The electricity system faces a problem that new generation alone cannot solve: demand does not arrive when the grid finds it convenient to supply power. A data center can need electricity throughout the day, a manufacturing site can maintain production through demanding shifts, and millions of cooling systems can reach their highest consumption when temperatures rise. Meanwhile, solar generation declines toward evening, transmission networks encounter congestion, and grid operators must preserve enough capacity to withstand sudden changes in supply or demand. Virtual power plants offer a way to coordinate distributed resources and turn selected electricity consumption into something grid operators can manage. The US Department of Energy’s virtual power plant strategy therefore addresses a fundamental question for the AI economy: how much additional demand can the grid accommodate if more of its existing resources become flexible?

The US Department of Energy’s recent update sets an ambition to deploy between 80 GW and 160 GW of virtual power plant capacity by 2030, equivalent to approximately 10% to 20% of peak electricity demand. The department also identifies potential annual grid cost savings of $10 billion, subject to the deployment of flexible resources and their ability to reduce operating costs or avoid or defer infrastructure investments. These figures represent a modeled opportunity rather than a guaranteed outcome, and the report does not assign the entire capacity target to data centers. The central challenge is converting this potential into dependable capacity that grid operators can call upon when the system needs it.

The Liftoff Math No One Has Mapped To Real Demand

The headline capacity range becomes more meaningful when compared with the electricity system’s projected peak demand. The DOE’s VPP Liftoff Update sets a deployment ambition of 80 GW to 160 GW by 2030 and describes that range as approximately 10% to 20% of peak load. Any comparison with a separate forecast of US peak demand must identify the forecast’s source, scenario, and definition of peak demand. The target describes the potential contribution of coordinated distributed resources rather than a guarantee that every enrolled megawatt will be available during every stressed operating period. A battery, thermostat, or industrial load may have different operating limits, response durations, and availability conditions. Grid planners must therefore assess the dependable contribution of a portfolio rather than treating its aggregate enrolled capacity as fully dispatchable under all conditions.

What the 80–160 GW Target Actually Represents

The arithmetic also reveals why installed capacity and dependable capacity cannot serve as interchangeable planning concepts. A battery can discharge only within its available energy reserve, while a thermostat can reduce cooling demand only within the limits of indoor temperatures, equipment operation, and customer participation. An industrial site may offer substantial curtailment during a defined operating window but little flexibility when production schedules or process requirements tighten. A data center may possess controllable cooling loads and stored energy yet remain unable to shift a particular computing workload because latency, data locality, or service-level requirements constrain scheduling. Grid planners must therefore examine the size of a resource portfolio alongside its duration, availability, location, response speed, and ability to repeat a dispatch event. Without that qualification, a large aggregate capacity figure could overstate how much electricity a system can reliably avoid consuming during its most difficult operating periods.

The difference between an ambition and a deployable resource portfolio also emerges from the relationship between annual electricity consumption and peak demand. Annual energy describes the electricity used over time, whereas peak demand captures the maximum rate at which the system must deliver power during a particular interval. A VPP can reduce peak demand without materially changing a customer’s total daily energy requirement if it moves consumption from a constrained period into a less demanding one. That capability matters when a grid experiences a temporary capacity shortage even though it has sufficient energy resources across the broader day. The DOE’s target consequently requires more than adding controllable devices to an enrollment database, because the portfolio must match the system’s actual needs in the right locations and operating windows.

Why AI-Era Demand Changes the Planning Equation

Data center demand complicates conventional capacity planning because computing infrastructure combines sustained electricity requirements with workloads that do not all respond to the same operational constraints. Some applications require immediate responses, continuous data movement, or tightly coordinated processing, while other jobs can tolerate scheduling changes or pauses within defined limits. Cooling systems, pumping equipment, and thermal storage also introduce controllable loads whose operating windows may differ from those of the computing equipment they support. A site can therefore contain several distinct flexibility opportunities rather than one uniform block of demand that an operator can switch off. The engineering task involves identifying which loads can move, how long they can move, what recovery demand follows, and which operating conditions prohibit intervention. That assessment turns a broad claim about flexible computing into an operational plan that can specify the amount of capacity available under particular grid conditions.

The DOE’s target does not establish that data centers must supply the full VPP capacity requirement, and the report does not assign the entire target to AI infrastructure. Residential equipment, commercial buildings, industrial processes, electric vehicles, batteries, and distributed generation can all contribute potential flexibility. Growing electricity demand from computing increases the value of assessing whether large data centers can help manage peak loads, but it does not prove that AI growth mathematically requires the entire VPP target. Large electricity users may offer concentrated loads, established monitoring systems, and engineering capabilities that support participation in demand-response programs. Their actual flexibility still depends on computing requirements, cooling design, backup arrangements, and commercial obligations. Data center participation is therefore one potential route to expanding dependable VPP capacity, rather than a documented prerequisite for achieving the DOE’s target.

Where That $10 Billion A Year Actually Comes From

The DOE’s potential annual savings estimate concerns the cost of serving electricity demand, rather than simply the amount of fuel that flexible resources might avoid burning. The department identifies potential benefits from reducing reliance on peaking resources and avoiding or deferring certain grid investments. Each pathway addresses a different part of the electricity system, so the savings depend on which investment or operating expense a VPP can actually displace. A utility that avoids running a peaker during a constrained evening may reduce operating costs without eliminating the need for the plant itself. A utility that can defer a proposed generating unit or network upgrade may capture a different form of value by postponing a major capital commitment. The $10 billion figure should therefore be presented as a modeled system-level opportunity, not a guaranteed annual saving for participating customers.

The Savings Come From Infrastructure That the Grid May No Longer Need to Build

The difference matters because conventional electricity procurement does not always expose the full system value of a resource that reduces demand during a difficult operating period. A power purchase agreement generally establishes commercial terms for electricity supply, while the costs of transmission reinforcement, local distribution upgrades, capacity adequacy, and balancing can sit in separate arrangements. The precise allocation varies by market, contract, and regulatory structure, so a lower energy price does not necessarily indicate that a project will reduce the total cost of serving a particular site. A VPP may create value by reducing the need for a costly network intervention even when its participants consume almost the same total amount of electricity over a complete operating cycle. That value depends on whether the flexibility arrives behind the constraint, during the relevant period, and at a level that changes the investment decision.

The department also describes VPP capacity as a potentially lower-cost alternative to conventional peaking resources under suitable conditions. Its program overview states that a VPP combining residential thermostats, water heaters, electric vehicle chargers, and behind-the-meter batteries could provide peaking capacity at roughly half the net cost to a utility of alternatives such as a utility-scale battery and a natural gas peaker plant. That comparison belongs to the department’s stated analytical case and should not be treated as a universal cost ratio for every technology, location, or dispatch duration. A VPP still requires controls, communications, aggregation, customer incentives, measurement, and settlement, while the underlying resources may have costs that participants must recover. A central plant, meanwhile, offers a different combination of dispatchability, duration, fuel requirements, and operational responsibilities that a demand-side portfolio cannot reproduce in every situation.

Why Flexibility Must Compete With New Central Capacity on Delivered Value

A credible comparison between VPPs and new central capacity begins with the grid service that each resource must provide. A generator can supply electricity when called, subject to its operating limits and fuel availability, whereas a demand-side resource creates capacity by reducing or shifting consumption relative to an established baseline. A battery can bridge that difference by delivering stored electricity, but its contribution depends on charging opportunities, state of charge, and the duration of the dispatch requirement. A coordinated portfolio can combine these capabilities, allowing one resource to reduce consumption while another supplies power or supports a local network constraint. The financial case must still account for the portfolio’s availability, replacement costs, customer payments, control infrastructure, and the possibility that several resources become unavailable during the same weather or market event.

The timing of investment creates another source of potential value that simple electricity-price comparisons can miss. A transmission or distribution upgrade may require planning, approvals, equipment procurement, construction, and coordination with other grid projects before the additional capacity becomes usable. If a VPP can reliably relieve the specific constraint that drives the upgrade, a utility may gain time to complete the project without exposing customers to the same level of near-term capacity pressure. That deferral does not necessarily cancel the upgrade, because sustained demand growth may eventually exceed the available flexibility or create a different constraint elsewhere on the network. The value instead comes from aligning capital spending with demonstrated need and reducing the risk of building infrastructure that the grid could avoid or postpone.

Thermostats Got Us To 37.5 GW, They Won’t Get Us To 160

The North American VPP market provides a useful starting point for understanding the distance between current deployment and the DOE’s target. Wood Mackenzie reported that operational VPP capacity reached 37.5 GW in its 2025 North America market assessment, following growth from 33 GW in the previous reporting period. The firm also reported that deployment activity expanded faster than total capacity, suggesting that the market added programs and commercial relationships without increasing aggregate capacity at the same pace. Its assessment identified utility program limits, capacity accreditation reforms, and market access barriers among the constraints affecting growth. The figures describe a market that already coordinates substantial distributed flexibility, but they do not establish that every enrolled resource can deliver equivalent capacity during a stressed evening or a prolonged supply shortage.

What the Existing VPP Market Actually Tells Us

The composition of existing VPP portfolios also complicates the assumption that residential thermostats alone can deliver the next phase of growth. Wood Mackenzie’s reporting identifies smart thermostats as an established technology and documents the presence of batteries and electric vehicles in VPP deployments. Residential equipment remains valuable because aggregators can coordinate many small loads, and thermal inertia allows some cooling demand to move across a limited period without an immediate loss of comfort. A thermostat’s available response depends on outdoor temperatures, building characteristics, customer settings, and the length of time that cooling can remain below its normal operating level. Batteries and water heaters offer different operating characteristics and may provide flexibility over different durations. Expanding the resource mix can improve the range of services available, but the capability of each portfolio must be assessed from its actual performance rather than assumed from its enrollment figures.

The market’s next stage depends on expanding the range of controllable assets rather than abandoning the household resources that established the model. Batteries can supply electricity directly during a peak, electric vehicles can alter charging schedules, and commercial buildings can coordinate heating, ventilation, and cooling systems around grid requirements. Industrial processes may offer larger individual blocks of controllable demand, although production commitments, safety requirements, and equipment limitations determine how much flexibility operators can release. Data centers add another category because their electrical demand combines computing workloads, cooling systems, power conversion, and stored energy within a tightly managed operating environment. Aggregators that combine these resources can build a portfolio with several complementary responses rather than relying on one device type to meet every dispatch requirement. Reaching the DOE’s target will therefore require a market structure that values resource diversity and verifies actual delivery, not merely a larger count of connected devices.

Why Enrollment Models Must Expand Beyond Household Controls

Residential VPP programs often begin with a straightforward relationship between a device owner, an aggregator, and a utility or electricity market. The aggregator obtains permission to control an eligible resource, communicates operating instructions, measures the response, and arranges compensation under the applicable program. That structure works well when the device has a clear operating envelope and the customer can accept modest changes to normal consumption. Large commercial and industrial sites require a more detailed arrangement because their loads may interact with production schedules, critical services, cooling requirements, backup systems, and internal energy-management controls. A site operator must understand which equipment the aggregator can influence, which conditions prohibit dispatch, and who bears responsibility if a grid request conflicts with the site’s operating requirements. The enrollment process consequently becomes a technical integration and commercial negotiation rather than a simple registration exercise.

The capacity contribution of a residential portfolio depends on whether its participants can respond together when the grid calls. Individual homes have different thermal conditions, occupancy patterns, equipment characteristics, and willingness to accept temporary changes in comfort. An aggregator can forecast expected availability and coordinate dispatch, but it must account for customer overrides, communication failures, and the possibility that the actual response falls below its forecast. Batteries and vehicle charging introduce different constraints because stored energy and future mobility needs affect their availability. Commercial sites may provide more concentrated resources, although production commitments and operational limits can restrict their response. Portfolio design must therefore combine resources with different operating characteristics and account for availability risks instead of assuming that residential flexibility is inherently incapable of longer-duration service.

The Load That Stopped Being Just A Load

A data center’s electricity demand usually enters grid planning as a requirement that the network must serve, rather than as a resource that can help balance the system. That approach reflects the operational importance of computing workloads, which may support applications that cannot tolerate interruptions or uncontrolled changes in processing schedules. Yet a site’s total electrical demand includes systems with different control requirements, thermal characteristics, and operating priorities. Computing equipment, cooling loops, pumps, heat rejection systems, batteries, and charging infrastructure do not necessarily need to follow identical operating schedules. An operator that understands these relationships can identify specific loads that may shift without affecting critical computing services or breaching operating limits. A data center can potentially contribute to demand reduction or load shifting, while batteries may supply stored electricity if the site has suitable equipment and available reserves.

Precooling offers one potential route to short-duration demand flexibility because a cooling system can sometimes remove heat before the grid enters its most constrained period. The site may then use its available thermal headroom for a limited time, allowing cooling equipment to reduce electrical demand while maintaining acceptable operating conditions. Thermal storage can extend this opportunity where the site’s design supports it, although its contribution depends on storage capacity, charging schedules, thermal losses, and operating requirements. Workload scheduling offers another route when computing tasks can move to a less constrained period or a suitable alternative location without violating service requirements. Electric vehicle charging and stationary batteries may provide additional flexibility where the site has suitable equipment, energy reserves, and control arrangements. These resources must operate as a coordinated portfolio because simultaneous cooling recovery, postponed computing, and battery recharging could erase some of the expected reduction.

The Evening Window That Defines The Whole Target

The evening peak exposes the limits of a grid strategy that focuses primarily on adding electricity supply without coordinating the timing of demand. California’s September 2022 heat wave provides a documented example of how extreme temperatures can increase electricity demand while solar generation declines later in the day. On September 6, 2022, the California ISO recorded an all-time peak demand of 51,426 MW. A separate Flex Alert covered the evening of September 9, during the continuing heat wave. These events illustrate why grid operators value resources that can reduce or shift consumption during defined periods of system stress. They do not establish that data centers must supply the required response or that a VPP portfolio could independently eliminate the risk of outages. Their relevance to the DOE’s target lies in the need to align available flexibility with the timing and location of actual grid constraints.

A data center participating in an evening demand-response program must determine which loads it can shift without undermining computing performance or thermal stability. Cooling systems may offer a temporary adjustment window, while flexible workloads may move to later hours if their deadlines and latency requirements permit a change in schedule. Batteries can support the response, but operators must preserve enough stored energy for backup requirements and other site-specific contingencies. The dispatch plan should identify the available reduction, expected duration, conditions that could limit delivery, and schedule for restoring normal operations. Grid operators can then evaluate the site’s contribution alongside residential demand response, commercial building controls, industrial curtailment, and other resources. The plan must also manage the rebound period, because synchronized recovery could create a new demand peak after the original constraint eases.

The Bottleneck Is Not Steel, It’s Sign-Ups

A grid can have connected batteries, controllable cooling equipment, smart thermostats, and flexible industrial loads without being able to dispatch them as one dependable resource. Each device may use a different communications protocol, operate under a separate customer agreement, or remain subject to program rules that restrict when an aggregator can call on it. The DOE’s original Liftoff report identifies simplified enrollment and more standardized VPP operations as important steps toward accelerating deployment. Those measures address a commercial and administrative problem as much as a technical one, because a resource cannot provide a market service until its owner has authorized participation and the relevant parties have established the conditions for control. A capable device outside an eligible program does not contribute to that program’s available dispatch capacity. The challenge involves connecting existing equipment to an operating arrangement that can reliably identify, authorize, coordinate, and verify its response.

The market figures illustrate why adding programs does not necessarily translate into equivalent growth in available capacity. Wood Mackenzie’s 2025 North America assessment reported 37.5 GW of behind-the-meter flexible capacity, up 13.7% year over year, while the number of deployments, unique offtakers, and monetized programs each increased by more than one-third. The report described a market that was broadening faster than it was deepening, with utility program caps, capacity accreditation reforms, and market barriers limiting the rate of capacity growth. This pattern suggests that interest in VPPs can spread across more customers and programs without every new relationship producing a large block of dependable capacity. Some deployments involve small resources, while others encounter restrictions on market participation or the ability to receive compensation for grid services.

Market Access Determines Whether Flexibility Has Commercial Value

The ability to control a resource does not automatically give its owner access to every market that values electricity flexibility. Utilities and grid operators use different procurement mechanisms, qualification requirements, measurement rules, and settlement arrangements, which can affect the services a VPP can offer. Federal Energy Regulatory Commission Order 2222 established a framework for distributed energy resource aggregations to participate in regional wholesale electricity markets, subject to implementation by the relevant market operators and other applicable requirements. The framework does not guarantee identical access across every region or eliminate local interconnection and distribution constraints. An aggregator must understand the market’s qualification rules and determine whether its resources can meet requirements for dispatch, measurement, and performance. The commercial route from a connected device to a recognized grid service therefore depends on the interaction between technology, regulation, and the market in which the resource operates.

Accreditation and measurement create another hurdle because planners need to understand how much dependable capacity an aggregated portfolio can contribute under stressed conditions. A portfolio’s nameplate capability may exceed the response it can deliver after accounting for device availability, dispatch duration, customer overrides, energy reserves, and correlated failures. An operator may need to demonstrate historical performance or satisfy program-specific tests before the market credits the portfolio with capacity. Data center resources introduce additional variables because workload flexibility, cooling headroom, and stored energy can change with computing demand and operating conditions. A robust arrangement should specify the response the site can provide, the conditions under which that response remains available, and the consequences of a shortfall. This process gives grid operators evidence to assess the portfolio before relying on it in resource planning.

No Hyperscaler Flex, No 160 GW — That’s The Quiet Conclusion

The DOE’s 80–160 GW ambition does not require data centers to supply the entire target, but its scale makes broad participation across resource types important. Residential batteries, smart appliances, vehicle charging, commercial buildings, and industrial demand response can all contribute to a coordinated portfolio. Growing electricity requirements from AI infrastructure make it increasingly important to examine whether large computing sites can help manage peak demand. A site that can shift suitable workloads, use stored cooling capacity, optimize auxiliary systems, or dispatch a battery may reduce the electricity it draws during a constrained period. The available response will vary with workload requirements, thermal design, backup policies, and the site’s commercial obligations. Data center participation becomes consequential when it provides dependable, measurable support that complements other VPP resources and matches the actual needs of the grid.

Why Large Computing Loads Must Join the Flexible Capacity Portfolio

The economic opportunity also depends on whether grid operators can procure that flexibility as a defined service rather than treating it as an informal request for cooperation. A data center may offer a limited reduction during a forecast evening peak, a scheduled change in selected workloads, or battery support under an agreed set of conditions. Each option carries different implications for energy consumption, computing performance, resilience, and the duration of the available response. The operator and aggregator must agree on measurement methods, notification requirements, dispatch limits, compensation, and the conditions that allow the site to decline a request. Grid planners must then determine whether that response can defer a specific investment or reduce reliance on expensive peak resources. A clear commercial arrangement helps both parties understand the value being exchanged without promising savings that the site’s actual performance or the grid’s operating conditions cannot support.

Hyperscaler participation also raises a question of operational responsibility that cannot be resolved through capacity forecasts alone. The grid operator needs confidence that a promised reduction will arrive when requested, while the data center operator must preserve critical services and maintain sufficient operating reserves. A program that ignores either requirement could create reliability risks or discourage participation after the first difficult dispatch event. The strongest designs would identify eligible loads in advance, maintain conservative limits, test the response under controlled conditions, and account for the energy needed to recover after dispatch. They would also coordinate with utility planning so that a site’s flexibility supports a defined network or capacity need rather than an abstract system-wide target. Such arrangements can make data centers a useful grid resource without implying that computing infrastructure can replace conventional generation or operate without firm electricity supply. 

What Must Change Before the 2030 Target Becomes Deliverable

The $10 billion annual savings estimate must remain tied to the investment and operating costs that VPPs can genuinely avoid. A portfolio that reduces peaker dispatch may lower operating expenses, while a portfolio that relieves a specific network constraint may support a decision to defer an upgrade. Those outcomes require separate evidence because the value of avoided fuel differs from the value of postponed capital spending. Planners should account for the cost of aggregating resources, compensating participants, maintaining controls, and preserving dependable performance when comparing VPPs with conventional alternatives. They should also avoid counting the same benefit across multiple programs or assuming that a temporary reduction eliminates infrastructure that long-term demand will eventually require. The department’s figure represents a potential system-level benefit under its stated assumptions, not a guaranteed saving for every utility, customer, or data center.

The 2030 deadline makes implementation speed important, but rapid deployment cannot substitute for the evidence needed to establish reliable capacity. The DOE’s ambition calls for 80–160 GW of VPPs, while Wood Mackenzie’s 2025 North American assessment reports 37.5 GW of behind-the-meter flexible capacity. These figures come from different analytical contexts and should not be interpreted as a direct measure of the capacity remaining to be built. They nevertheless illustrate the scale of the deployment challenge and the need to expand participation, deepen existing portfolios, and improve access to markets that compensate grid services. Data centers can contribute if operators convert eligible computing and supporting loads into measurable, contractually defined flexibility while preserving essential operating requirements. The central condition for reaching the target is not that every data center becomes a battery, but that a sufficiently diverse portfolio provides dependable capacity when the grid needs it.