Executive Monday Insights
AI power demand is becoming a capacity allocation problem.
The visible issue sits in the electricity system. Data centers need power. Utilities need generation, transmission, interconnection capacity, capital approvals, and regulatory permission. Customers want certainty. Communities want lower bills and reliable service.
The deeper issue is organizational.
When power becomes scarce, companies need to decide which growth commitments deserve capacity, which customer promises can be made, which sites should be prioritized, which infrastructure investments are worth accelerating, and which risks should be absorbed, delayed, priced, or rejected.
Those decisions rarely sit in one place.
Reuters reported on May 16, 2026, citing the Financial Times, that NextEra Energy and Dominion Energy were in talks over a potential tie-up that could create a utility valued at roughly $400 billion including debt. The strategic logic is not hard to see. Dominion serves Virginia, one of the most important data-center regions in the United States, while NextEra has positioned itself around large-scale power growth. Reuters also noted that U.S. power consumption reached a second consecutive record high in 2025 and is expected to rise further as AI data centers add demand.
Four days earlier, Reuters reported that the U.S. Energy Information Administration expected U.S. power demand to rise from 4,195 billion kilowatt-hours in 2025 to 4,248 billion in 2026 and 4,379 billion in 2027. The International Energy Agency has pointed to the same structural pressure globally: data-center electricity consumption is projected to double by 2030, reaching around 945 terawatt-hours in its base case.
The numbers matter. But for executives, the more useful question is not whether demand is rising. It is how the organization decides what to do when demand outruns the capacity available to serve it.
The capacity signal arrives before the decision model
Power constraints do not create a single decision. They create a queue of competing claims.
A hyperscale customer wants a firm commitment. A manufacturing site wants a faster grid connection. A regulator wants evidence that residential customers will not carry disproportionate cost. Finance wants return discipline. Operations wants reliability. Public affairs wants a defensible community position. Commercial teams want to secure the growth account before a competitor does.
Each claim can be legitimate.
The problem begins when the organization treats capacity as an input to be sourced rather than a constraint to be governed.
In many companies, energy is handled through procurement, real estate, sustainability, infrastructure, or operations. Growth commitments sit with commercial leadership. Capital decisions sit elsewhere. Risk sits partly with legal, partly with finance, partly with public affairs. Customer exceptions are often negotiated near the account, while the consequences appear later in network planning, cost allocation, service reliability, or margin.
The organization can be busy and still not converge.
One group negotiates the demand. Another validates feasibility. A third models cost. A fourth manages regulatory exposure. A fifth explains the decision externally. The leadership team sees activity, but the capacity decision itself remains distributed across functions that are not accountable for the same outcome.
That creates decision distance.
Under moderate demand, the distance may be hidden. Under AI-driven electricity demand, it becomes expensive. Commitments are made before capacity is secured. Capacity is reserved before the economic logic is clear. Capital projects advance before customer value is tested. Community and regulatory risks are discovered late. Senior leaders are then asked to reconcile promises the operating model allowed different parts of the company to make separately.
Scarcity changes the economics of authority
When capacity is abundant, many decisions can be made locally. Sales can pursue demand. Infrastructure can respond. Finance can review investment cases. Operations can adjust sequencing. The system may be inefficient, but the inefficiency is tolerable.
Scarcity changes that.
A megawatt allocated to one customer, site, region, or product roadmap is no longer neutral. It is a strategic commitment. It carries opportunity cost, capital cost, regulatory exposure, delivery risk, and reputational risk. It can also affect existing customers who never participated in the decision.
That is why AI power demand creates a leadership design issue.
If authority stays too local, teams may optimize for account growth, site speed, or project approval while weakening enterprise economics. If authority moves too far upward, the organization gains formal control but slows the decisions that require local facts. Leaders then become the queue managers for every contested capacity claim.
Neither pattern works well for long.
The better question is where bounded authority should sit. Central leadership needs to define the economic and risk constraints: capital thresholds, customer prioritization logic, reliability requirements, cost-recovery principles, regulatory boundaries, and community commitments. Outcome-owning teams then need enough authority to make integrated decisions inside those constraints.
That distinction matters.
A team responsible for a data-center region cannot wait for separate approvals from commercial, grid planning, finance, legal, sustainability, and public affairs every time a capacity decision changes. But it also cannot be allowed to promise capacity as if the only objective were revenue growth.
The decision needs to close where the facts are understood and where the trade-offs are visible.
The structural mistake
The common mistake is to respond to capacity scarcity by adding governance.
More steering committees. More approval gates. More executive reviews. More investment cases. More scenario decks.
Some of that may be necessary. Scarce infrastructure should not be allocated casually. But additional governance often increases visibility without improving decision quality. It tells leaders where the pressure is. It does not necessarily give the organization a better mechanism for resolving the pressure.
Capacity allocation requires convergence across several questions:
- Which demand is strategically worth serving?
- Which commitments can be made without weakening reliability or cost discipline?
- Who can reserve capacity, and under which conditions?
- What must be true before commercial teams can promise delivery?
- When should the company refuse demand, delay demand, or price scarcity explicitly?
- Who owns the final economic consequence if the decision proves wrong?
If these questions are answered in different parts of the organization, the company does not have a capacity allocation model. It has capacity exposure reporting.
Reporting is useful. It is not enough.
What high-performing organizations will do differently
The companies that handle AI power demand well will treat capacity as a managed strategic constraint.
They will not allow the demand signal to move faster than the decision rights attached to it.
Commercial teams will know which customers qualify for scarce capacity and which do not. Infrastructure teams will know which projects are economically and strategically protected. Finance will be involved early enough to shape choices, not late enough only to challenge them. Regulatory and public-affairs input will be part of the capacity decision before the company has created commitments it cannot defend.
This is not collaboration as a slogan. It is a different operating rhythm.
A useful starting point is to map the path from demand signal to capacity commitment. Where does the request first appear? Who validates the load requirement? Who tests the economics? Who checks grid feasibility? Who evaluates regulatory and community exposure? Who can say yes? Who can say no? Who can override the answer? Who carries the cost if the commitment later proves uneconomic?
The map will usually show that the formal decision is later than leaders think.
It will also show where work is being repeated. Commercial teams build their view. Infrastructure teams build another. Finance builds a third. Legal and regulatory teams add constraints after the economic case has already gained momentum. By the time the decision reaches senior leadership, the organization is no longer deciding cleanly. It is unwinding partial commitments.
What to measure
Capacity allocation should be measured as an operating capability, not only as project delivery or capital spend.
Leaders need earlier indicators than utilization, backlog, or missed connection dates.
The first is time from demand signal to capacity decision. This measures how long the organization takes to move from customer or market demand to an approved, rejected, delayed, or conditional capacity commitment.
The second is decision distance. How many structural handoffs sit between demand identification and authority to commit capacity?
The third is reopened commitments. If capacity decisions are repeatedly revisited after finance, grid planning, regulatory, or customer facts arrive late, the operating model is integrating the right perspectives too slowly.
The fourth is exception rate. Scarce capacity invites exceptions. Some will be commercially justified. Too many exceptions indicate that the allocation rules are unclear or that leaders are avoiding the hard prioritization decision.
The fifth is realized economic impact. Did the capacity commitment produce the expected margin, strategic value, reliability outcome, or customer result? Or did the company simply win demand that became expensive to serve?
These measures will not remove scarcity. They will show whether scarcity is being governed deliberately or absorbed through delay, escalation, and rework.
The leadership question
AI power demand will keep creating pressure on utilities, data-center operators, manufacturers, regulators, and communities. The infrastructure question is real. More power has to be generated, transmitted, connected, financed, permitted, and defended.
But executives should not let the infrastructure challenge hide the operating-model challenge.
Scarce capacity forces choices. Choices require authority. Authority needs to sit close enough to the facts to be useful and close enough to enterprise economics to be disciplined.
The practical question is simple:
How does your organization decide who gets scarce capacity?
If the answer is a sequence of functional reviews, the decision is probably slower than the market signal. If the answer is “senior leadership decides,” the organization may have visibility without operating capability.
The better test is whether the organization can convert a demand signal into a capacity commitment with clear economics, clear constraints, and clear ownership.
AI may be creating the demand. Slow capacity allocation is designed internally.
Sources: Reuters reporting on NextEra/Dominion talks; Reuters reporting on U.S. EIA electricity-demand forecasts; International Energy Agency analysis on energy demand from AI and data centers.
If you want to get additional inspiration and support, then let's have a conversation.
To receive a new edition every week, we invite you to sign up to the Executive Monday Insights Newsletter
You can find other articled here.
Comments are closed.