Hyper-Local: Bringing Forecasts into the Context of Your Deal
- John Harrington
- 2 days ago
- 2 min read
At a specific bus in southern Dallas County, the ERCOT 2025 planning model shows approximately 915.8 MW of Available Transfer Capability for generation injection — and 0 MW for additional load. LandGate's project database maps a 550 MW planned hyperscale data center development to that same node. Half a gigawatt of demand is queued for a location the planning model shows cannot currently take it. The regional ERCOT local forecast shows steady demand growth across North Texas while not acknowledging the complete lack of capacity in many parts of the grid. That gap is what this paper is about.
Forecasts are built to provide directional guidance across a wide area. The relevance to any single asset decays as the survey area widens, and the decisions that determine whether a deal performs (where to site, what to pay, whether to lend) are made at one substation, on one node. This paper makes three arguments:
A zonal forecast applied uniformly assigns most individual nodes almost nothing. In the Oklahoma example, a 5.12% zonal growth rate implies roughly 7.7 MW of two-year growth at a substation. A single data center interconnecting at that substation delivers 150 MW, nearly twenty times the forecast expectation, and close to half the entire metro's projected growth landing in one place. The forecast is accurate at zonal scale but missing context at the asset level.
The cost shows up in two distinct forms, and both are measurable in advance. Where interconnection is granted, the cost arrives as congestion: in the Oklahoma example, $1.6MM per year on a 20 MW industrial load, roughly a 27% increase on its energy spend. Where the network cannot accommodate the load, the cost arrives instead as denial, delay, or network upgrade obligations; the condition the Dallas County planning model documents. Project-level and grid-level data identify which path an asset travels before capital is committed.
Relief is also hyper-local, and equally below the radar to a regional view. A 250 MW solar farm interconnecting at the same constrained substation restores roughly 220 MW of midday headroom and cuts the congestion cost by about 57%. It does not lower the zone's demand growth by a single megawatt. Only node-level analysis distinguishes the two, and only node-level evidence of acute local need moves a relief project to the front of an interconnection queue where roughly two of every three projects withdraw before reaching operation.
The combination of Wood Mackenzie's forecasting and LandGate's project, substation, and nodal data closes the distance between a directionally correct outlook and the specific site a decision rests on. The forecast is not wrong. It is incomplete without context.
