An AI data-center project can have financing secured and GPUs on order while its electricity, cooling and network dependencies aren’t yet ready to work together. A signed capacity agreement cannot establish when that site will support the intended workload. Infrastructure leaders need to distinguish commercial commitment from physical readiness before either becomes the basis for further spending.
The Site Must Support the Promised Operating Load
For Neil P. Osnato, founder of Persistence Analytics Group LLC, the evidence behind an infrastructure decision must support the specific claim being made. He draws a practical distinction: a power purchase agreement does not establish that electricity will be available when commissioning requires it. Equipment orders likewise do not demonstrate that a campus can operate the proposed cluster at its intended duty cycle. Procurement, construction and operational readiness each need their own verification.
The pressure sits at the interfaces. Electricity, cooling, networking and equipment must become usable in a configuration that supports the customer’s expected workload. Evidence that one dependency has advanced should not be allowed to stand in for evidence that the whole facility is ready.
For leadership, the useful question is: ready for what? A campus approaching completion and a campus capable of sustaining the planned operating load represent different investment propositions. Approval records should make clear which proposition the available evidence supports, rather than allowing a broad label such as “capacity secured” to cover both.
Infrastructure Commitments Run On Different Clocks
Osnato points to a mismatch between the duration of compute agreements, financing, equipment cycles and physical infrastructure. Servers may pass through technology changes while buildings, electrical systems and transmission investments remain in service. Enterprise customers, meanwhile, are still determining their usage needs and willingness to pay.
That mismatch creates an architectural concern as well as a financial one. A long-lived facility may be justified using assumptions about equipment and workloads that change much sooner. Leaders should therefore examine which parts of the investment case depend on the initial configuration remaining economically useful.
A customer contract provides evidence of commitment. It does not establish that the customer’s applications will justify the same volume of compute throughout the agreement, or that renewal terms will remain comparable. Osnato’s argument is to evaluate the agreement’s legal strength separately from the durability of its usage and revenue assumptions.
The distinction also changes how resilience should be assessed. A project may retain enforceable payment rights after its operating assumptions weaken. That protection deserves credit, but it leaves a separate question about the continued usefulness of the installed assets.
Keep Load Claims Separate Through Each Decision Gate
Osnato describes a progression from a customer’s requested load through supporting evidence, physical availability, energization, actual use and demand that persists. Each stage requires its own records. If those stages are treated as interchangeable, a forecast can take on the authority of an operating fact as it moves between engineering, finance and leadership teams.
A practical review can organize those distinctions without creating another broad assurance label:
| Claim under review | Evidence the decision needs |
|---|---|
| Customer capacity commitment | Contract terms, expected usage and the customer’s ability to meet its obligations |
| Physical readiness | Support for electricity, cooling, network and equipment availability |
| Operating demand | Evidence of actual use, distinct from reserved capacity |
| Continued asset value | Support for sustained demand and the assumptions behind renewal |
| Changed project conditions | A reassessment before the next capital commitment |
Grid planning faces a related coordination problem. Persistence Analytics Group has provided comments to MISO’s Large Load Working Group concerning forecasts, application procedures, studies and cost visibility. Osnato identifies a recurring concern: earlier planning decisions can lose alignment with a project when its proposed load, schedule or other substantive conditions change.
For infrastructure teams, that makes the handoff between studies and deployment a governance issue. The original evidence may have been adequate for the original plan. A revised plan needs an explicit assessment of whether that evidence still applies.
Revalidation Should Govern the Next Commitment
Osnato uses hypothetical changes such as a slower customer ramp, later power availability or materially different equipment to explain the need for reassessment. Each change might be manageable, yet still undermine an assumption used to authorize construction or financing. The control belongs before the next commitment.
Contractual protections remain useful within that process. Guarantees, collateral and termination terms can distribute or reduce exposure. Osnato cautions, though, that recovery still depends on their terms, enforceability and the resources behind them. Several agreements may rely on the same customers, providers or funding relationships, so the diversification they provide may be limited.
Infrastructure leadership should connect the engineering record to the capital decision. Engineers establish technical feasibility, legal teams examine rights, and finance teams assess cash flows. The approval process must then determine whether those findings, taken together, support the proposed investment. If power timing, equipment configuration or customer demand changes, the next authorization should make clear which assumptions are affected, what evidence is still missing and who accepts the remaining exposure.
