Who Can Block a Data Center

Executives usually ask about AI infrastructure in terms of models, clouds, vendors, chips, and workloads. Those details matter. They also rest on an idea that is getting harder to defend: that the physical side is simply ready for purchase.

That idea is weakening.

AI is turning data centers into visible industrial infrastructure. Sites need land, substations, transmission, cooling, crews, long contracts, and local political support. Capacity arrives as a request for power and space, not as a software subscription.

That shifts the strategy problem. A company that treats AI capacity as simple procurement may discover that capacity also depends on approvals. A utility commission, town council, governor, grid operator, water district, lender, or nearby residents can block progress when they see no local benefit.

The C-suite should treat compute as a permissioned operating asset, not an abstract input.

The Signal

Demand is rising quickly, and the binding constraints are moving from chips alone to the full stack of infrastructure approvals.

International Energy Agency estimates put data-center electricity use at about 415 terawatt-hours in 2024, roughly 1.5 percent of global consumption, with a base-case path toward roughly double by 2030. A more recent agency summary puts 2025 data-center electricity use near 485 terawatt-hours and 2030 demand near 950 terawatt-hours. AI-focused data centers are growing much faster than the total.

In the United States the pressure is more concentrated. A U.S. Department of Energy (DOE) Lawrence Berkeley National Laboratory report said U.S. data centers consumed about 4.4 percent of total U.S. electricity in 2023 and could reach 6.7 to 12 percent by 2028. The Electric Power Research Institute (EPRI) 2026 update puts the 2030 range higher still: 9 to 17 percent of U.S. electricity, depending on how many projects clear supply-chain and permitting hurdles.

Capital belongs in the same frame. McKinsey estimates that global data-center demand could almost triple from 2025 to 2030, from about 82 gigawatts to about 220 gigawatts. Goldman Sachs Research says private-market financing will become more important because hyperscalers may spend roughly $5.3 trillion on AI and data centers by 2030. Another Goldman note put 2026 hyperscaler AI capex consensus at $527 billion and warned that supply bottlenecks and investor appetite may constrain spending more than balance sheets.

Politics is next. The White House's 2026 Ratepayer Protection Pledge is a useful sign that data-center growth has become a household-bill issue. The pledge asks major AI and data-center companies to build, bring, or buy new power supply, pay for power-delivery upgrades, and negotiate separate rate structures so ordinary households avoid subsidizing the load. An Associated Press report this week described legislation aimed at forcing AI data centers to obtain electricity and water from private sources. More than 20 Florida counties and municipalities had rejected, delayed, or paused major data-center projects for at least a year.

Local planning is where a strategic constraint becomes visible.

The Veto Map

Boards do not need utility expertise. They do need a map of where the AI roadmap can be stopped.

The Strategy Error

The easy error is to call this a sustainability issue and hand it to the environmental, social, and governance team. That frame is too small. Complementarity is more useful.

AI advantage depends on complementary assets: data rights, workflow redesign, risk controls, distribution, talent, and now physical capacity. When one complement is scarce, the value of the others changes. Better models buy little when inference economics are unattractive. Stronger agent roadmaps stall when production capacity sits behind a delayed substation. Vendor relationships leave exposure intact when the vendor also competes for power, cooling gear, construction labor, and capital.

The data-center veto belongs in corporate strategy for that reason. It changes build-versus-buy choices, workload placement, which AI use cases deserve scarce capacity, procurement diligence, resilience design, and how the company explains AI to regulators, customers, employees, and communities.

The highest-risk companies hold two incompatible beliefs at once: that AI will soon touch core operations, and that the physical cost and permission structure behind that AI can stay invisible.

What Buyers Should Ask Vendors

Enterprise buyers need not interrogate every megawatt behind a vendor. They should stop treating capacity as magic.

Ask four questions before moving a strategic workflow onto an AI platform.

Where is the capacity assumption? If the product roadmap depends on larger models, lower latency, richer context, more agents, or heavier inference, the buyer should know whether the vendor can support that growth without sudden pricing, availability, or service-level changes.

Who bears infrastructure cost volatility? If power, equipment, and financing costs rise, the buyer needs a clear answer: absorb margin pressure, change pricing, narrow usage, or shift architecture.

What is the fallback architecture? Critical workflows need a plan for degraded model quality, lower-volume inference, regional capacity limits, or slower rollout. AI resilience includes uptime. It also means the business process can keep running when compute is constrained.

Which claims rest on demand evidence rather than demos? Capex can signal confidence. Capex can also signal that the industry is racing ahead of monetization evidence. Buyers should separate useful capacity from speculative abundance.

The Board Agenda

The board-level version is simple: require every material AI program to name its physical dependency.

That dependency may be direct, as when a company builds or leases dedicated infrastructure. It may be indirect, as when a company depends on a hyperscaler's buildout. Either way, the operating plan should answer five questions.

1. Name the workflows important enough to deserve scarce and possibly expensive AI capacity.
2. Name the vendors or infrastructure partners that control the capacity path.
3. List the power, cooling, permitting, and tariff assumptions behind the roadmap.
4. State the plan if capacity is delayed, rationed, repriced, or politically constrained.
5. Define evidence that would slow the AI buildout rather than spend through the constraint.

The last question is the most important. Infrastructure enthusiasm often turns forecasts into commitments before demand quality is proven. Stage commitments like real options: buy learning, preserve flexibility, and commit heavily only where usage evidence and operating advantage are both strong.

The Practical Move

This week, take the three AI initiatives most likely to become operationally important and add a one-page physical-dependency appendix to each business case.

List the expected compute pattern, vendor capacity assumption, cost-volatility exposure, region constraints, fallback mode, and the person who revisits the assumption every quarter.

That exercise will feel premature for small pilots. Treat it as a test of seriousness. An initiative too small to justify an infrastructure assumption is probably too small to carry a transformation story. An initiative big enough to matter is big enough to expose the company to the data-center veto.

AI strategy used to sound weightless. Substations, cooling systems, construction schedules, utility tariffs, and voters now sit under it.

Companies that see that first will write better AI roadmaps. Others may discover that their most important AI decision was made somewhere else.