grid costs, redirecting those savings back to electricity consumers. And the good news is that we already have a strong foundation for VPP lift-off—any utility with a demand response programme can evolve their programme into a VPP.

To deliver on their promise, a VPP needs to harness, orchestrate, and optimise energy assets across multiple customer classes, device types, and use cases. With multiple resources working together, VPP portfolios maximise programme enrollments, expanding grid services and ensuring maximum flexibility. They should also include real-time monitoring and control to enable utilities to respond quickly to changes in demand or supply. Grid services aren’t just limited to peak shaving, either. Best-in-class solutions use machine learning algorithms to predict energy demand, optimise energy storage and dispatch and reduce energy waste. With an integrated demand stack including VPPs, grid operators can manage local network constraints and wholesale price hedging, while achieving reliable capacity year-round.

For an example of VPP in action, take Puget Sound Energy, Washington state’s largest utility. Washington, along with the broader Pacific Northwest region, faces predicted growth in energy demand of more than 30% in the next decade, driven by factors such as data centers and electrification. The utility is leveraging a VPP to deliver 100MW of flexible capacity—enough to power approximately 100,000 homes. PSE’s VPP is a comprehensive programme that centralises the enrollment, dispatch and assessment of individual and combined demand response programmes (DR) across PSE’s portfolio. In 2023 alone, PSE scaled the programme from zero to 30MWs in just a few months by tapping into its demand stack, including energy efficiency, residential and commercial DR, battery storage and electric vehicle programmes.

To address the growing power demands driven by AI, utilities must begin deploying solutions now, and Virtual Power Plants (VPPs) offer the most practical near-term approach to quickly deliver the flexibility needed for this challenge.

Have you read:
TransnetBW gives sonnen greenlight for virtual power plant expansion
PG&E launches ‘first of its kind’ virtual power plant

Exploring data centres as grid assets

Building VPPs around data centres alone isn’t enough; they must also be integrated into flexibility management strategies. Standard data centres, data centres specialising in AI, and chip foundries are projected to increase demand from 130 terawatt-hours (TWh) in 2023 to 307 TWh in 2030. This massive growth in demand can be a net positive for the grid if it is paired with greater flexibility. Including data centres in VPPs can provide substantial, reliable flexibility during times of peak demand.

New research from Uplight and See Change Institute found that making this a reality relies heavily on two things: education and communication. The qualitative study found that while data centre managers largely understand the idea of demand flexibility, they know less about VPPs or specific flexibility programmes. With a bit more education around the mechanics, they express some interest in participating, albeit with some stipulations.

Their hesitations? Participating without significant benefits in general one-size-fits-all programmes. Not all data centres are the same, so load flexibility programmes should also be tailored to their needs while ensuring reliability and cost-effectiveness.

From the research, it became clear that if utilities want to flip the paradigm of data centres from grid challenge to asset, a tailored, customised approach to engaging data centre energy managers is required. And, that connection should start as a new facility (or expansion) is in the planning phase, giving plenty of time to help establish a positive relationship and best collaborate (and educate) on VPPs and other energy programme solutions.

Together, by leveraging the power of VPPs and reconsidering data centres as a potential asset, utilities can play a pivotal role in navigating this unprecedented moment of transformation in the energy landscape. These strategic shifts will not only help meet the rising demand for clean energy in light of the AI boom, but also help alleviate the grid’s pressing infrastructure needs. In doing so, utilities can continue to provide the reliability, flexibility, and safety that customers depend on. Embracing these innovative solutions will enable a forward-thinking energy system that benefits everyone—from consumers and hyperscalers to grid operators and the environment—and foster a more sustainable and equitable energy future.

About the author

Hannah Bascom leads the growth team at Uplight. Prior to joining Uplight, Bascom spent 20 years building partnerships and new markets in the climate and social impact space.