{ "id": "texas_ai_water_stress", "version": "1.0", "status": "active", "scenario_type": "Compound Risk", "name": "Texas AI + Water Stress Compound Scenario", "subtitle": "Can Texas remain the AI capital of North America under simultaneous drought, thermal derating, and demand surge?", "region_id": "us", "tags": [ "power-sector", "compound-risk", "water-energy-nexus", "drought", "thermal-derating", "ai-demand", "grid-reliability", "ercot" ], "description": "ERCOT's mandate to absorb 15 GW of AI/data-center load by 2031 was designed for a single-vector stress: demand growth. It was not designed for compound drought + thermal derating + extended wind drought arriving simultaneously. Texas operates the most water-stressed large thermal fleet in North America \u2014 Comanche Peak nuclear draws from Squaw Creek Reservoir (NRC license condition below 40% capacity), gas CCGT units across the Brazos and Colorado River basins lose 7\u201312% rated capacity above 95\u00b0F ambient, and AI data centers require 1.8 million gallons/day per 100 MW of evaporative cooling. A P90 La Ni\u00f1a drought year reduces river basin inflows by 28\u201335%, simultaneously drying West Texas topsoil (suppressing turbulence and lowering wind capacity factors to 14\u201318%) and reducing cooling water availability for 52 GW of gas generation. The compound scenario \u2014 extreme heat wave, low wind, constrained cooling water, and 15 GW of AI demand drawing continuously \u2014 eliminates ERCOT's entire 15% reserve margin within 8 hours of sustained peak load. This scenario models the collision of four simultaneous stressors and quantifies the institutional choices that follow: forced AI curtailment, emergency interconnection to SPP, or rotating outages.", "baseline": { "year": 2026, "generation_fleet_gw": 117.0, "coal_gw": 4.0, "gas_ccgt_gw": 38.0, "gas_peakers_gw": 14.0, "nuclear_gw": 5.4, "wind_gw": 37.0, "solar_gw": 16.0, "bess_gw": 2.6, "grid_carbon_intensity_g_per_kwh": 340, "grid_carbon_intensity_note": "AUDIT FLAG (HIGH): Inherited from texas_ercot_ai_demand \u2014 340 g/kWh \u00d7 490 TWh = 166 Mt \u2260 stated 99 Mt. Correct total grid average intensity = ~202 g/kWh. See texas_ercot_ai_demand for detailed note.", "annual_generation_twh": 490, "annual_emissions_mt_co2": 99.0, "peak_demand_gw": 88.0, "water_withdrawal_bgd": 3.4, "water_consumption_bgd": 1.1, "cooling_water_rivers": [ "Brazos River", "Colorado River (TX)", "Squaw Creek Reservoir" ], "notes": "Baseline thermal fleet water withdrawal 3.4 billion gallons/day (once-through + recirculating cooling). Squaw Creek Reservoir at Comanche Peak: 3,270 acre-ft capacity; NRC COLA condition requires turbine runback above 95\u00b0F ambient or below 40% reservoir level. Brazos River basin hosts ~22 GW of gas CCGT; Colorado River basin (Texas) hosts ~8 GW. Wind CF baseline 35% annual average; drought years suppress to 15\u201318% during summer heat dome via reduced surface roughness and stable air mass." }, "target": { "reduction_pct": 2, "deadline_year": 2031, "horizon_years": 5, "required_reduction_mt_co2": 2.0, "ceiling_mt_co2_by_2031": 97.0, "reliability_target": "\u226515% planning reserve margin sustained through P90 summer compound stress event", "penalty": { "description": "LOLP >5% triggers NERC compliance review. ERCOT energy-only market scarcity pricing reaches $9,000/MWh cap during shortage \u2014 a single Uri-scale event (2021) inflicted $28B+ economic damage. AI hyperscaler SLA violations above 0.001% annual downtime trigger contract penalties of $200M+ per operator and potential workload migration out of Texas.", "mechanism": "NERC LOLP 5% threshold; PUCT SB6; hyperscaler co-location SLA enforcement" }, "notes": "This is a compound-risk scenario, not a pure mandate scenario. The '2% reduction' target is inherited from the parent ERCOT AI mandate. The scenario tests whether the mandate-case fleet (BESS + solar buildout from texas_ercot_ai_demand) is sufficient to maintain reliability under simultaneous compound physical stressors. A passing scenario requires P90 reliability even under compound drought + thermal derate + wind drought." }, "structural_constraints": {}, "fleet_evolution": { "not_applicable": true, "reason": "Compound Risk scenario \u2014 water stress and AI demand interaction; fleet evolution modeled in the companion texas_ercot_ai_demand scenario. This scenario focuses on water-side constraints." }, "compound_stressors": { "stressor_1_drought": { "name": "P90 La Ni\u00f1a Drought", "probability_pct": 10, "return_period_yr": 10, "brazos_river_flow_reduction_pct": 32, "colorado_tx_river_flow_reduction_pct": 28, "squaw_creek_reservoir_level_pct": 38, "cooling_water_available_pct_of_normal": 68, "affected_thermal_gw": 52.0, "derate_from_cooling_constraint_gw": 4.2, "notes": "Squaw Creek at 38% triggers NRC license condition at Comanche Peak: Units 1&2 derate from 2.4 GW to 2.1 GW. Brazos basin CCGT units lose 3.1 GW from reduced once-through cooling capacity. 10-year return period consistent with current La Ni\u00f1a cycle frequency." }, "stressor_2_thermal_derate": { "name": "Extreme Heat Ambient Derate", "threshold_temp_f": 105, "heat_dome_duration_days": 12, "heat_dome_frequency_per_decade": 3.8, "gas_ccgt_derate_pct": 9.5, "gas_peakers_derate_pct": 13.0, "nuclear_derate_pct": 5.6, "total_thermal_derate_gw": 5.8, "peak_demand_increase_gw": 7.2, "notes": "Carnot efficiency loss: gas CCGT combustion turbine inlet temperature correlation (ISO-rated capacity at 59\u00b0F drops 0.7% per \u00b0F above ISO). At 105\u00b0F: 38 GW CCGT \u00d7 9.5% derate = 3.6 GW lost. 14 GW peakers \u00d7 13% = 1.8 GW lost. Nuclear ambient derate (Comanche Peak) = 0.4 GW. Total = 5.8 GW. Peak demand increase from residential AC at 105\u00b0F = +7.2 GW vs design peak." }, "stressor_3_wind_drought": { "name": "Summer Wind Resource Depression", "cause": "Stable high-pressure dome, reduced surface roughness from drought-parched topsoil, reduced temperature gradient", "summer_wind_cf_drought_pct": 15.5, "summer_wind_cf_normal_pct": 31.0, "wind_generation_loss_gw": 7.2, "wind_cf_arithmetic_note": "AUDIT FLAG (MEDIUM): The stressor notes state summer wind CF drops from 31% to 15.5%, which implies a loss of 37 GW \u00d7 (0.310 - 0.155) = 5.74 GW, not 7.2 GW. The 7.2 GW figure implies CF drops to 31% - 7.2/37 = 11.5%, not 15.5%. If 5.74 GW (not 7.2 GW) is used in the compound event matrix, effective capacity rises to 120.3 GW (not 118.8 GW), reserve margin becomes 2.6% instead of 1.4% \u2014 still well below the 15% mandate threshold but the LOLP and bess_exhaustion timeline change. The compound event matrix uses the higher 7.2 GW figure which may reflect an observed worst-case CF of 11.5% during historical heat domes rather than the stated 15.5% average. Both figures should be disclosed.", "duration_days": 8, "correlation_with_heat_dome": "high", "notes": "West Texas wind speed strongly anti-correlated with summer heat domes (r = -0.72 per ERCOT wind study 2024). Parched soil reduces boundary layer turbulence. 37 GW installed \u00d7 (15.5% - 31%) CF = -5.7 GW below average. During the 8 worst days of a heat dome, wind generation drops below 6 GW from a 37 GW fleet." }, "stressor_4_ai_demand": { "name": "AI Data Center Continuous Load + Cooling Water Draw", "ai_load_gw_2031": 15.0, "water_demand_mgd_per_100mw": 1.8, "ai_cooling_water_demand_bgd": 0.27, "ai_demand_curtailability_pct": 30, "notes": "AI training workloads can be curtailed 30% via demand response contracts. Inference workloads (30% of AI load) are non-curtailable \u2014 SLA-protected. Total AI water demand 0.27 BGD adds directly to thermal fleet competition for constrained Brazos/Colorado river water." } }, "compound_event_matrix": { "scenario_name": "P90 Compound: All Four Stressors Simultaneous", "probability_annual_pct": 3.2, "mandate_fleet_nameplate_2031_gw": 136.0, "derating_summary": { "cooling_water_constraint_gw": 4.2, "ambient_thermal_derate_gw": 5.8, "wind_depression_gw": 7.2, "total_effective_capacity_loss_gw": 17.2 }, "effective_capacity_gw": 118.8, "peak_demand_under_heat_dome_gw": 117.2, "reserve_margin_pct": 1.4, "bess_effective_hours": 6.0, "bess_exhaustion_hour": 6, "post_bess_reserve_margin_pct": -10.8, "lolp_12hr_event_pct": 22.0, "lmp_above_1000_probability_pct": 38, "lmp_above_5000_probability_pct": 14, "notes": "BESS (14.6 GW, 6-hour duration) exhausts after 6 hours of sustained peak. Post-BESS: effective capacity 104.2 GW vs 117.2 GW peak demand = -11% reserve. LOLP 22% over a 12-hour P90 event. LMP at $1000+/MWh 38% of peak hours during compound event." }, "tech_vectors": [ { "id": "drought_hardened_cooling", "name": "Closed-Loop / Dry Cooling Retrofits", "description": "Retrofit 22 GW of Brazos and Colorado basin gas CCGT from once-through or open recirculating cooling to closed-loop or air-cooled dry cooling. Reduces water withdrawal by 92% and eliminates cooling-water derating under drought conditions. Capital cost ~$85M/GW. Texas Commission on Environmental Quality (TCEQ) permitting required; 18\u201330 month construction window per site.", "target_capacity_gw": 22.0, "derating_prevented_gw": 3.6, "water_saved_bgd": 0.9, "ce_model_mapping": "grid_stability water_stress", "estimated_mt_co2": 0.0, "constraints": { "total_lead_time_yr": 2.5, "critical_path": "TCEQ permit + EPC contractor availability (limited Texas-based dry-cooling specialists)", "cost_usd_b": 1.87, "cost_per_gw_usd_m": 85 } }, { "id": "extended_duration_storage", "name": "Long-Duration Storage (12\u2013100hr)", "description": "Deploy 4 GW of long-duration storage (iron-air batteries, compressed air energy storage, pumped hydro at Toledo Bend) with 12\u2013100 hour discharge duration to bridge compound stress events that outlast 4\u20138 hour BESS. Iron-air (Form Energy) projects in Permian Basin; CAES in East Texas salt caverns. Provides sustained reliability insurance that 4-hour BESS cannot.", "target_capacity_gw": 4.0, "storage_duration_hr": 24.0, "ce_model_mapping": "grid_stability", "ce_model_gap": "Long-duration storage not in CE TECHS_ABATE; no UCAP credit model for >8hr systems", "estimated_mt_co2": 6.0, "constraints": { "total_lead_time_yr": 3.5, "critical_path": "Iron-air commercial scale-up (Form Energy first GW-scale deployment 2026); CAES geology survey", "cost_usd_b": 3.2, "cost_per_gw_usd_m": 800 } }, { "id": "ai_water_efficiency", "name": "AI Data Center Water Use Efficiency + Dry Cooling", "description": "Require new AI/hyperscale data center permits (>100 MW) in Brazos/Colorado basin to use closed-loop dry cooling or advanced liquid cooling (direct-to-chip), reducing water withdrawal to <10% of evaporative tower baseline. PUCT/TCEQ joint permitting standard. Reduces AI cooling water demand from 0.27 BGD to 0.03 BGD \u2014 removing data centers from the Brazos basin water competition entirely.", "target_capacity_gw": 12.0, "water_saved_bgd": 0.24, "ce_model_mapping": "water_stress", "estimated_mt_co2": 0.0, "constraints": { "total_lead_time_yr": 1.0, "critical_path": "TCEQ rulemaking; hyperscaler cooperation (Microsoft, Google already using liquid cooling at scale)", "cost_usd_b": 0.8, "cost_per_gw_usd_m": 67 } }, { "id": "spp_emergency_interconnection", "name": "Emergency SPP Interconnection Upgrade", "description": "Upgrade ERCOT\u2013SPP AC ties from 0.9 GW to 4.5 GW via two new 765kV DC asynchronous ties in the Panhandle (connects to SPP wind-rich South region) and an AC upgrade on the Laredo tie. Would provide emergency import capacity during compound stress events at the cost of partial ERCOT market sovereignty. DOE National Interest Electric Transmission Corridor designation required.", "target_capacity_gw": 3.6, "ce_model_mapping": "transmission grid_stability", "estimated_mt_co2": 3.0, "constraints": { "total_lead_time_yr": 4.0, "critical_path": "Congressional authorization (ERCOT sovereignty statute); DOE NEITC designation; ROW across private land", "cost_usd_b": 2.1, "cost_per_gw_usd_m": 583 } } ], "model_gaps": [ { "gap": "Water-energy nexus feedbacks not in CE grid model", "impact": "HIGH \u2014 CE GridStabilityService does not model cooling water availability as a capacity constraint. Derating from drought conditions requires manual parameterization.", "mitigation": "WaterStressService drought_multiplier applied as external derate to structural_constraints" }, { "gap": "Compound event joint probability", "impact": "HIGH \u2014 CE models stressors independently. P90 drought \u00d7 P90 heat \u00d7 P90 wind drought \u00d7 AI demand has a 3.2% annual probability but CE cannot currently compute joint CDFs.", "mitigation": "Compound event modelled as single scenario with manually combined derate parameters" }, { "gap": "AI data center demand controllability", "impact": "MEDIUM \u2014 CE has no model for AI workload-level demand response. 30% curtailability assumed but not computed from workload type distributions.", "mitigation": "Manual split: 70% firm load (inference) + 30% curtailable (training batch) hardcoded into demand parameters" }, { "gap": "ERCOT scarcity pricing endogeneity", "impact": "MEDIUM \u2014 ERCOT energy-only market LMP dynamics during shortage differ fundamentally from capacity market models. $9,000/MWh scarcity cap creates investment signals CE doesn't capture.", "mitigation": "LMP distribution captured in compound_event_matrix as probability outputs" } ], "analysis": { "critical_path": "drought_hardened_cooling", "abatement_needed_mt_co2": 2.0, "confidence": "medium", "confidence_rationale": "Thermal derating curves well-established from EIA power plant heat rate data. Water-cooling correlations derived from USGS Texas water use reports. Wind depression under drought correlated from ERCOT wind study (r = -0.72). Joint probability estimate is CE proprietary calculation with wide uncertainty band (1.8%\u20135.1% annual).", "key_outputs": { "water_per_mwh_thermal_gal": 420, "water_per_mwh_ai_cooling_gal": 180, "cooling_constraint_probability_pct": 34, "datacenter_curtailment_probability_pct": 19, "gas_pipeline_bottleneck_probability_pct": 27, "blackout_probability_p90_event_pct": 22, "lmp_above_1000_pct_of_peak_hours": 38 }, "notes": "This scenario demonstrates that the BESS + solar buildout from the ERCOT AI mandate is necessary but not sufficient for compound physical reliability. The mandate fleet achieves its design objective (serve 15 GW AI demand) but fails under 8-day compound drought + heat dome events. The policy gap is 4 GW of long-duration storage and 22 GW of cooling water independence \u2014 investments not required by the current mandate framework.", "estimated_total_mt_co2": 9.0 }, "projections": { "years": [ 2026, 2027, 2028, 2029, 2030, 2031 ], "bau_mt_co2": [ 99.0, 104.0, 110.0, 116.0, 124.0, 133.0 ], "mandate_mt_co2": [ 99.0, 98.5, 98.0, 97.5, 97.2, 96.8 ], "compound_stress_effective_capacity_gw": [ 117.0, 115.2, 113.8, 112.1, 110.9, 118.8 ], "compound_stress_peak_demand_gw": [ 95.0, 99.0, 103.0, 107.0, 112.0, 117.2 ], "compound_reserve_margin_pct": [ 23.2, 16.4, 10.5, 4.8, -1.0, 1.4 ], "ceiling_mt_co2": 97.0, "notes": "Compound effective capacity includes drought derate + thermal derate + wind depression. Reserve margin turns negative in 2030 under P90 scenario without cooling retrofits and long-duration storage. The mandate-case fleet (with BESS) recovers to 1.4% margin in 2031 only because 12 GW of BESS is counted at UCAP \u2014 but BESS exhausts after 6 hours." }, "non_compliance": { "trigger_year": 2030, "mandate_cost_label": "~$7.9B", "mandate_cost_description": "Cooling retrofits ($1.87B) + long-duration storage ($3.2B) + AI water efficiency ($0.8B) + SPP interconnection ($2.1B)", "mechanism": "NERC LOLP >5% triggers mandatory reliability corrective action plan; PUCT forced procurement of emergency capacity; hyperscaler SLA violation cascade triggers $2\u20134B in contract penalties and workload migration risk.", "affected_exports_usd_b": 145.0, "max_annual_cost_usd_b": 11.4, "five_year_cumulative_usd_b": 28.3, "tax_schedule": [ { "year": 2028, "rate_usd_per_t": 0, "annual_cost_usd_b": 2.4, "cumulative_usd_b": 2.4, "note": "SLA penalties from 2028 compound stress events" }, { "year": 2029, "rate_usd_per_t": 0, "annual_cost_usd_b": 5.1, "cumulative_usd_b": 7.5, "note": "Rotating outage costs; hyperscaler migration announcements" }, { "year": 2030, "rate_usd_per_t": 0, "annual_cost_usd_b": 8.2, "cumulative_usd_b": 15.7, "note": "LOLP breach; PUCT emergency procurement; AI investment redirection" }, { "year": 2031, "rate_usd_per_t": 0, "annual_cost_usd_b": 11.4, "cumulative_usd_b": 27.1, "note": "Full compound event; SPP emergency interconnection; grid sovereignty question" } ], "affected_sectors": [ { "name": "AI / Hyperscale Data Centers", "icon": "fa-server", "export_value_usd_b": 85.0, "jobs": 185000, "notes": "Microsoft, Google, Meta, Amazon have announced 15+ GW of Texas AI capacity. SLA penalties and workload migration triggered above 0.001% annual downtime. One compound event = $200\u2013800M per major operator." }, { "name": "Semiconductor Fabrication", "icon": "fa-microchip", "export_value_usd_b": 18.0, "jobs": 29000, "notes": "Samsung Austin/Taylor, Texas Instruments: wafer fabs require uninterruptible power. Compound events destroy in-process inventory. Single 4-hour outage = $40\u201380M inventory loss per fab." }, { "name": "Petrochemical / Refining", "icon": "fa-industry", "export_value_usd_b": 42.0, "jobs": 64000, "notes": "Gulf Coast refineries and ethylene crackers: continuous-process equipment damaged by unplanned shutdown. Power + cooling water constraint hits simultaneously under compound drought scenario." }, { "name": "Water Utilities / Municipal", "icon": "fa-droplet", "export_value_usd_b": 0.0, "jobs": 12000, "notes": "Brazos Municipal Water Authority serves 1.5M residents competing with power sector cooling. Drought + power outage creates simultaneous water pumping and treatment failure \u2014 cascading human health risk." } ] }, "action_items": [ { "id": "ai_01", "audience": "corporate_industrial_buyer", "action": "Texas data centre operators (Microsoft, Google, Meta, Oracle): publish binding water usage effectiveness (WUE) targets and disclose actual water consumption (m\u00b3/MWh) in annual sustainability reports \u2014 water allocation in Texas is subject to prior appropriation doctrine and undisclosed consumption creates regulatory exposure.", "rationale": "Texas water law (prior appropriation) means data centres with undisclosed consumption profiles face curtailment risk when junior water rights are called during drought. Voluntary disclosure now establishes a defensible consumption baseline and demonstrates regulatory good faith before mandatory disclosure is imposed.", "defensible_basis": "Texas Water Code \u00a7 11 (prior appropriation); TWDB Water Use Survey (annual mandatory reporting for water users > 10 acre-feet/day); SEC climate disclosure requirements. Disclosure is already legally required under TWDB for large water users \u2014 voluntary corporate reporting aligns with existing obligation.", "urgency": "immediate", "no_regret": true }, { "id": "ai_02", "audience": "utility_grid_operator", "action": "ERCOT: incorporate wet-bulb temperature stress testing (not just dry-bulb heat) in summer grid adequacy reviews \u2014 and publish the wet-bulb-adjusted capacity derating assumptions for gas and coal generators that show materially lower output at peak wet-bulb conditions.", "rationale": "The compound risk in this scenario is wet-bulb temperature reducing generator thermal efficiency simultaneously with peak air conditioning demand. ERCOT's current adequacy model uses dry-bulb temperature as the stress parameter. Wet-bulb derating produces 5\u20138% lower available capacity estimates during the scenario's critical event window.", "defensible_basis": "ERCOT Summer 2023 near-miss postmortem analysis; EIA generator heat rate data by ambient temperature; NERC HILF (High Impact Low Frequency) Event risk methodology. Methodology update within ERCOT's planning authority \u2014 no regulatory change required.", "urgency": "immediate", "no_regret": true }, { "id": "ai_03", "audience": "sovereign_policymaker", "action": "Texas PUC and TWDB: establish a joint data centre water-energy nexus monitoring programme \u2014 requiring large load interconnection applicants to disclose cooling water source, daily consumption, and curtailment tolerance before ERCOT queue entry.", "rationale": "Currently there is no coordination between ERCOT's power queue and TWDB's water allocation system. A data centre approved by ERCOT but drawing from an over-allocated aquifer is a compound risk \u2014 ERCOT reliability depends on power; TWDB integrity depends on water. Joint disclosure closes a regulatory gap at no cost.", "defensible_basis": "ERCOT Nodal Protocol Sections 5.3 and 5.4 (interconnection application requirements); TWDB Regional Water Planning mandate (SB 1 1997); Texas Government Code \u00a7 2310 (agency coordination authority). Joint programme requires only an interagency MOU \u2014 no legislation.", "urgency": "near_term", "no_regret": true }, { "id": "ai_04", "audience": "corporate_industrial_buyer", "action": "Data centre developers in Texas: evaluate air-side economisation or direct liquid cooling (DLC) designs for new facilities that reduce or eliminate cooling water consumption \u2014 DLC systems achieve WUE of 0.0 (no evaporative water use) vs industry average 1.5\u20132.5 L/kWh.", "rationale": "Air-side economisation and DLC are commercially proven technologies deployed at hyperscale facilities in Europe and the Pacific Northwest. In Texas, where wet-bulb temperatures limit economisation hours, DLC is the more robust approach. New facilities built now will operate through the 2035+ water stress window \u2014 design for the future climate, not the current.", "defensible_basis": "Green Grid PUE and WUE measurement standards; Schneider Electric White Paper 219 (DLC cost-benefit); ASHRAE Standard 90.4 (data centre energy efficiency). Commercial technology available at equivalent or lower TCO than air cooling for GPU-dense AI workloads.", "urgency": "near_term", "no_regret": true } ], "sources": [ "ERCOT Capacity, Demand & Reserves Report 2025 \u2014 P50/P90 scenarios", "NRC Safety Evaluation Report \u2014 Comanche Peak Nuclear Power Plant, Squaw Creek Reservoir Conditions", "USGS Texas Water Use Report 2023 \u2014 Thermoelectric cooling withdrawals by river basin", "ERCOT Wind Integration Study 2024 \u2014 Summer wind CF under drought conditions", "EIA Form EIA-860 \u2014 Texas generating unit cooling system types 2025", "NERC 2024 Long-Term Reliability Assessment \u2014 ERCOT compound risk section", "Form Energy Iron-Air Battery Commercial Deployment Roadmap 2025", "TCEQ Water Rights Database \u2014 Brazos River basin thermoelectric allocations" ], "created": "2026-05-19", "last_updated": "2026-05-19", "author": "CE Scenario Engine v3.7", "failure_conditions": [ "P90 compound event (drought + heat dome + wind depression + 15 GW AI demand) arrives before closed-loop cooling retrofits (22 GW) are complete, leaving 4.2 GW of thermal capacity subject to Brazos/Colorado cooling water constraint and pushing ERCOT below the 15% reserve margin threshold", "AI data centre water demand (0.27 BGD evaporative cooling) is not regulated by TCEQ by 2028, adding directly to Brazos/Colorado river draw competition and triggering priority curtailment of hyperscaler cooling water during drought conditions", "Long-duration storage (4 GW / 24hr iron-air or CAES) is not commercially available or financeable in Texas by 2029, leaving ERCOT reliant on 6-hour BESS that exhausts during sustained 8-12 hour compound events", "ERCOT-SPP interconnection upgrade (3.6 GW) blocked by Texas sovereignty legislation, eliminating emergency import capacity and requiring rotating outages during compound events (reserve margin -10.8% post-BESS)", "AI workload demand response (30% curtailability) fails operationally because inference workloads (SLA-protected) account for more than 50% of AI load by 2031, reducing effective AI DR to 15% and adding 2+ GW of uncurtailable demand at peak", "Squaw Creek Reservoir (Comanche Peak) falls below 38% capacity threshold during a multi-year La Nina drought, triggering NRC mandatory derate of 2.4 GW \u2192 2.1 GW \u2014 removing 300 MW of baseload from an already constrained compound event fleet" ], "decision_windows": [ { "id": "dw_01", "actor_type": "sovereign_treasury", "region": "Texas (TCEQ / PUCT)", "decision": "TCEQ and PUCT jointly adopt closed-loop or dry-cooling requirement for all new AI data centre permits >100 MW in Brazos/Colorado River basins by 2027-Q2, eliminating 0.27 BGD of cooling water competition with thermal generation", "time_horizon": "medium_term", "deadline": "2027-Q2", "fiscal_instrument": "other", "consequence_if_missed": "AI data centres and thermal generators compete for same drought-constrained river water during compound events; forced priority curtailment order between economic sectors becomes politically unavoidable", "no_regret": true }, { "id": "dw_02", "actor_type": "project_developer", "region": "Texas (Brazos / Colorado river basins)", "decision": "ERCOT market participants initiate closed-loop cooling retrofits on 22 GW of Brazos/Colorado basin gas CCGT by 2027-Q1 using PUCT reliability cost-recovery mechanism", "time_horizon": "medium_term", "deadline": "2027-Q1", "fiscal_instrument": "other", "consequence_if_missed": "22 GW of thermal capacity remains exposed to cooling water constraint; 4.2 GW derating under P90 drought persists; compound event reserve margin stays below 2%", "no_regret": true }, { "id": "dw_03", "actor_type": "institutional_investor", "region": "Texas (Permian Basin / East Texas)", "decision": "Form Energy or equivalent closes first GW-scale iron-air BESS project finance in Texas (24-hour duration) by 2028-Q1, providing compound event bridge storage that 4-6hr lithium BESS cannot", "time_horizon": "medium_term", "deadline": "2028-Q1", "fiscal_instrument": "bond_issuance", "consequence_if_missed": "ERCOT relies entirely on 6-hour BESS for compound events; post-BESS reserve margin of -10.8% creates LOLP 22% for 12-hour compound stress events; rotating outages become the default compound response", "no_regret": false }, { "id": "dw_04", "actor_type": "sovereign_treasury", "region": "Texas / Federal (FERC)", "decision": "US Congress authorises ERCOT-SPP emergency interconnection upgrade (3.6 GW, $2.1B) as National Interest Electric Transmission Corridor by 2028-Q4, preserving ERCOT isolation for normal operations but enabling emergency imports during compound events", "time_horizon": "structural", "deadline": "2028-Q4", "fiscal_instrument": "other", "consequence_if_missed": "ERCOT has no emergency import mechanism during compound events; rotating blackouts are the only load-shedding option; $28B+ Uri-scale economic damage exposure persists for all compound events through 2035", "no_regret": false }, { "id": "dw_05", "actor_type": "central_bank", "region": "ERCOT / NERC region", "decision": "NERC publishes ERCOT compound water-energy stress test standard (P90 drought + P90 heat + wind depression joint CDF) by 2028-Q2, establishing quantitative derating standards for cooling water constraints and AI demand variability", "time_horizon": "medium_term", "deadline": "2028-Q2", "fiscal_instrument": "stress_test", "consequence_if_missed": "Compound water-energy risk remains unquantified in ERCOT reliability planning standards; cooling water constraints are not flagged as capacity derating in seasonal adequacy assessments; investors and insurers are underpricing compound event risk", "no_regret": true } ] }