America is preparing for an electricity demand shock. But one of the most consequential constraints on how quickly the grid can respond is barely discussed: the time it takes utilities to make the procurement decisions required to turn planned infrastructure into operating infrastructure.

NERC’s 2025 Long-Term Reliability Assessment makes the urgency difficult to ignore. Summer peak demand is forecast to rise by 224 GW over the next decade—24% above 2025 peak demand and roughly 70% higher than the growth projected just one year earlier. New data centers are the primary source of that growth. At the same time, NERC warns that uncertainty and lag in the pace of new resource additions are creating heightened reliability concerns. Its own summary identifies supply-chain constraints, interconnection processes and approval delays as threats to infrastructure growth.

This changes the procurement conversation. Procurement-cycle delay is no longer simply an administrative inefficiency. When infrastructure required for reliability cannot be sourced, evaluated, contracted and mobilized at the speed at which AI compute demand is exploding, procurement becomes part of the risk equation.

The Grid Has a Decision-Capacity Problem

There is physical capacity. There is manufacturing capacity. There is workforce capacity. And there is decision-making capacity—the easiest constraint to overlook because it does not appear on a one-line diagram.

We are spending enormous energy—figuratively and literally—debating where the next wave of electricity will come from. AI and data centers are driving load growth at a scale most utilities have not planned for in decades, stacked on top of aging infrastructure, electrification, grid hardening, constrained equipment and interconnection backlogs. Almost all of the conversation focuses on physical supply: more generation, more transmission, more substations and more transformers.

But before a single megawatt is built, thousands of decisions have to happen.

Engineering alternatives are weighed. Equipment is sourced. Contractors are vetted. Technical proposals are reviewed. Pricing is normalized. Safety qualifications are verified. Commercial exceptions are understood. Risk is evaluated. And every consequential decision must be documented well enough to withstand internal governance, regulatory scrutiny and, potentially, years of hindsight.

A major procurement can generate tens of thousands of pages across proposals, spreadsheets, schedules, certifications and commercial terms. Much of that information is reviewed by the organization’s most experienced engineers, project managers, procurement professionals and subject-matter experts—the same scarce people the industry needs focused on engineering judgment and execution.

There is nothing wrong with rigor. These decisions shape safety, reliability, customer costs and infrastructure that may operate for decades. The problem is that we routinely use scarce human expertise for two fundamentally different jobs: processing information and exercising judgment. They are not the same.

That distinction matters now because the grid’s physical capacity is not the only capacity under strain. We also have a decision-capacity constraint.

The Reliability Consequence of Delay

Consider the mismatch NERC is describing. Demand is accelerating faster than previously forecast while the infrastructure intended to serve that demand faces delays in approvals, interconnection, supply chains and project delivery. NERC’s 2025 assessment shows only 71 GW of approved summer resource additions over ten years against 224 GW of projected summer demand growth, with another 266 GW categorized as prospective rather than approved.

That does not mean procurement alone causes the gap. It clearly does not. Permitting, interconnection queues, manufacturing capacity, siting, financing, workforce constraints and regulatory approvals are all major contributors.

But procurement sits directly inside the critical path of many projects. A transformer cannot be installed until it is sourced. A transmission contractor cannot mobilize until it is selected and contracted. Engineering services cannot begin until qualifications, technical approaches, commercial terms and pricing have been evaluated. When those decisions take months longer than necessary, the schedule impact compounds across a capital portfolio.

A sourcing decision that takes months longer than necessary can add costs beyond procurement labor and may contribute to project-cost increases that ultimately affect customers. It can consume construction season, defer energization, expose a project to escalation and delay the moment new capacity becomes available to the system.

That is why procurement-cycle time deserves to be measured as an infrastructure-delivery and reliability variable—not merely as a procurement KPI.

AI Should Increase Labor Leverage, Not Replace Judgment

The instinct in many AI conversations is to ask what work can be automated away. For utilities, that is the wrong framing. This industry does not have a surplus of experienced engineers, project managers or procurement professionals waiting to be displaced. It has a shortage of them.

The opportunity is labor leverage.

Nor is the answer to hand a general-purpose AI assistant to a 20-year transmission engineer and expect that person to become an agent builder on the side. General-purpose models are powerful, but utility procurement is a regulated, evidence-intensive environment. A model given incomplete context or trusted without traceability can confidently misread a requirement, miss an exception or compare pricing on an invalid assumption. In a casual workflow, that is an inconvenience. In a multibillion-dollar capital program, the error may surface after award, when it is dramatically more expensive to correct.

The appropriate architecture keeps humans accountable for judgment while machines perform the information-processing work around them: structuring solicitation requirements, mapping vendor responses back to scope, identifying exceptions, normalizing commercial and pricing information, surfacing risk, preserving source evidence and creating a traceable record of how an evaluation reached its conclusions.

Speed Without Defensibility Is Not the Answer

AI-enabled procurement tools can help utilities structure solicitation requirements, evaluate vendor submissions against technical and commercial criteria, identify exceptions and preserve a traceable record of evaluation decisions. These capabilities could reduce the time spent processing large volumes of documentation while allowing engineers and procurement professionals to retain responsibility for technical judgment. For utilities, the value of these tools should be measured not only in time saved, but also in the quality, transparency and defensibility of procurement decisions.

There is an important trap here. Faster procurement by itself is not the objective.

That means measuring more than hours saved. A serious procurement-modernization program should ask: How long does it take to move from an approved scope to a complete solicitation? How long from bid close to a defensible evaluation? How much senior engineering time is consumed processing documents rather than exercising technical judgment? How consistently are vendor answers mapped to the actual scope and evaluation criteria? Can every material finding be traced back to source evidence? And can the organization reuse intelligence from prior procurements rather than beginning from zero?

These are measurable operating variables. Increasingly, they are also reliability variables.

A New Reliability Question for NERC and the Industry

NERC’s role is not to prescribe a particular commercial technology, nor should utilities be required to purchase a named AI platform. The more important question is technology-neutral: Should avoidable procurement-cycle delay on reliability-critical infrastructure be formally measured and managed as a grid reliability risk?

NERC already evaluates whether generation, transmission and other resources will arrive in time to serve forecast demand. The next step should be to understand more precisely where infrastructure schedules are being lost between project identification and construction: permitting, interconnection, equipment availability, internal approvals, solicitation development, bid evaluation, contracting and mobilization.

Many of these tasks may be suitable for AI assistance, provided that human judgment and accountability remain central.

The industry would benefit from a common framework for measuring these delays and from case studies demonstrating where modern procurement systems can materially compress cycle time without compromising governance or regulatory defensibility.

The hypothesis is testable. Take real utility procurement events. Establish the conventional baseline. Measure elapsed time, expert hours, evaluation quality, traceability and exceptions identified. Run the same work through an AI-enabled, human-in-the-loop process. Then publish the results.

If the evidence shows that weeks or months can be removed from critical procurement cycles while producing a stronger audit trail, the implication is larger than procurement productivity. It means the industry has found a way to increase infrastructure-delivery capacity without waiting years to manufacture another transformer, train another engineer or build another transmission line.

We Need to Widen the Definition of Grid Capacity

We cannot manufacture decades of engineering experience overnight. We cannot expand teams every time capital workloads spike. And we should not weaken the governance that protects reliability and customers merely to move faster.

What we can do is give the people already making these decisions dramatically more leverage.

AI is creating one of the largest new sources of electricity demand in modern history. It can also help the industry build the infrastructure required to serve that demand. But that will require us to stop treating procurement as back-office administration and recognize a harder truth:

When the grid cannot make and execute critical infrastructure decisions fast enough to keep pace with demand, procurement-cycle delay becomes a major grid reliability risk.