Years ago, I ordered a shipment of concrete for a project. It was certified, paid for and aboard a ship headed to port. The supplier was a large company, and my order was part of an even larger project. I believed that connection gave the delivery additional security. Other people had more riding on that shipment than I did.
The broker called me three times as the situation changed.
The first call was good news. The shipment was in transit, on time and expected in port the next day.
On the second call, he asked whether I would be willing to sell my order and accept replacement material a few days later. I said no. My staff and outside contractors were already coordinated around the promised delivery date. He understood and said he would let the interested buyer know.
On the third call, he apologized. The ship's captain had contacted him to report that the other buyer had purchased the entire vessel and its cargo. The ship was being diverted. My concrete was still certified. It was still paid for. It was still physically aboard a ship. It simply was not coming to me.
The broker had done nothing wrong. He kept me informed as the facts changed. Whatever remedy the contract may have provided later, it could not put the concrete on my jobsite when the work was scheduled.
A promise and a physical fact are not the same thing
That experience left me with a lesson that applies directly to the present infrastructure buildout: operationally, a delivery should not be treated as a fact until you control it. Even physical possession is not proof that it will perform the function for which it was purchased.
Commissioning establishes the next indispensable truth: whether the assembled system performs under the conditions it was designed to withstand. A transformer on a pad, a cooling system in a building or a rack full of equipment may be physically present. Until the complete system has been tested under operating conditions, the need has not yet been satisfied.
A contractual promise is not the same as physical control.
This distinction becomes critical when many well-financed projects are pursuing the same constrained equipment, utility capacity, skilled labor and delivery routes. A purchase order can reserve a place in line. It cannot guarantee that the line itself will remain intact.
Every project may be legitimate
The AI infrastructure buildout is often discussed as though the central problem is deciding whether an individual project is real. That is only part of the question. A developer may have land, financing, a customer, a sound design and credible contractors. Its demand may be legitimate. Its competitors may be able to say the same thing.
The weakness appears when those individually reasonable plans are added together. Several projects may be counting on the same future megawatt, transformer manufacturing slot, turbine, transmission improvement, water source, fiber route or specialized workforce. Every claim may be serious. The shared system may still be unable to satisfy them all within the promised time.
The constraint is not simply how many megawatts exist. It is whether the required power is deliverable at the necessary place and time, under the expected load profile and contingency. Firm demand, phased demand and demand that can be interrupted do not place the same claim on the system.
That is not necessarily fraud, and it does not require anyone to be foolish. It is a coordination failure. Each participant can make a rational decision from inside its own boundary while the combined set of decisions becomes physically impossible.
The most important question is therefore not whether each project has a credible plan. It is whether all the credible plans can be true at the same time.
The same test must be applied to demand. A customer's expression of interest is not revenue, just as an equipment order is not operating capacity. Before capital becomes difficult to reverse, the project must establish what demand is contracted, who carries the credit, what conditions allow the customer to withdraw and whether the revenue clock can survive the delivery clock.
Braking is not the same as planning
States and grid operators are beginning to respond, but they are responding from different starting points.
Texas has a more self-contained grid planning boundary. That gives ERCOT a clearer view of large load requests within its footprint, although relative isolation can also limit outside support. ERCOT is implementing its Batch Zero process to assess eligible projects together and distinguish more mature projects from speculative requests. That is a useful form of braking. It forces the system to ask what is real before treating every requested megawatt as equally ready.
Virginia sits inside a much more interdependent system. Local governments control much of the land use process, utilities serve the load, state regulators examine major approvals and cost recovery, and PJM coordinates multistate transmission planning, wholesale markets and reliability. Those parties can slow projects, examine costs and reconsider how new demand is served. No one of them can reconcile every dependency within Virginia's borders.
Both forms of braking may be necessary. Neither, by itself, is a complete plan. Slowing projects can protect a system from immediate overload. It does not determine which projects can be delivered as complete operating systems, how shared resources are being counted or what happens when a promised dependency fails elsewhere.
Self-preservation is a rational first response. Forward planning begins when the system can see beyond the immediate constraint and understand the entire chain of claims.
A regional plan does not control a global supply chain
It is possible to imagine a North American system with abundant land, generation, fuel, water, transmission corridors and construction capability. Canada, the United States and Mexico together possess enormous physical resources.
That still would not make the system independent.
The equipment that converts those resources into operating capacity depends on supply chains that extend around the world. Transformers, switchgear, power electronics, semiconductors, cooling components, controls and specialized parts may depend on manufacturers, minerals, ports and shipping routes outside North America. Even domestically assembled equipment can be stopped by one obscure component produced elsewhere.
A project may control its site and financing while remaining exposed to a vessel being rerouted, a factory being allocated to a larger buyer, a trade restriction, a war-risk insurance decision or a component that never arrives.
Independence is therefore the wrong absolute. Resilience is the more useful objective: knowing which dependencies cannot be eliminated, creating alternatives where possible and making the remaining exposure visible before commitments become irreversible.
The missing function
Pieces of this function already exist within utilities, grid operators, regulators, local governments, lenders and private contracts. What is missing is an integrated view capable of seeing when multiple legitimate claims are being made against resources that cannot satisfy them all at once, and testing whether those separate assumptions can all be true together.
That function should not decide which private company deserves to win. Competition is real, and it matters. If one developer can secure equipment, execute well and bring a viable project online while another cannot, the government should not erase that distinction in the name of coordination.
Any allocation of scarce shared capacity will affect competitive outcomes. The legitimate role is not to favor one company. It is to apply transparent, neutral standards for readiness, cost responsibility and access to constrained shared resources.
The boundary lies between commercial risk and shared-system risk. Companies should be free to compete with their own capital, their own contracts and their own execution. They should not be free to make competing bets against the same constrained shared capacity, the same transmission assumption or a dependency whose failure will be paid for by customers and communities that never chose the bet.
Companies should be free to compete with their own money. They should not be free to make competing bets against the same constrained shared capacity.
The missing function must be able to:
- See material claims against shared resources.
- Distinguish mature demand from speculative reservations.
- Distinguish firm, phased and interruptible demand.
- Identify where several projects depend on the same capacity.
- Test whether delivery schedules can coexist physically.
- Expose dependencies that sit outside regional control.
- Require alternatives before a single failure becomes fatal.
- Allocate each risk to the party best able to control, mitigate or absorb it, and make any public exposure explicit.
- Update the picture as technology, demand and supply change.
This is not a proposal to direct private investment. It is shared resource underwriting. The purpose is not to choose a commercial winner. It is to prevent several private plans from quietly spending the same constrained shared capacity and leaving customers or communities with the loss.
It does not necessarily require one new institution or competitors sharing sensitive information with one another. It requires existing institutions to exchange enough verified, protected and appropriately aggregated information to see the complete exposure.
Seeing the complete chain
Infrastructure plans are usually presented in pieces. Land is discussed as real estate. Power is discussed as utility service. Equipment is discussed as procurement. Capital is discussed as financing. Operations are discussed as a later staffing matter.
The project succeeds only when those pieces become one complete operating chain.
That chain must extend past delivery. Equipment must arrive, be installed, communicate with the other systems, survive failure scenarios and be maintainable by the people who will operate it. Spare parts, service support, documentation and trained staff are not secondary considerations. They are part of deliverable operating capacity.
Optionality must also be real. A physical asset may have an alternative use, but only if that use remains technically, geographically and contractually possible. Recovering part of an investment by selling stranded equipment is not the return investors expected. It is damage control.
Preserving alternatives must therefore be a standard used during design and contracting, not a separate exercise performed after the original plan fails.
A better definition of deliverable capacity
Announced gigawatts are not delivered capacity. Interconnection requests are not delivered capacity. Capital commitments, land positions and equipment orders do not create deliverable operating capacity when viewed separately.
Deliverable operating capacity is the number of complete operating systems that can be delivered with every indispensable dependency accounted for.
This definition changes the conversation. It requires us to count the same manufacturing capacity, delivery window or shared electrical resource once, not several times through several project plans. It requires a credible path from generation to the rack, not simply a future service date. It requires confirmation that the system can be operated, maintained and repaired after the construction team leaves.
It also makes uncertainty visible. A project can move forward with unresolved questions, but those questions should be priced, assigned and paired with decision points. Uncertainty is not the problem. Treating uncertainty as certainty is.
Leadership begins when certainty ends
No institution will eliminate competition, geopolitical risk or technological change. No model will make every clock move at the same speed. The answer is not a predetermined plan that remains correct regardless of conditions.
The answer is a disciplined process that keeps testing physical truth, brings the necessary parties together and preserves the next decision before the current one removes it.
The government should not decide which competitor deserves to succeed. The market can make that decision.
The public system must ensure that private competition does not place the same shared resources at risk several times or leave the consequences of failure with people who never chose the bet.
The question is no longer whether each project has a credible plan.
It is whether all those plans can be true at the same time, and who pays when they are not.
⚓