AI Capex Absorption Score

Problem solved

Hyperscalers announce gigawatt-scale data center commitments faster than any single region's grid, water, talent pipeline, and permitting capacity can keep up. ACAS scores absorption probability for a named site against a four-axis stack (power, water, talent, permitting) and produces a delivery-risk classification with a gap-to-COD timeline and an explicit mitigation table.

Inputs

• Power available and queue position (utility filings, ERCOT, PJM, CAISO, MISO interconnection trackers)
• Water rights and aquifer stress (USGS, EPA WaterSense, regional water authority data, state engineer filings)
• Engineering talent depth (BLS QCEW occupational, IEEE membership density, regional university pipelines)
• Permitting throughput (state and county filings, prior data center COD-to-permit timelines)
• Transformer slot reservation and lead time (utility procurement, GE Vernova, Hitachi Energy, Siemens Energy disclosures)
• Transmission upgrade timeline (RTO planning queue, FERC Order 2023 cluster studies)
• Tax incentive package (state and county MOU, abatement, sales tax exemption)
• Community opposition signal (FERC and PUC docket comments, local press coverage, council vote tallies)

Outputs

• Delivery-risk class (low, medium, high) for the named site
• Individual axis scores (power, water, talent, permitting) on a 0 to 100 calibrated band
• Gap-to-COD timeline (months to commercial operation date) under base, stretch, and stress
• Mitigation ranking by axis with capex and timeline cost
• Scenario sensitivity to ERCOT-style large flexible load reform, FERC Order 2023 cluster timing, and water curtailment
• Behind-the-fence generation feasibility flag (gas peakers, on-site solar, SMR optionality)
• Replication package with named source links and rebuild scripts

Method

1. Step 1. Lock the named site and its rated load. Resolve net vs. gross MW, ramp profile, and PUE assumption.
2. Step 2. Pull power axis: queue position, transformer slot reservation, transmission upgrade timeline, large-load tariff exposure.
3. Step 3. Pull water axis: water rights, aquifer status, cooling architecture (closed-loop, evaporative, dry), reclaimed-water availability.
4. Step 4. Pull talent axis: engineering depth (electrical, mechanical, controls, fiber, SCADA), regional commute geography, hyperscaler talent saturation.
5. Step 5. Pull permitting axis: state and county throughput, prior data center timelines, environmental review tier, community-opposition signal.
6. Step 6. Score on a 0 to 100 calibrated band per axis, then combine with explicit weights (default 35 power, 25 water, 20 talent, 20 permitting).
7. Step 7. Run interconnection queue stress and mitigation table; output replication package with citations.

Assumptions

• ERCOT, PJM, CAISO, MISO interconnection queue data is current as of the run date.
• Water rights are taken at face value from state engineer filings; pending litigation is flagged but not adjudicated.
• Talent pipeline is regional QCEW; cross-state recruitment is not modeled as a base-case mitigation.
• Default axis weighting is 35 power, 25 water, 20 talent, 20 permitting; users can re-run at custom weights.
• Behind-the-fence generation is treated as a mitigation, not a base-case input.

Limitations

• The framework does not model behind-the-fence generation in the base case; it appears only in mitigation.
• Community opposition is a signal, not a probability; the framework reports the signal but cannot resolve the political outcome.
• The four-axis weighting is debatable. CAISO and ERCOT users typically reweight power higher; arid-region users typically reweight water higher.
• Transformer lead-time data is procurement-confidential in many cases; the framework uses public OEM disclosures as a floor.
• FERC Order 2023 cluster-study timing is RTO-specific and continues to evolve.

Example application

Applied to a hyperscaler siting decision in ERCOT 2026: a 1.2 GW campus near Abilene scores high power axis (queue position favorable, transformer slot reserved), medium water axis (Edwards-Trinity aquifer stress flagged), low talent axis (regional engineering depth thin without commute reach to Dallas-Fort Worth), and medium permitting axis (county throughput recent but environmental review tier escalates with cooling architecture). The composite delivery-risk class is medium, with talent and water as the binding axes and a 14 month gap-to-COD under base case. See Texas ERCOT AI siting in 2026: power, water, and the absorption ceiling.