The data center conversation is usually framed as a real-estate, semiconductor, or cloud question. For leaders, it is increasingly a coordination problem across electricity, water, land use, permitting, cybersecurity, workforce, financing, and the communities that host the infrastructure.
DOE’s current AI agenda treats data-center growth as a grid, permitting, and community coordination problem. Its 2026 materials note that data centers could consume up to 9% of U.S. electricity by 2030, while a July partnership in Kentucky pairs an anticipated 1.8-gigawatt AI and high-performance-computing campus with generation, transmission, and storage planning. The strategic question becomes broader: how can an organization grow compute without creating a reliability, affordability, or legitimacy problem elsewhere in the system?
Compute is now an infrastructure portfolio
A site decision can commit the enterprise to years of energy, network, cooling, and supply-chain dependencies. Those dependencies should be visible before the business case is approved. A useful portfolio map includes load shape, interconnection assumptions, backup strategy, water and thermal constraints, fiber diversity, equipment lifecycle, security boundaries, and the conditions that could make the project politically or operationally infeasible.
The point is not to slow expansion. It is to make the full system legible enough to choose the right sequence: where to build, what to defer, which partnerships to form, and what flexibility the design needs to preserve.
Permitting is part of the operating model
Permitting is sometimes assigned to a specialist function while the program team continues planning as if approval were a date on a schedule. That creates avoidable rework. Community concerns, environmental evidence, grid studies, and local workforce commitments affect the design itself.
Treat the permitting path as a service with users, decisions, evidence, dependencies, and feedback. Give the program a shared fact base and a clear escalation rhythm. When stakeholders raise a concern, the organization should be able to show how it changes the design or why it does not—not simply route the question to another team.
The grid connection is another governance surface. FERC Order No. 1920 requires regional transmission providers to plan for long-term needs and establish cost-allocation methods for facilities selected through that process. For data-center leaders, that means interconnection and transmission assumptions belong in the operating case, not in a late-stage utility conversation.
Efficiency is a management capability
Power and compute efficiency are not only hardware metrics. They are choices about workload placement, scheduling, cooling, data movement, model use, and the operating rules that determine when capacity is added. Leaders need a common language for cost, carbon, reliability, and service performance so that a local optimization does not create system-level risk.
What leaders can do now
Before committing the next major capacity decision, convene a system review across technology, energy, facilities, finance, risk, public affairs, and operations. Map the critical assumptions, the reversible and irreversible choices, and the evidence required to move each gate. Then publish a small set of measures that show whether growth is improving capability without exporting hidden cost.
Global Enterprise helps leaders connect digital infrastructure strategy to the operating model, evidence system, and stakeholder decisions required for durable growth.