OpenAI Publishes Community Compact With Power, Water and Disclosure Commitments for Its Data Centers

OpenAI has published a National Community Compact setting out what communities can expect from its future data centers, with commitments on electricity costs, water use and disclosure. The company says it will pay the full cost of the electricity its facilities consume, along with the infrastructure buildouts and upgrades that power providers need to serve them. It also says its data center operations will not affect the electricity costs of nearby communities. (Engadget)

The compact is accompanied by a Compute Campus Policy Blueprint, which proposes policies intended to give communities a voice in developments where they live and to shield them from new costs those developments would create. The blueprint is a proposal. Its adoption depends on lawmakers, regulators and utilities, none of whom are bound by the company's document.

On water, OpenAI pledged to protect local resources and to develop a water plan for each community where it is building facilities. The company said it will prioritize closed-loop or low-water cooling, in which water circulates through a sealed circuit that absorbs and removes heat without being consumed, much like the coolant loop in a car engine. It also said it will publish water and electricity use for each data center site. Per-site reporting is the element most amenable to independent checking, because it creates a baseline that local authorities and journalists can compare against utility interconnection filings and water-permit figures.

OpenAI also said it will not sign non-disclosure agreements with public officials, and that it will create jobs for local workers and businesses. Those are the two commitments most directly tied to the way large projects have often been negotiated, frequently through confidential site-selection processes that local officials could not discuss publicly.

The company said it had been working on these data center updates for months and had planned them before the 2026 U.S. midterm election results were released. A Reuters report from July described opposition to AI data centers as one of the rare issues galvanizing voters across the political spectrum in the midterms, and that pressure has been building in statehouses and county commissions. Resource questions have also surfaced outside the United States. A separate Reuters report from August described Google's $15 billion India data centre project as facing water and wildlife concerns.

The compact sits alongside earlier commitments. In March, the White House published its Ratepayer Protection Pledge, under which signing companies agree to protect American consumers from price hikes attributable to data center energy use. OpenAI's Effingham County project in Georgia was announced in July with $80 million in community benefits over the life of the project, according to the company. OpenAI has also said its Stargate campus in Port Washington, Wisconsin, will generate more than 1,000 long-term jobs.

On November 2, OpenAI introduced its Grid Protectors Program, under which the company will work with utilities to help future compute campuses support grid reliability. That program is the operational counterpart to the cost pledge. A campus that can shed or shift load during stress events, reducing or moving its power draw when the grid is strained, is a different proposition to a utility than one that draws a flat, uninterrupted block of megawatts. Separately, OpenAI's Supplier Code of Conduct, updated in October, requires suppliers to identify, assess and manage environmental risks associated with resource-intensive operations, which extends the water and energy commitments into the supply chain.

In my view, the central question for engineers and planners is whether these pledges become enforceable terms. A commitment to pay the full cost of electricity and grid upgrades is only as meaningful as the tariff structures, interconnection agreements and cost-allocation rules that define it. Per-site disclosure of water and power draw, if it arrives in a consistent format, would let a reader test the claims against metered data rather than against the document itself.

The broader context here is that data center siting has become a question of local infrastructure as much as cloud capacity. Closed-loop cooling trades water for energy and capital cost, and grid-aware load management asks campuses to behave more like flexible industrial customers than fixed loads. If the compact's commitments hold, they could make large campuses easier to site, with clearer cost allocation and a known set of resource obligations. That outcome would benefit both the operators and the communities hosting them, and it is the version of this story worth watching closely.