|
Voice of the Citadel
Getting your Trinity Audio player ready...
|
AI, Nuclear, and America’s New Energy Bargain in 2026
In Brief
- The critical question for an AI investment is when dependable electricity will reach the site, not how much power has been proposed or contracted
- Large electricity users face growing expectations to fund the supply and infrastructure their expansion requires without shifting those costs to existing customers
- A successful project must deliver three outcomes that can be measured separately: reliable service for the buyer, useful capacity for the grid, and benefits consumers can verify
The most consequential date in an AI investment plan may now be the day reliable electricity becomes available.
In July, PJM reported that resources committed for the 2028/2029 delivery year fell 6,831 megawatts short of its reliability requirement. PJM cautioned that the shortfall did not necessarily mean it could not serve customers, but it would leave slimmer reserves and greater operating risk. Nine days later, the White House announced that more than 200 additional utilities, data center developers, cooperatives, and states had joined its Ratepayer Protection Pledge. The announcement reinforced the expectation that large data center operators fund the electricity and infrastructure their growth requires.
The developments bring the same issue into focus from different directions. A company needs confidence that power will be available when its investment is ready to operate. Communities and customers sharing that electricity system need confidence that growth will strengthen the grid without shifting costs onto them. Both claims depend on what is built, when it becomes usable, who pays, and how performance is measured.
In 2024, we examined the pressure that electric vehicles and AI could place on electricity systems. In 2025, we proposed a Revive, Rebuild, Reshape framework linking nuclear investment to grid modernization and operational coordination. The next step is to turn that framework into agreements and operating capabilities that can deliver at a specific location, on a credible schedule, and under real conditions.
That is the test for America’s new energy bargain. A power contract can support an investment, but it cannot alone establish that dependable capacity has been added or that neighboring customers will benefit. Those outcomes have to be designed into the project and demonstrated as it operates.
America’s New Energy Bargain
The federal Ratepayer Protection Pledge, published in March 2026, states the bargain in public terms. It calls on participating large data center operators to build, bring, or buy new or otherwise additive power; pay for the delivery infrastructure their projects require; and negotiate rates for power and related infrastructure brought online to serve them, whether or not they ultimately use it. Where possible, it also calls for capacity that serves the broader public and for backup resources to support the grid during scarcity.
The pledge establishes expectations. Its commercial force depends on the tariffs, contracts, and utility arrangements that assign costs, delivery obligations, and consequences for a particular project. A data center need not own a generating plant to help finance one. What matters is whether its arrangements establish who will fund the required supply and network upgrades, who bears the risk of delay or lower demand, and what the project will contribute beyond its own operations.
A long-term purchase agreement may give a generator the revenue certainty to invest and give a buyer a path toward dependable supply. Neither outcome follows from the announcement alone. Contracted power, generating capacity placed into service, and benefits received by other customers are different measures. The investment case must show how one will lead to the next.
That distinction becomes more consequential as electricity demand grows. National projections establish the scale of the challenge, but a project succeeds or fails at a particular location and on a particular timetable.
The test of America’s new energy bargain is what gets built, when it becomes dependable, who pays, and whether other customers benefit.
Demand Is National; the Constraint Is Local
EPRI’s 2026 analysis projects that data centers could account for 9% to 17% of
U.S. electricity use by 2030, up from the 4% to 5% it estimated then. The wide
range reflects uncertainty about which proposed facilities will be built and how rapidly they will be equipped and used. It also conceals substantial differences among states and local electricity systems.
Greater AI efficiency does not settle the demand question. The International
Energy Agency reports steep declines in energy use per AI task alongside rapid adoption and growth in more energy-intensive uses, including reasoning, video generation, and agentic work. Lower energy use per task can also make that task cheaper and expand its use. Planners need to test both efficiency gains and growth in work volume and intensity, rather than assume one will cancel out the other.
The broader electrification we examined in 2024 remains relevant, but different loads create different planning problems. Residential electric vehicle charging is dispersed and can often move to different hours. A computing campus concentrates demand at one site and may require substantial continuous service. Neither a national data center forecast nor a global electric vehicle sales figure tells a utility what it must deliver to a specific project.
The developer and its utility therefore need a shared, staged account of the project’s likely demand. Corporate investment committees and boards of directors need it to assess capital commitments. Regulators and siting authorities need evidence appropriate to the decisions before them. In each case, announced campus capacity, installed computing equipment, contracted electric service, and actual electricity use should remain distinct.
That discipline protects both the project and the wider system. A campus can be ready before its power supply is available. Generation and network upgrades can also be completed before the anticipated demand arrives. Phased commitments and clear consequences for delay reduce the risk that either gap becomes a cost borne by parties who did not agree to it.
Once demand is tied to a site and a schedule, the decisive question is no longer how many megawatts have been proposed. It is when dependable service can begin.
The Delivery Schedule Is Part of the Product
PJM’s August 2026 interconnection update illustrates the distance between a
development pipeline and operating capacity. It reported that 715 proposed
projects, representing 201.5 gigawatts of nameplate capacity, had qualified for
study in the first cycle of its reformed process. PJM separately reported 51 gigawatts of generation under signed interconnection agreements, while
cautioning that many of those projects remained unbuilt or delayed by permitting and supply chain obstacles.
These figures describe different groups of projects at different stages, rather than power available for immediate use. Qualification for study does not establish a connection date. A signed agreement does not establish that a facility has been constructed, commissioned, and demonstrated to operate reliably. Each milestone reduces a different risk.
For a corporate buyer, the power plan must trace the full path to service: generation, required network upgrades, equipment orders, permits, construction, testing, and commissioning. It must identify the dependencies that determine the opening date and the alternatives available if a critical milestone slips. The same schedule should govern decisions about the campus, computing equipment, and electricity needed to operate them.
Timing changes the economics. An attractive electricity price has limited value if the business cannot operate when planned. A more expensive arrangement may be stronger if it provides credible earlier access and appropriate reliability. Compare alternatives on the expected total cost of dependable service over the project’s operating life, including connection costs, backup arrangements, delay exposure, operating constraints, and contractual obligations.
Finance, operations, technology, procurement, and facilities leaders need to make those choices against one schedule and one risk account. That standard applies to every supply option. It is especially important for nuclear investments, whose commercial promise and delivery timelines can differ substantially by project.
Nuclear’s Opportunity Depends on Execution
The delivery schedule determines how to value nuclear power. Contracting output from an operating plant, expanding an existing facility, restarting a closed plant, and building an advanced reactor are different propositions. Each can contribute to growth, but each has a different path to dependable service.
Constellation’s August 6 update makes the distinction concrete. The company said its 176-megawatt agreement with Walmart would enable a 30-megawatt expansion at the Dresden Clean Energy Center. The agreement establishes a commercial commitment; the expansion is the additional generating capacity it is expected to support. Both numbers matter, but they measure different things.
In the same update, Constellation reported progress toward restarting the Crane Clean Energy Center. Federal Energy Regulatory Commission approval of a waiver to transfer interconnection rights and a Nuclear Regulatory Commission fuel-license amendment cleared important hurdles. Constellation continued to target a 2027 restart. Those approvals improve the project’s prospects; they do not make the plant an operating resource today.
Advanced reactors operate on a longer clock. In March, the Nuclear Regulatory Commission issued a construction permit for TerraPower’s 345-megawatt Natrium project in Wyoming. The Department of Energy identified 2030 as an expected completion year and noted that a separate operating license would be required before the plant could run. In September, TerraPower and Bechtel parted ways over the next phase of construction. TerraPower said it still expected construction to be complete in 2030 and commercial operation in 2031. A permit, a construction target, and dependable electricity available to a buyer are separate milestones.
Together, these projects show how nuclear investment can preserve, restore, and add dependable power. They also show why a buyer cannot treat a future milestone as current supply. A campus scheduled to open before a restart or new reactor enters service needs a credible arrangement for the intervening years. Utilities and investors need contingencies if licensing, construction, or commissioning takes longer than planned.
The strategic case for nuclear is strongest when each facility’s contribution is tied to its actual delivery path. The next decision is how to combine those contributions with other resources so the full power plan can serve the business as it grows.
Build a Portfolio That Can Perform and Adapt
The broader supply mix is changing alongside nuclear investment. In its September 2026 outlook, the U.S. Energy Information Administration forecast almost 2%
growth in electricity sales in 2026, with solar generation growing 21%, wind 7%, and natural gas generation 2%. Those national, annual forecasts describe the direction of supply. They do not establish that a particular campus can obtain power at its connection point during every hour it needs to operate.
Portfolio design begins with the service requirement. A buyer needs to understand its operating hours, interruption tolerance, expansion schedule, emissions objectives, and ability to adjust demand. It can then assess generation, storage, grid access, backup resources, and flexible operations against the same requirements and delivery schedule.
Onsite generation may seem like a faster route when grid connections are delayed, but installed capacity cannot be assumed to be available for other customers. The International Energy Agency estimates that providing reliable onsite gas power for critical, variable data center demand can require generating infrastructure 30% to 70% above the campus’s demand. It also points to turbine supply constraints that may complicate deployment. Equipment reserved to keep the campus running during maintenance or an outage is not automatically dependable surplus for the surrounding grid.
Grid value begins with what a campus can reliably offer after meeting its own needs: how much, for how long, and under whose dispatch authority.
Any promise to export electricity must therefore specify how much can be delivered, when it is available, how long it can be sustained, and who controls its dispatch. The answer depends on fuel, equipment condition, network capacity, operating restrictions, and the campus’s own needs. Demand flexibility offers another possible contribution, but it must also be designed and tested against the conditions under which the grid would call on it.
Once the campus’s needs are covered, the remaining question is what its portfolio can reliably offer the grid under stress. That answer must carry into the project’s agreements with the utility and host community.
Community Benefit Must Be Designed Into the Project
Those agreements must turn a potential contribution into an assessable obligation. They should define what will be available to the grid, when it can be called upon, who has dispatch authority, and how its value will reach other customers.
That last question deserves particular care. Selling electricity into a wholesale market does not, by itself, lower a nearby household’s bill. A direct financial benefit requires an identifiable route, such as a customer credit, payment to a participating household, or a rate arrangement that allocates savings. Improved reliability and avoided infrastructure costs may also benefit customers, but they need their own measures and should be described on their own terms.
Large buyers can contribute through flexibility as well as supply. In March, Google reported that it had incorporated one gigawatt of data center demand-response capacity into long-term utility contracts. Its approach includes shifting or limiting some machine-learning workloads when the electricity system needs relief. In June, Google and Voltus announced a three-year agreement intended to unlock up to 100 megawatts of capacity from flexible resources in the PJM region, including batteries and smart thermostats. That arrangement provides for payments to participating homes and businesses.
These are two distinct ways to support the grid: a large customer can change its own consumption, and it can help mobilize resources across the community. In either case, announced capacity is a commitment to be tested against actual performance. Useful measures include how much demand was reduced, or power was delivered, for how long, and what participating customers were paid.
The community bargain becomes credible when people can follow that chain from commitment to operation to benefit.
Trust Requires Operating Proof
Electricity investment has become more visible to consumers as bills rise and large projects seek access to local grids. In a Pew Research Center survey conducted in March 2026, most respondents who reported higher home energy costs viewed data center energy use and grid upgrade costs as at least minor reasons for the increases. The survey records perceptions, not the cause of any particular bill. It shows why promises about cost protection and community benefit need evidence people can understand.
Transparency begins with clear distinctions. Forecast demand is not electricity consumed. Contracted power is not power delivered. Installed equipment is not necessarily dependable capacity. A promise to reduce demand during grid stress is not an operating capability until it has been tested under defined conditions. Reporting that combines these stages into one favorable number obscures the result.
The underlying records should connect the original commitment with what happened: when supply became available, how the facility’s demand developed, whether flexibility performed when called upon, what infrastructure costs were incurred, and what benefits reached other customers. Commercially sensitive details need not be public for the resulting account to be credible. Each audience should be able to trace a reported outcome to the same operating evidence.
Visibility must also lead to action. Each material commitment needs an owner, a performance threshold, a test, and a decision process when results differ from the plan. If a generation milestone slips, demand rises faster than expected, or a flexibility test fails, leaders need to see the exception soon enough to adjust the operating plan and address its financial consequences.
Consumers who contribute through batteries, electric vehicles, smart equipment, or changes in when they use power need the same clarity. Participation terms should explain what customers can control, how performance and compensation are calculated, how personal information is handled, and when a customer can change or end participation. Consumer visibility means more than seeing a system-wide megawatt figure. It means understanding what one’s participation accomplished and what one received for it.
Consumer visibility means more than seeing a system-wide megawatt figure. It means understanding what one’s participation accomplished and what one received for it.
Executives, utilities, regulators, and communities will need different views of this information, but those views should reconcile. Executives need to govern growth, cost, and operating risk. Utilities and regulators need to assess dependable resources and cost allocation. Residents and businesses need an intelligible account of what was promised, what was delivered, who paid, and who benefited.
A commitment may change as a project develops. Trust depends on making the change visible, assigning responsibility, and showing how the resulting decision was made. That operating discipline gives the Revive, Rebuild, Reshape agenda practical meaning.
Revive, Rebuild, Reshape: The 2026 Execution Agenda
The Revive, Rebuild, Reshape framework introduced in our 2025 article remains useful. In 2026, each part needs a delivery test: what capability will it provide, when will it be available, and how will you demonstrate its performance?
Revive begins with existing assets that could supply dependable power sooner than a new build. Assess operating plants, viable restarts, and capacity expansions for achievable output, readiness date, remaining approvals, and network requirements. The test is how much power can enter service when the system needs it.
Rebuild aligns new generation, storage, and network investment with substantiated demand. Capital should be released against defined development milestones, with contingencies for equipment delays, slower customer expansion, and changes in the required supply mix. The commercial agreements supporting long-lived infrastructure must also remain workable if the buyer’s plans change.
Reshape concerns how a large customer operates once it connects. Flexibility has value only if the customer can reduce or move demand without disrupting essential services, and if that response can be called upon and verified. It requires more than a contractual promise. Workloads, equipment controls, backup arrangements, operating authority, and tests must support the obligation.
PJM’s August 13 Interim Resource Adequacy Service proposal illustrates the direction of travel. As proposed, certain new large customers that do not bring sufficient additional supply would reduce or transfer demand during severe grid conditions before traditional customers faced interruption. States and utilities would identify affected loads and allocate retail costs. For a corporate buyer, the immediate question is whether its facilities can deliver the flexibility its power arrangement assumes.
An enterprise cannot commit flexibility it cannot identify, control, and verify. That is the operating problem we described for GryphX Energy in 2025. Across distributed facilities, the questions are which loads can respond, who may direct them, whether they responded, and what happened to service and cost. Those records should reconcile with utility data and the agreed measures of grid and community benefit. They can also support a clear account of what participating consumers contributed and how they were compensated.
Boards of directors should receive a consolidated view of dependable power available by date, total service cost, exposure to delay, tested demand flexibility, and verified community benefit. These measures reveal choices in the original business case. Earlier access may require greater flexibility or higher cost. A community commitment may require additional equipment, network investment, or a different revenue allocation. Leaders can govern those choices when the commercial plan and operating evidence tell the same story.
Turning Private Growth Into Public Strength
Large electricity users can support infrastructure that serves more than their own expansion. Their long-term commitments can help finance generation and grid investment. Nuclear projects can preserve, restore, or add dependable supply on different schedules. A portfolio of resources and flexible demand can help the system operate through periods of stress.
The value of those contributions must be established separately. A purchase agreement does not demonstrate that new capacity has entered service. Installed equipment does not establish that power can be delivered when the grid needs it. An announced community benefit does not establish that a customer received it.
America’s new energy bargain will succeed when those gaps are closed in practice. An enterprise must be able to show when dependable power became available and at what cost. The electricity system must show what capacity or flexibility the investment added. Communities and consumers must be able to see what they paid and what they received. When all three accounts reconcile, private growth can help create the capacity for broader growth.
Put Your Power Plan to the Test
A credible plan connects the date dependable power reaches the site with who pays, what the grid can rely on, and what customers can verify. Gryphon Citadel and GryphX Energy can help you test those commitments against the project’s delivery schedule and operating capabilities.
About Gryphon Citadel
Gryphon Citadel is a management consulting firm headquartered in the Philadelphia area, with a European office in Zurich, Switzerland. We help leaders make consequential choices and build the capabilities to carry them through. Our work spans strategic planning and execution, digital and organizational transformation, AI and advanced technology, performance improvement, supply chain and manufacturing optimization, workforce development, and operational planning and control.
We bring these disciplines together to address complex business challenges from diagnosis through implementation. By aligning leadership priorities, operating capabilities, and measures of success, we help clients adapt to changing conditions and achieve results they can see in their organizations’ performance.
Learn more at gryphoncitadel.com.
