OpenAI Community Compact: Data Center NDAs, Power Costs

OpenAI released a set of public commitments on Thursday, November 5, 2026, that changes how it says it will negotiate and operate data centers in the United States. Called the National Community Compact, the document commits the company to stop using nondisclosure agreements with public officials, pay the full cost of the electricity and grid infrastructure its facilities require, and publish site-level data on water and power use.

The announcement came two days after U.S. voters, in the midterm elections, backed restrictions on data center development or rejected pro-development proposals in 40 of 41 local ballot measures across 10 states, according to tallies cited by CNBC and Ballotpedia. One outlet counted 39 measures, so the exact total varies by source. The next day, AI and chip stocks had their worst session since April 2025 as investors priced in a harder political climate for the AI buildout.

What OpenAI is promising

According to OpenAI and coverage of the announcement, the compact contains several main commitments:

• Power costs. OpenAI says it will pay the full electricity and required infrastructure costs at every data center it owns and operates, so local ratepayers do not cover the cost of new grid capacity built to serve its campuses.
• No NDAs with officials. The company says it will no longer use nondisclosure agreements with public officials on data center projects.
• Water planning. OpenAI will develop a water plan for each campus.
• Public reporting. The company says it will publish site-level water and electricity use.
• Pre-construction disclosure. OpenAI will release construction plans and independent noise studies before a facility begins operating.

The compact also includes a policy ask. OpenAI urged state utility commissions to require large electricity customers, including data centers, to cover the costs they add to the grid. That proposal would apply to the whole industry, not just OpenAI.

The Georgia backdrop

Several outlets connected the NDA commitment to a dispute in Effingham County, Georgia, where OpenAI announced Project Camellia in July. The campus is planned at 3.2 gigawatts and is the company’s first self-built data center (see our coverage of OpenAI’s Project Camellia). Residents there have alleged that officials used NDAs, code names and closed-door meetings to advance the OpenAI-linked campus before the public could react.

That kind of secrecy is common in data center site selection. Developers often negotiate land purchases, tax abatements and utility agreements through shell companies or under confidentiality agreements, in part to keep competitors from learning their plans and in part to avoid driving up land prices. Opponents say the practice keeps residents from weighing in until decisions are effectively final.

Georgia’s politics have also changed. In Tuesday’s elections, Democratic candidates who campaigned partly on pausing data center development won governor’s races in Georgia and Texas, offices the party had not held in decades. Goldman Sachs estimated that Democratic statehouse wins in Texas, Michigan and Georgia could put roughly 45% of planned U.S. data center capacity at risk.

Following Amazon

OpenAI is not the first large builder to drop official NDAs. Amazon said in October that it would stop using nondisclosure agreements with public officials on data center projects. The two moves suggest the largest AI infrastructure buyers now see secrecy in site negotiations as a liability rather than a competitive advantage.

The push on power costs follows a broader regulatory trend. In July, New York became the first state to pause new hyperscale data center permits while it writes a new framework covering grid, water and ratepayer impacts (see New York’s hyperscale moratorium). In September, Texas paused new large-load grid connections and ordered an audit of an interconnection queue that had swollen to an estimated 474 gigawatts (see Texas’s data center power freeze). The central political question in both states was whether household and small-business ratepayers would end up paying for transmission and generation built for AI campuses.

What the compact does not cover

Coverage of the announcement noted several limits.

It is voluntary. The pledges are not legally binding, and OpenAI did not attach any enforcement mechanism or independent auditor to them.

There are no dollar figures. The compact commits OpenAI to pay full power and infrastructure costs but does not say how much that will be or how the costs will be calculated. In practice, utilities and regulators decide which grid upgrades are attributable to a specific customer, and those calculations are often disputed.

It covers owned sites only. The commitments apply to data centers that OpenAI owns and operates. Much of the company’s compute comes from partners and cloud providers that build and run facilities on its behalf. The compact does not clearly say whether the same terms apply to those third-party builds. That matters because only a portion of OpenAI’s planned capacity is self-built.

Industry and political context

The compact arrives during a difficult period for OpenAI’s infrastructure plans. The company has outlined plans to spend hundreds of billions of dollars on compute through the end of the decade (see OpenAI’s $750 billion compute plan), and those plans depend on getting permits, power contracts and local approval in multiple states at once.

Analysts quoted after the election said they doubt any new federal or state legislation could meaningfully stop data center construction. They expect more hearings, investigations and local fights instead. The ballot results suggest those local fights are now the bigger obstacle: voters in many communities have shown they are willing to vote projects down when given the chance.

What it means

The compact is OpenAI’s attempt to get ahead of the local opposition rather than fight it project by project. Two of the commitments, paying full power costs and dropping official NDAs, address the two most common complaints from opponents: that residents subsidize data centers through higher electricity bills, and that projects are approved before the public learns about them.

Who gains. Local officials get political cover. Approving a project is easier when the developer has publicly committed to pay its own grid costs and disclose its water use. Utility commissions get a large company publicly supporting the cost-allocation rules many of them are already considering. And OpenAI gains a reputational advantage over rivals that still use confidential site negotiations.

Who loses. If utility commissions adopt OpenAI’s proposal that large customers pay for the grid capacity they add, it would raise costs for every data center developer, not only OpenAI. Smaller developers and neoclouds with thinner margins would feel that more than a company that has raised tens of billions of dollars. In that sense, the policy ask also gives OpenAI a competitive advantage.

The credibility test. Because the compact is voluntary, its value depends on what OpenAI does next. The clearest signals will be:

• Effingham County. Whether OpenAI publishes construction plans, noise studies and water data for Project Camellia, and whether residents and local officials accept the commitments.
• Partner sites. Whether OpenAI extends the terms to data centers built for it by cloud providers and developers, which would cover far more of its capacity.
• Utility filings. Whether OpenAI’s power-cost pledge shows up in actual tariff agreements with utilities, with specific numbers.
• Copycats. Whether Microsoft, Google, Meta and Anthropic adopt similar commitments. Amazon’s October move and OpenAI’s compact could make no-NDA, pay-your-own-way terms a baseline for the industry.

The AI industry’s constraint is increasingly not chips or capital but approval from local communities. OpenAI’s compact shows the company recognizes that. Whether voters accept it will become clearer as the next projects go before local officials.