OpenAI has published a "National Community Compact," which sets out its commitments to communities near the data center sites it will own and operate in the future, along with a "Compute Campus Policy Blueprint," a set of policy proposals it wants enacted into law. The commitments cover a wide range of areas, including bearing the costs of power supply, a ban on requiring public officials to sign nondisclosure agreements (NDAs), and disclosure of noise and water use. However, only one site, Effingham County, Georgia, has a specific figure for demand flexibility on power. That site can reduce demand by up to 1,000MW, equal to about 31% of its expected power demand of 3,200MW. The shared document lays out basic principles, and mechanisms to ensure the commitments are carried out appear to be set separately.
The commitments apply to "future sites OpenAI owns and operates"
OpenAI says it will prepare a separate document tailored to local conditions for each data center site it owns and operates going forward. The main shared commitments are five: (1) paying the full cost of power supply and necessary infrastructure; (2) financial guarantees so that other electricity customers do not bear costs if a project is delayed, scaled back, or canceled; (3) a ban on requiring NDAs from public officials; (4) disclosure of noise, water, and power use; and (5) job plans and annual progress reports. For NDAs, exceptions are allowed only for information that needs protection, such as national security information and trade secrets. For noise, OpenAI says it will commission independent studies before operations begin and every year afterward, and publish the results. However, the document does not clearly say whether the commitments apply to sites run by OpenAI's partner companies.
Some sites have specific amounts for community support. In Effingham County, OpenAI will contribute $4.8 million to school meals, following input from residents, supporting all 14,000 students in the county. In Pike County, Ohio, it plans to contribute $6 million to child care support together with SB Energy. At ¥150 to the dollar, those come to roughly ¥720 million and ¥900 million, respectively. However, these are site-specific measures, not conditions that apply to every site.
OpenAI says it has also signed the White House-led "Ratepayer Protection Pledge." According to an article published by the American Nuclear Society on July 29, 2026, signatories of the pledge have grown to 281 across 23 states. Ari Peskoe of Harvard University, meanwhile, has described such pledges as "symbolic." As voluntary pledges multiply, what matters is which parts of the commitments can be verified through concrete figures.
Some items besides power have numerical targets. In the "Compute Campus Policy Blueprint," OpenAI projects that the Pike County site will create 2,500 long-term operations jobs (average annual salary of $80,000) and 26,000 construction-related jobs. These are OpenAI's own projections, however, not results verified by a third party. The document also does not describe any specific mechanism to guarantee that the job targets are met. Having published numerical targets is different from having guaranteed their achievement.
Only one site has a specific figure for demand flexibility
The shared power-related commitments in the "National Community Compact" are four: (1) paying all costs of power supply; (2) providing financial guarantees ahead of large capital investments; (3) supporting long-term planning by telling utilities how much power is needed and when; and (4) participating in demand flexibility programs. The shared document, however, does not include specific costs, guarantee periods, or megawatt figures for the scale of demand flexibility. Those terms will be set individually for each site.
It is worth noting that demand flexibility and cost coverage are different commitments. Demand flexibility is a mechanism for reducing a data center's power consumption when electricity supply is tight, and it says nothing about how much of the cost the company will bear. OpenAI promises to pay "all" of the costs required for power supply but does not give a specific amount. Meanwhile, the Effingham County contract is the only one for which the scale of demand flexibility has been disclosed in megawatts.
In Effingham County, Georgia Power uses an arrangement that lets it reduce power supply to the data center during periods of high strain on the grid. According to the company, this right is based on a 25-year contract. If demand can be flexibly reduced, the utility can build less new generating capacity, it says. In the shared document, OpenAI says it can generally respond to demand reductions with "one hour's notice." However, the specific notice conditions in the Effingham County contract have not been made public.
According to Georgia Power's announcement, the demand adjustable under this contract is up to 1,000MW, equal to 31.25% of the expected demand of 3,200MW. By simple calculation, even at the maximum reduction, supply to the remaining 2,200MW of demand would continue. But 1,000MW is only the upper limit of what can be curtailed, and it is not clear how often or for how many hours reductions will actually occur. The 3,200MW is also a planning assumption, not power consumption measured after operations begin.
Georgia Power describes the contract as "among the largest demand flexibility agreements in the country for a single facility." It also announced that the contract passed review by the Georgia Public Service Commission (PSC). The company says a standard household using 1,000kWh a month could save about $15 a month, or $180 a year, starting in 2029, up from a previous projection of $102 a year. However, that estimate does not come from the OpenAI contract alone; it also factors in contracts with other large customers and future growth in power demand. All of these are Georgia Power's projections, and no independent third-party verification has been confirmed.
Where do the commitments become binding?
There is one consideration in judging the shared document as a statement of principles. The demand flexibility in Effingham County is not merely a voluntary pledge but is based on a 25-year contract that passed PSC review. According to Georgia Power, the contract follows rules for large customers that the PSC approved in January 2025. In other words, at least in Effingham County, there is a contractual obligation backed by specific figures.
The absence of specific figures in the shared document is also not necessarily a problem. The document sets out principles common to future sites that differ in location and scale, so leaving out individual conditions could be seen as a natural design. In that case, however, local residents cannot confirm the scale of demand flexibility or the content of financial guarantees from the shared document alone. To learn the specific terms, they would need to check each site's local document or the contract approved by the state regulator.
What matters is that the binding force of the Effingham County contract comes not from OpenAI's shared document itself but from the state's regulatory system and the contract with the utility. The shared document does not specify any mechanism that guarantees comparable performance. It therefore cannot be concluded from the publication of the shared document alone that specific power cost obligations and demand flexibility terms have been settled at every site.
OpenAI itself acknowledges in the "Compute Campus Policy Blueprint" that voluntary commitments alone cannot keep pace with the scale of current data center development. The document asks state public utility commissions to make large customers pay the additional costs they impose on the grid and to build systems that protect other customers if a project is canceled. In other words, the idea is to leave it to each state's regulatory system to give power-related commitments specific terms and binding force. This assessment, however, covers only commitments on power supply; job creation and local investment need to be examined separately.
What to watch: site-by-site figures and follow-through
To judge how effective OpenAI's commitments will be, two points matter. One is whether each site's local document includes specific terms, such as megawatt figures for the scale of demand flexibility and the amount and duration of financial guarantees. The other is whether OpenAI's promised annual reports publish measured site-level values for noise, water use, and power use. If these become available, outsiders will be able to verify how well the commitments are being carried out.
The same questions apply when Japanese municipalities and businesses evaluate similar pledges on data center development: whether specific figures, the period of application, and the conditions for triggering demand flexibility are spelled out. If OpenAI discloses site-level terms going forward and state regulators apply similar rules to other large customers, local residents will be able to check effectiveness not just through a company's voluntary pledges but through obligations grounded in contracts and regulation.