Available capacity is a planning or map signal. Deliverable capacity is power the utility can serve to a specific parcel on a dated path, with known works, cost ownership, and conditions. Confusing the two is one of the most expensive mistakes in data center site selection.
Developers should treat available capacity as a reason to investigate, not a reason to commit. The commitment question is narrower: can this load be energized here, by this date, under terms the capital stack can survive?
Why the distinction matters now
Large-load demand is rising faster than many grids can absorb on historic timelines. Public planning reports and market outlooks have been clear that data center load creates forecasting and delivery pressure that older industrial underwriting did not face.
In that environment, a site can look power-rich on a slide and still fail when the utility path, upgrade scope, or energization date is tested. The underwriting job is to convert a capacity story into a delivery case.
Definitions
Available capacity
A signal that some amount of electrical capacity exists in a territory, corridor, substation area, or planning study. It may be seasonal, contingent, reserved, shared, or dependent on future upgrades. It is often presented without parcel-level service terms.
Deliverable capacity
Capacity that can be served to the project parcel for a defined load profile, under identified technical and commercial conditions, with a credible path to energization. Deliverable capacity names the utility, the delivery point logic, the works required, who pays, what remains conditional, and when first useful power can arrive.
If those elements are missing, you still have availability language. You do not yet have a deliverable power case.
Side-by-side comparison
| Dimension | Available capacity | Deliverable capacity |
|---|---|---|
| What it answers | Is there a power story nearby? | Can this parcel get firm service on our timeline? |
| Typical evidence | Maps, IRP excerpts, broker claims, early utility notes | Studies, written feasibility, service terms, construction programme |
| Parcel link | Often regional or substation-adjacent | Tied to the site and load case |
| Timing | Vague or optimistic | Dated milestones with dependencies |
| Cost ownership | Often unclear | Explicit or explicitly unknown and flagged |
| Decision use | Screen in or screen out | Pursue, monitor, renegotiate, or kill |
Evidence levels for power claims
Rank every power claim before you spend on control or deep diligence.
Marketing claim: nearby lines, campus narrative, "power available" language
Desktop infrastructure read: substations, voltage, territory, public planning context
Utility conversation: useful, still low confidence until written
Written feasibility or enquiry response: better, still conditional
System or facilities study: stronger technical basis, check assumptions and age
Executed arrangements: commercial path with obligations and remedies
Delivery programme and energization evidence: the controlling schedule becomes visible
Move a site forward only when the evidence level matches the next dollar of risk. An option deposit may tolerate a lower rung than a non-refundable deposit or construction financing assumption.
How to convert availability into a deliverable case
Run the same sequence on every candidate:
Fix the load profile: day-one MW, stabilized MW, ramp, redundancy, cooling, and backup plan.
Identify the serving utility and likely delivery path for that parcel.
Separate capacity that exists today from capacity that requires upgrades.
List required works, lead times, approvals, and cost allocation.
Write the energization chain as dependencies, not a single marketing date.
Record unresolved conditions in one ledger the IC can read.
A practical one-line record looks like this:
Parcel → utility → requested MW → deliverable MW claim → works → cost owner → target date → evidence date → open conditions
If the team cannot complete the line, the site remains in available-capacity territory.
Pursue, monitor, or kill
Pursue when deliverable evidence supports the load and date closely enough for the next control step, and residual gaps are named.
Monitor when the path is plausible but depends on upgrades, queue movement, tariff clarity, or political process that is not yet bankable.
Kill when the site depends on unowned upgrades, undated energization, reserved capacity that is not transferable, or conditions that break the commercial case.
Do not keep "maybe power" sites in the live pursuit column without an owner and a revisit trigger. Soft power language becomes soft pipeline.
What AI can and cannot do
AI can gather filings, extract queue and planning tables, compare sites, maintain the evidence ledger, and flag when a study is stale or a condition changed.
AI cannot create deliverable capacity. It cannot promise that a utility will serve the load, that an upgrade will clear, or that a contingent date will hold. Experts still place the utility calls and own the risk call.
Decision rule
If the power case only proves that capacity exists somewhere nearby, keep screening. If the power case proves how this parcel gets served, when, at what works and cost conditions, you may be looking at deliverable capacity.
Site selection should advance on deliverable capacity, then validate the rest of the site and bundle the work required to close.