The Race for Power: How the AI Infrastructure Boom Is Reshaping Power Development in the United States

Background

Prior to the recent rise in load demand driven by the AI infrastructure boom, many power projects in the United States followed a traditional development timeline and approach – a developer would obtain all necessary permits, land and contracted offtake to support construction financing for the power project. In a typical project, the power end user would not share in the development risk or costs. As end users of data centers compete to secure significant amounts of reliable power on compressed timelines, the traditional allocation of risk between power developer and power end user is shifting. For data center end users, certainty and speed of delivery can matter as much as (and in some cases, more than) price. At the same time, data center developers are facing increasing headwinds to their ability to deliver power solutions for large-scale data center projects on budget and on time, including rising prices, supply chain constraints, permitting delays, constrained transmission and interconnection infrastructure and changes to regulatory requirements.

The Race for Power

Against this backdrop, some end users are willing to contribute to the improvement of grid infrastructure and accept a greater share of development and regulatory risk to secure power for their projects.

For example, under Google’s electric service arrangement with Xcel Energy for its new Minnesota data center, Google will pay new grid infrastructure costs associated with its project and Google and Xcel Energy are partnering to bring approximately 1,900 MW of new clean energy resources to the grid. Georgia Power’s 25-year agreement to serve OpenAI’s 3.2 GW “Project Camellia” provides another recent example. Based on publicly disclosed terms, OpenAI will bear the infrastructure and electric service costs required to serve the project and has committed to provide up to 1,000 MW of flexible demand response. These arrangements illustrate how end users may be willing to assume both incremental infrastructure costs and, in some cases, greater load-side flexibility to secure long-term access to power at significant scale.

These types of arrangements may also be required in certain jurisdictions, as communities are increasingly demanding that data centers must bring their own power, or pay for the infrastructure improvements necessitated by higher load brought by a data center. In March 2026, seven leading technology companies accepted the White House’s Ratepayer Protection Pledge, committing, among other things, to build, bring or buy the generation required to satisfy their data center demand and pay for associated power delivery infrastructure upgrades.

Development Risk Moves Across the Table

A related dynamic is emerging between end users and power project developers. Whereas an end user’s main objective may be to secure power by a particular date, pushing to allocate development risk to the developer may be counterproductive to that objective if doing so increases the risk that the data center capacity will not be available on the required timetable.

Power project developers have traditionally borne significant development risk associated with obtaining required permits, achieving related milestones and securing interconnection. Permitting delays, interconnection queue delays, network upgrade requirements and transmission constraints can each materially affect project costs and schedules. As competition for new generation and grid capacity intensifies, however, there are signs that end users may be willing to share more of the power project development risk, through milestone extensions, cost-sharing arrangements, equipment reimbursement agreements or other negotiated protections, rather than leaving the associated exposure primarily with the developer.

Change-in-law exposure provides another example. Tariffs, tax-credit eligibility and other regulatory changes can materially alter power project economics after contract execution, but before procurement or construction is complete. Tenants that are very focused on certainty of power supply are increasingly willing to address regulatory costs through price adjustments, cost-sharing mechanisms, reopening rights and other contractual protections.

The result is not necessarily that end users assume responsibility for developing or completing the power project. Rather, the allocation of contractual and economic consequences of risks that neither party can fully control may be increasingly negotiated, with bespoke solutions being required to address the goals of the power project developer and end user. This need for creative solutions arises where the power project developer and the data center developer are third parties, but also for behind the meter projects where the power generation and the data center have common ownership, which are becoming much more common.

Bankability as a Shared Objective

For a power project seeking non-recourse or limited recourse project financing, senior creditors will want to be satisfied that risks outside the developer’s control have been adequately addressed and are unlikely to prevent the project from reaching commercial operation or materially impair the revenues supporting repayment of project debt. The interests of the power end user and the power project lender are highly aligned, but increasingly end users are willing to take on some of the development and regulatory risk to facilitate raising of financing by the power project developer. Bankability can therefore become an increasingly shared objective between power project developers and data center developers. Where allocating a particular risk entirely to the power project developer could impair the power project’s ability to obtain financing (at all or on acceptable terms) or reach commercial operation, an end user seeking certainty of supply may have more incentive to accept or share that risk.

Outlook

Increasing power demand driven by (among other things) the AI infrastructure boom is driving industry participants to find new and creative ways to allocate risks and rewards. On the load side, some data center developers and end users are assuming significant infrastructure costs and operational obligations associated with serving their demand. On the generation side, the urgency of securing power can support greater sharing of development and regulatory risks where doing so preserves power project economics, financing and/or delivery schedules.

For project participants, the practical consequence is that power purchase negotiations are increasingly focused beyond achieving the best price outcome. Interconnection and infrastructure costs, permitting milestones, change-in-law protections, tax-credit eligibility and contract tenor can each determine whether a project is capable of being financed and delivered on the timetable required by the buyer.

As the race for power continues, successful transactions may increasingly depend on allocating the power development risks in a manner that both protects the end user and leaves the power project capable of attracting the necessary investment to reach commercial operation.

* * *