An AI Data Centre Is a Community Deal

The United Kingdom's Lanarkshire AI Growth Zone uses another model: government-backed industrial policy, private investment, a community fund, apprenticeships and conditional delivery milestones. The announcement separates future permanent AI and data-centre roles from the larger construction workforce and states that the jobs and investment figures were supplied by the developer. That distinction is essential when leaders assess the local return. ³⁸

Across these cases, local value is no longer a charitable add-on. It is part of the delivery architecture. Skills determine who can operate the facility. Grid and water arrangements determine timing and cost. Community confidence affects planning and political durability. Local compute access determines whether the region hosts infrastructure or builds capability around it.

The compact should name six things. First, infrastructure: new supply, connection timing, transmission, water use and who carries each cost. Second, work: construction jobs, permanent roles, apprenticeships, completion rates and pathways into local employment. Third, local capability: procurement, research access, startup compute and education partnerships.

Fourth, community governance: which local, civil-society and First Nations constituencies have standing, how concerns are resolved and what information remains public. Fifth, outcomes: the small set of measures reported annually by an owner independent enough to challenge the sponsor. Sixth, reset rights: the thresholds for renegotiation when demand, water, grid costs, ownership or project timing changes.

Each commitment needs a baseline, a deadline and an accountable party. A fund announced is not benefit delivered. A training place is not a durable job. A power purchase agreement is not proof that household or system costs are protected. The compact turns attractive claims into decisions that can be governed.

The International Energy Agency estimates that data-centre electricity demand could approach 950 terawatt-hours by 2030, with AI-focused demand rising especially quickly. Its analysis also stresses geographic concentration, infrastructure lead times and growing concern about social acceptability. A global percentage can look manageable while a local cluster creates a significant connection, generation or affordability problem. ⁴⁵

That is why one market cannot stand in for another. In the United States, the bargain may revolve around utilities, tax arrangements, local employment and new generation. In the United Kingdom and Europe, planning, heat reuse, industrial policy and sovereign capacity can feature more strongly.

In parts of Asia, Africa, the Middle East and Latin America, grid capacity, water availability, land, local-language institutions and access to domestic compute can change the order of priorities. Leaders should use the same compact fields everywhere but negotiate the thresholds locally.

A workable partnership needs more than the developer and the approving authority. The utility or system operator must make connection timing and cost visible. Workforce and education institutions must connect training to actual roles. Community and First Nations representatives need standing before the proposal hardens. An independent measurement owner must publish progress without allowing the sponsor to grade itself.

Developer initiatives show what may be possible. OpenAI's Stargate Community programme describes site-specific plans, infrastructure and water commitments, workforce engagement and community investment. Stargate Norway couples planned capacity with renewable power, closed-loop cooling, excess-heat possibilities and priority access for the local ecosystem. These are supplier announcements, not independent outcome studies, so they are useful design inputs rather than proof. ⁶⁷

The strategic advantage is not simply fewer objections. A stronger coalition can surface grid, water, workforce and legitimacy risks early enough to change the design. It can also make local benefit more credible to employees, customers, investors and governments.

Choose the organisation's largest planned AI-infrastructure commitment. In the first ten days, map the parties whose systems or consent affect delivery and separate supplier claims from confirmed baselines. By day 20, draft the six compact fields with costs, measures, owners and regional variants. By day 30, test the draft with one representative from the utility or public system, workforce or education, and the affected community.

Return to the investment committee with three visible items: unresolved infrastructure costs, promises without an outcome measure and constituencies with no standing. Approve only when the economic case includes those dependencies and a named executive owns the public compact.

The aim is not to slow every project. It is to prevent a fast announcement from creating a slow delivery problem. The organisations that negotiate local value early will be better placed to secure capacity, talent and confidence when infrastructure competition intensifies.