The Longest Lead Time Isn't Hardware: Reading Power Queues Before You Commit to a Site
Photo: Hirho, CC BY-SA 4.0, via Wikimedia Commons
For twenty years, the critical path to new capacity ran through procurement. Servers had lead times, networks had lead times, and the discipline of infrastructure planning was largely the discipline of ordering early enough. That assumption has inverted. In most US markets today, the longest item on the schedule is not a rack, a switch, or a storage array. It is an energized service entrance.
Boards have not yet fully internalized the change. Capacity plans still tend to be built upward from workload forecasts and downward from capital budgets, with power treated as a facility detail to be settled during construction. In 2026, power is the constraint that sets the calendar, and the calendar sets what the business can actually do.
What a queue position really is
A utility interconnection request is not an order. It is a place in a line, and the line is governed by study processes that consume years. A typical sequence runs through a feasibility or system impact study, then a facilities study, each requiring a deposit and each capable of producing restudy obligations when other requests ahead of yours change the model. Deposits have grown materially in several regions as grid operators attempt to separate speculative requests from funded projects, and withdrawal after a certain stage can carry financial penalties.
The practical consequence is that a queue position behaves like an option. It has value, it has carrying cost, and it decays if not exercised. Enterprises that understand this request capacity earlier than they can justify it on workload grounds, accepting the deposit as the price of optionality. Enterprises that treat interconnection as a downstream procurement step discover that the decision they wanted to make in 2026 was effectively made when the request was or was not filed.
Assignability deserves early legal attention. If a site is sold, a lease is assigned, or the entity holding the request is restructured, the ability to transfer the queue position determines whether the option survives the transaction.
The regions are not one market
Planning built on a national average will mislead. The organized markets differ in process, timeline, and exposure. PJM's large-load addition procedures, ERCOT's interconnection rules for very large loads, MISO's generator and load queue reform work, SPP's and CAISO's study backlogs, and the vertically integrated Southeast, where state commissions increasingly scrutinize large-load tariffs, each produce a different answer to the same question. Some jurisdictions now require large loads to bring their own generation or to fund firm capacity before service is granted. Others allow a phased energization that materially improves the schedule.
Site selection should therefore begin with a power-mapping exercise rather than a real estate shortlist: which substations have headroom, which utilities have demonstrated throughput on large-load requests, and which regulatory environments have shown a willingness to approve the kind of load profile the business actually runs.
Behind-the-meter is a bridge, not a bypass
On-site generation has moved from a sustainability gesture to a scheduling instrument. Reciprocating engines, aeroderivative turbines, fuel cells, and battery storage can energize a facility years before a utility service would. That capability is real, and it is routinely oversold.
The constraints arrive in a different order than most plans assume. Air permits and emissions limits often bind before the equipment does. Fuel supply, whether gas pipeline capacity or on-site storage, introduces a second utility relationship with its own queue. Maintenance contracts for prime-power operation are a different product from standby service, with different cost structures. And a behind-the-meter plant that runs continuously may change the site's regulatory classification and its relationship with the local utility in ways that were not anticipated during financial modeling.
Treat on-site generation as a bridge with a defined amortization period and a defined exit, not as a permanent answer to a grid problem you no longer wish to think about.
Cloud placement is a power decision too
Hyperscale providers absorb interconnection risk at a scale no individual enterprise can replicate, which is a legitimate reason to place elastic workloads with them. But the absorption is not unlimited and not uniform. Availability zone capacity in constrained regions is finite, commitments are increasingly structured around reserved capacity rather than on-demand elasticity, and providers signal constraint through pricing, quota, and contract terms long before they signal it in a press release.
A portfolio that assumes unlimited regional elasticity is, in effect, assuming the provider solved a problem the enterprise declined to model. Capacity reviews should test that assumption explicitly: which workloads depend on a single constrained region, what reserved capacity exists in writing, and how quickly a failover region could absorb the load if it had to.
Questions for the next capacity review
- Which of our assumed 2028 capacity depends on an interconnection request we have not filed?
- What is the current queue position in each relevant market, and what deposits and restudy obligations accompany it?
- Can the position be assigned, and does our lease or entity structure preserve that right?
- What is the energization date we would bet capital on, as distinct from the date in the utility's acknowledgment letter?
- Which workloads could move to a provider region with genuine headroom, and which cannot move at all?
- If we bridge with on-site generation, what is the permit path, and what is the amortization period?
None of these questions is comfortable, and none is answered by the facilities group alone. Power has become an enterprise strategy problem because it now determines the outer boundary of what the business can commit to. The organizations that plan around that boundary will look conservative in the boardroom for about eighteen months, and prescient for considerably longer.