Deliverability remains the dominant constraint. In ERCOT North, much of the new AI-driven demand is landing in areas that depend on imported power. Raising the deliverability factor (D) from 0.85 to 1.00 increases DCRM from 4.586 GW (Scenario A) to 5.484 GW (Scenario F),which is nearly 1 GW swing without adding any new capacity. This reflects what we’re seeing on the ground: even when generation exists somewhere on the grid, getting it to a data center campus at the right time is often the limiting factor.
Peak coincidence acts as a conditional relief valve. When data center load overlaps less with system peak (through AI-aware scheduling or flexible workload shifting), the grid absorbs more without needing to expand. Reducing CF from 1.0 to 0.8 (Scenario B to E) adds ~50 MW of headroom. In high-performing scenarios that affect compounds, Scenario F, with low coincidence and strong delivery, stretches the DCRM ceiling to nearly 5.5 GW. It suggests that real-world strategies like scheduling inference overnight or batch training workloads off-peak could meaningfully ease stress if grid flexibility and incentives are aligned.
Climate events compress the margin from both ends. In scenarios like G and C, the grid faces a double hit: available supply drops (high V), while baseline demand rises due to heat-driven cooling needs (ΔB = 5,000–6,000 MW). Even with full coincidence, that extra base load eats into available headroom fast, pulling DCRM down to as low as 2.973 GW. This mirrors what’s already happening in summer peak alerts where firm capacity may exist, but volatility + urban cooling demand crowd out growth.
What this reveals is a system with very little structural slack. No single constraint pushes it past the threshold, but together, they bend the margin rapidly. DCRM behaves like a constrained slack function, quantifying how much capacity remains once the system accounts for delivery limits, demand timing, and weather-driven volatility. Instead of a binary verdict, it shows how close the system is to tipping, and what kinds of stress combinations matter most.
FINAL THOUGHTS
ERCOT North may not be the most at-risk region, but it captures the core insight behind DCRM: when AI- driven demand, transmission constraints, and planning uncertainty converge, the margin for error narrows quickly. Under base assumptions, DCRM shows the region can absorb up to 4.59 GW of new AI-scale data center load, a figure that may appear sufficient, but already mirrors volumes in the interconnection queue.
Sources & References:
1. ERCOT 2025 Capacity, Demand and Reserves (CDR) Report (May 2025)
2. ERCOT Generation Interconnection Queue 3. ERCOT 2025 Long-Term Load & Forecast Methodology (LTLF) 4. NERC 2025 Summer Reliability Assessment 5. NERC 2025 Long-Term Reliability Assessment 6. NERC Planning Reserve Margin Guideline 7. ERCOT Monthly Outlook / MORA (May 2025)
The data, comments and/or opinions contained herein are provided solely for informational purposes by ADM Investor Services, Inc. (“ADMIS”) and in no way should be construed to be data, comments or opinions of the Archer Daniels Midland Company. This report includes information from sources believed to be reliable and accurate as of the date of this publication, but no independent verification has been made and we do not guarantee its accuracy or completeness. Opinions expressed are subject to change without notice. This report should not be construed as a request to engage in any transaction involving the purchase or sale of a futures contract and/or commodity option thereon. The risk of loss in trading futures contracts or commodity options can be substantial, and investors should carefully consider the inherent risks of such an investment in light of their financial condition. Any reproduction or retransmission of this report without the express written consent of ADMIS is strictly prohibited. Again, the data, comments and/or opinions contained herein are provided by ADMIS and NOT the Archer Daniels Midland Company. Copyright (c) ADM Investor Services, Inc.
In other words, the strain is already materializing. Traditional reserve margin metrics assume power flows freely across space and time. DCRM makes explicit what those models miss, which is that electrons don’t move freely, and new loads won’t politely arrive off-peak.
COMING UP: THE GLOBAL STRESS TEST
In the last Power Bytes note, I made the case that the AI race is increasingly shaped by physical systems, not just ambition or capital. This note delivered on the next step: introducing DCRM, the tool built to expose where those physical limits are forming. And after testing it on ERCOT North, we can finally apply it to the question it was built for. In the next piece, I’ll use DCRM to assess the U.S., China, and Europe, comparing how much strain each system can absorb, how infrastructure is pacing (or not) with AI expansion, and where the structural gaps are largest.
If you’d like to dive deeper into the DCRM framework or following the AI × energy space/working in the sector, I’d be happy to connect — you can find a time here.
Aishwarya Mahesh E:
aishwarya.mahesh@
admis.com T: 2127851555
21 | ADMISI - The Ghost In The Machine | Q3 Edition 2025
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42
