OpenAI’s Data-Centre Compact Sets a Useful Baseline
Emmanuel Chidi · Published November 6, 2026 · Business & Policy
OpenAI has published a community compact for future compute campuses it owns and operates. The commitments include paying the full cost of electricity and supporting infrastructure, publishing site-level water and power use, commissioning independent noise studies and avoiding nondisclosure agreements with public officials except for narrow legitimate protections.
Those promises are aimed at U.S. projects, but they arrive days after Canada published its own responsible data-centre development principles. The overlap creates a useful checklist for Canadian communities evaluating large AI facilities. It also exposes the limits of voluntary promises.
What OpenAI is promising
OpenAI’s National Community Compact and policy blueprint say every future campus it owns and operates will receive a local agreement. The company says it will:
- cover the electricity and infrastructure costs created by the project;
- provide financial assurances before major utility spending begins;
- publish site-level water and electricity use;
- prioritize closed-loop or low-water cooling;
- publish construction plans and noise studies before operation and report annually afterward;
- disclose expected jobs and report on hiring and apprenticeships; and
- allow community input to shape local investments.
Independent reporting from Data Center Dynamics highlights an important boundary: the compact covers sites OpenAI owns and operates, so it may not extend to capacity leased from partners. The Washington Post also reported that the commitments followed local voter resistance to data-centre development.
Where Canada’s principles line up
Canada’s Responsible Data Centre Development Principles call for early engagement with host communities and Indigenous rights holders, transparent information on power, water, sound and emissions, local procurement and training, and measurable benefits for Canada.
Both documents therefore agree on four basics:
- large users should not quietly shift grid costs to households;
- water, power and noise impacts should be disclosed at the project level;
- communities should be involved before plans are fixed; and
- jobs and local benefits should be measured, not merely advertised.
Canada adds a national-resilience lens. Its principles say facilities serving sensitive data, public services and critical infrastructure should meet requirements for Canadian legal protection, cybersecurity, continuity and reliable Canadian access. OpenAI’s compact is more specific about financial assurances, annual reporting and limits on official NDAs.
The gaps Canadian permits still need to close
Neither document is a substitute for enforceable utility decisions, environmental reviews, zoning or procurement contracts. A strong Canadian approval should turn broad principles into project-level conditions:
- a published electricity forecast and a clear rule for who pays for new generation and transmission;
- site-level water budgets with drought and emergency plans;
- independent noise baselines and complaint procedures;
- public incentive terms with clawbacks when job or investment promises are missed;
- Indigenous consultation and benefit arrangements where rights and lands are affected; and
- continuity and Canadian-access terms for capacity receiving public support.
Mapletechie’s earlier reporting on Anthropic’s proposed Nova Scotia data centre found that a real lease can coexist with unresolved timelines, power plans and community benefits. Our review of DayOne’s data-centre filing also showed how quickly power costs and debt grow with capacity.
OpenAI’s compact is useful because it makes several expectations concrete. Canadian regulators should treat it as a floor companies can already meet, then add enforceable local, Indigenous, grid and sovereignty requirements rather than relying on a corporate promise alone.
Tags: OpenAI, data centres, AI infrastructure, electricity, Canada