DER Weekly Digest — Week Ending November 6, 2026


This week: the largest demand-flexibility commitment this series has ever reported and the strictest eligibility rules it has ever been able to quote arrived within forty-eight hours of each other. On November 5–6 OpenAI announced “Grid Protectors,” under which Georgia Power holds the right to curtail up to 1,000 MW at OpenAI’s Effingham County campus — roughly 31% of that site’s projected ~3,200 MW — under a 25-year agreement approved in August, with power phasing in from 2028 to 2032. Georgia Power declined to say when it may call the resource or how the 1,000 MW will be measured and verified. Two days earlier, on November 3, the Florida PSC voted 5-0 to reject Duke Energy Florida’s large-load tariff, the first compliance filing tested under SB 484, finding Duke had not shown it would “reasonably ensure” each large load bears its full cost of service. And a targeted search this week finally surfaced PJM’s qualifying-offset criteria — the sharpest unanswered question in this series since October 29 — in the draft manual language from PJM’s own stakeholder process: demand-side offsets require a signed commitment with new load that already holds an executed service agreement, carry a minimum three years of registration history, demand binding site-level commitments for the full fifteen-year term with letters of intent not accepted, and exclude locations that supported a Demand Resource in a prior delivery year unless the increase comes from new generation or storage investment at that location.

Read together, these three are one development. The demand side’s ceiling just rose to a gigawatt and its floor just became documentary. A thousand megawatts pledged for twenty-five years is the most valuable flexibility commitment on the North American record; a thousand megawatts with no disclosed call window, no stated measurement point and no published verification method is, in capacity terms, an assertion. PJM’s criteria are the mirror image: narrow, burdensome, and for the first time specific enough to test a portfolio against. The two documents describe the same transaction from opposite ends — one all scale and no evidence, the other all evidence and little room.

That sharpens, rather than replaces, the frame this series carried out of October. Last week’s reading was that the demand side is being paid more and trusted less, and that the contested term is no longer the megawatt but the measurement point. This week supplies the proof at both extremes. Georgia Power will not name the measurement point on 1,000 MW. PJM names it precisely — an executed EDC agreement covering metering, telemetry and baseline measurement, site by site — and in naming it, excludes most of what an aggregator already owns. The practical instruction follows directly: the dormant-site question is answered, and the answer is no. A location that supported a Demand Resource in a prior delivery year does not qualify as an offset unless new generation or storage was installed there. Incremental battery additions at existing sites do qualify. Re-enrollment of idle capacity does not.


⚡ Virtual Power Plants & Demand Flexibility

FranklinWH put a per-kilowatt price on residential battery flexibility in two more utility territories, and dropped the solar requirement. On October 27 FranklinWH launched a residential-battery demand-response program with Entergy Louisiana paying $1,000 at enrollment (if applied within a year of installation) plus an annual performance payment of $125 per average kW delivered over the year, paid by January 31 as a Visa prepaid card. Entergy may call events any day, year-round, standing down only when a severe storm is forecast so customer backup is preserved. The parallel Alliant Energy program in Iowa and Wisconsin pays $100/kWh up to $1,500 one-time at enrollment plus $100 per average kW discharged on utility call, with the household choosing how much stored energy to make available. Entergy projects 155 MW of dispatchable VPP capacity by 2030 across all its VPP programs. Enrollment runs through a Derapi integration connecting FranklinWH’s 15 kWh aPower 2 LFP systems — scalable to 225 kWh, fifteen units per aGate — into Entergy’s Smart Hours platform.

Two design choices here are worth copying and one is worth pricing carefully. Worth copying: participants need no solar and qualify if the battery charges directly from the grid under their interconnection service agreement. That is a program valuing pure load-shifting storage rather than solar self-consumption, and it widens the addressable fleet to every customer willing to buy a battery — the single largest expansion of residential VPP eligibility in any program this series has tracked. Also worth copying: the storm stand-down. Preserving customer backup during forecast severe weather removes the objection that kills residential enrollment, and it costs the utility nothing in the hours it actually needs the resource.

Worth pricing carefully: the compensation structure is per-average-kW-over-the-year, not per-event-kW. A resource that performs at 9 kW in four events and 0 kW in the remaining calls earns on the annual average, which means the incentive rewards availability across the full call season rather than depth in any single event — the opposite of how a capacity market accredits. Model the two side by side before assuming the payment and the accreditation move together. And note what the program discloses that most do not: a dollar figure per kW. At $125 per average kW/year, Entergy is paying roughly $125/kW-year for residential battery flexibility against the ~$66/kW-year avoided-cost anchor this series carries — about 1.9×, which says either that acquisition cost is being bought down deliberately or that Entergy’s own avoided cost is materially above the generic benchmark. Ask which, in any jurisdiction where you are defending a lower number. (Source: MGRID, November 5 — November 5 entry)

New York is consolidating residential direct load control into one statewide platform, and funding it from the state budget rather than from rates. The New York PSC approved modifications across National Grid’s three entities — The Brooklyn Union Gas Company, KeySpan Gas East (KEDLI) and Niagara Mohawk (NMPC) — that reduce the number of required program tests each year and lower enrollment thresholds, and advanced a new statewide Excelsior Power Program: a centralized platform for enrolling residential electric and gas direct-load-control devices providing both summer and winter load relief. New participating smart-thermostat customers receive a $25/month bill credit in their first year (customers already enrolled in a utility thermostat or Bring-Your-Own-Thermostat program before the 2026–2027 state budget are excluded from the enhanced incentive), backed by $33M in state-budget funding, with a Phase 1 implementation plan now open for public comment. The Commission also recommended migrating multifamily customers out of National Grid’s thermostat program into Excelsior.

The structural move is the platform, and it addresses the cost line that actually sinks residential DR programs. Customer acquisition and DLC administration are fixed costs borne separately by every utility running its own thermostat program; a statewide enrollment platform spreads them across the state’s load and standardizes measurement and verification in one place — which is also the precondition for a defensible aggregate capacity claim. The winter-plus-summer scope matters more than it looks: a dual-peaking system that can only call summer events is accrediting half a resource, and New York’s winter peak is the half the reliability case now turns on. The multifamily migration is the quiet win — this series flagged a multifamily enrollment channel as a standing requirement two weeks ago, and a statewide platform with a multifamily pathway is the cheapest version of it.

The caveat is the familiar one and it is disqualifying for citation purposes: the release discloses no MW load-relief target, no enrollment count and no cost-effectiveness finding. Treat Excelsior as a program-design template and a funding precedent — $33M of state budget money rather than a rate-case ask is the part to show your own commission — and not as a capacity number until Phase 1 produces one. (Source: EIN Presswire, October 16 — November 2 entry)

Ember put a ceiling on data-center load-shifting, and it is low enough to change how large-load DR should be sold. An Ember analysis finds that only about 3% of global data-center capacity can realistically shift its electricity use by 2030, because roughly 87% runs cloud services, corporate IT and AI inference that are hard to pause or relocate, leaving the ~13% that is AI training as the genuinely shiftable slice. The consequence: pure load-shifting trims U.S. peak demand only about 0.1–0.6% by 2030 — Texas up to ~0.4%, data-center-heavy Ireland and Malaysia up to ~1.5%. The worked example is the useful part. Two Texas AI campuses — Stargate Abilene and Comanche Circle — adding about 5.6 GW of load (averaging ~2.7 GW at 48% utilization) could shed only ~0.55 GW through load-shifting during a stress event, versus the full 2.7 GW (~2% of ERCOT’s projected 2030 peak) if shifted entirely to onsite generation. A companion Currence projection holds that non-firm interconnection rules could create up to 285 GWh of U.S. battery demand by 2035, PJM more than half of it, with required curtailment durations ranging from ~1.3 hours under uniform ERCOT rules to 5.3–7 hours under “last-in-first-curtailed” regimes.

This is the number to hold against every gigawatt-scale flexibility pledge, including the one leading this digest. Ember’s finding is not that large-load flexibility is small — it is that flexibility delivered by rescheduling compute is small, and flexibility delivered by on-site storage and generation is roughly five times larger at the same site. The 0.55-versus-2.7-GW split in the Texas example is the entire argument in one figure. For anyone building a large-load DR product the implication is a reallocation, not a retreat: sell the storage and the on-site generation, treat workload shifting as the thin incremental layer, and derate any claim that does not say which of the two is producing the megawatts. And read the Currence duration range as a procurement spec — a curtailment regime that can hold a load down for 5 to 7 hours requires a fundamentally different asset than one that calls 1.3 hours, and the rule that sets the duration is written before the battery is sized. (Source: Latitude Media, October 29 — November 2 entry)

GridPoint is targeting 2 GW of commercial storage across 20,000+ sites it already controls — which makes the controllable-load question the interesting one. GridPoint will add on-site battery storage to its commercial energy platform, aiming for 2 GW of distributed capacity across a network of more than 20,000 commercial sites including multi-site operators such as Panda Express, Walgreens and Wendy’s, offered as a subscription mirroring the model it already uses to manage HVAC, refrigeration and lighting loads. The value stack is bill optimization, peak-demand reduction, grid services and resilience, with stored energy earning payments through utility and capacity programs. GridPoint says customers typically save ~20% today and qualifying sites could see up to an additional 15% from batteries, against commercial rates that rose 42% over five years across its sites. Returning co-founder and CEO Adrian De Luca attributes the timing to falling battery input costs and clearer investment-tax-credit financing including transferability; the firm will integrate off-the-shelf batteries rather than build its own.

The aggregation thesis here is stronger than the storage announcement, and it is the part a utility should engage with. A 2 GW battery target across 20,000 sites is a plan, not a fleet. What already exists is control of thermal, refrigeration and lighting load at 20,000 commercial sites — and refrigeration in particular is thermal storage that is already paid for. Adding batteries to a portfolio that is already dispatchable produces multi-objective dispatch across two asset classes behind one meter at one site, simultaneously pursuing demand-charge reduction, energy arbitrage and a capacity-program commitment. That is the hardest function in a DERMS specification and the one most often assumed rather than tested: when the bill-optimization objective and the grid-service call conflict, which wins, who decides, and what does the utility see? Ask it of any aggregator offering a managed C&I portfolio, and require the arbitration logic in writing.

The caveats are substantial: the 2 GW is a target rather than installed capacity, and the announcement names no pricing, no deployment timeline, no utility programs and no battery or financing partners — so the pace, and the share that will actually bid into capacity or DR markets, are unproven. (Source: MGRID, November 5 — November 5 entry)


🔌 DERMS & Grid Integration Technology

CAISO’s board approved a large-load interconnection framework whose only flexible product is an instrumented, remotely disconnectable connection — and the telemetry requirements are a DERMS specification. The CAISO Board of Governors approved on October 28 a large-load interconnection package defining a “large load” as an end user with ≥50 MW peak at a single site — CAISO kept 50 MW over PG&E’s proposed 20 MW, reserving the right to revisit — and offering a single flexible path: Flexible Interim Load Interconnection (FILI), under which a load accepts enforceable, grid-condition-based curtailability until its associated network upgrades are complete, backed by required hourly forecasts, telemetry, phasor-measurement-unit-grade monitoring in the transmission owner’s tariff, and CAISO’s ability to remotely disconnect the load to preserve reliability. The August straw proposal’s permanent flexible option (FLIP) was dropped. CAISO had to file tariff language at FERC by November 16 (Docket EL26-71), responding to FERC’s June show-cause order directing CAISO and five other operators to prove large loads can connect quickly without shifting costs. Study windows are compressed to 60–90 days, a timeline Microsoft and others questioned. CAISO cites a California Energy Commission forecast of only 1.8 GW of in-grid data-center growth by 2030 and 4.7 GW by 2035 — modest against ERCOT’s 191.8 GW of large-load studies.

Three things to take from this, in descending order of durability. First, the requirements list is the most specific public statement of what a flexible large load must be instrumented to do, and it maps almost line-for-line onto a large-load-management or DERMS functional spec: hourly forecast submission, continuous telemetry, PMU-grade monitoring, and a remote-disconnect path the operator controls. Put all four in RFP evaluation criteria, and note that PMU-grade monitoring at a load connection is a materially higher bar than SCADA polling — it is the instrument that settles whether a curtailment actually happened, at the resolution a disturbance analysis requires. Second, dropping FLIP matters more than approving FILI. With the permanent option gone, flexibility in CAISO is a bridge to firm service rather than a durable resource class — a load that accepts curtailability does so to connect sooner, not to be paid for flexibility in perpetuity, and the commitment expires when its upgrades finish. That is a fundamentally different product from a DR contract, and it accrues no accreditation record. Third, most of the cost and operating detail sits in transmission-owner tariffs that have not been written, which is where the terms a customer actually experiences will be set.

The strategic read is that speed-to-power is now the currency flexibility is bought with, not money — and CAISO’s own numbers explain why California can afford that posture. 1.8 GW by 2030 against ERCOT’s 191.8 GW of study requests is a market with time to be careful. Utilities facing an ERCOT-shaped queue cannot copy the patience, only the instrumentation. (Source: MGRID, October 30 — November 3 entry; the November 16 FERC compliance deadline has been a standing tracked item in this series since the June show-cause orders)

A 12 MW / 60 MWh battery is being used to enforce a 7 MW utility import limit for eighteen months — which is flexible interconnection, implemented behind the meter, without waiting for a tariff. Critical Loop and developer SRG Commercial announced on October 27 a 25 MW temporary power system for a six-building advanced-manufacturing campus near Long Beach, pairing 12 MW / 60 MWh of battery storage — two CLB-5100 units per building, with Cygnus controllers that charge when grid capacity is available and discharge when building demand exceeds the import limit — with about 7 MW of firm plus 2 MW of flexible SCE service and 3.5 MW of normally-disconnected standby generation, designed to run at least 18 months until SCE completes two new feeders. Critical Loop owns, operates and maintains the batteries and sells the system as a managed service that relocates to SRG’s next site when the bridge ends. SRG’s “Power-Ready Industrial Portfolio” model resolves interconnection constraints across its pipeline before tenants lease — the tenant is reported as Anduril Industries, a ~$1B, 1.18M-sq-ft campus, with first buildings energizing ~February 2027.

The control stack is the transferable part, and it is the same stack CAISO just mandated, built by a private developer instead of an operator. Enforcing an import cap, arbitrating charge and discharge against available feeder capacity, and islanding to generation on outage are precisely the functional requirements a utility-grade DERMS or large-load-management platform must support — and they map onto the OpenADR / IEEE 2030.5 dispatch-and-constraint functions this series tracks. The difference is who holds the control: here a third party enforces the utility’s limit on the utility’s behalf, under a non-export Rule 21 application still under SCE review. SCE must approve the configuration, sign the interconnection agreement and grant permission to operate before energization.

Two engineering facts in this project are worth carrying into any import-limited design. The 60 MWh must recharge over the same ~7 MW feed, which at best is roughly nine hours of continuous import at full feeder capacity — so the recharge window, not the discharge rating, sets what the system can actually do day after day. And it is unclear who bears the cost if SCE’s feeders slip past 18 months, which is the real risk allocation in every bridge-power arrangement and the question to settle in the contract rather than the schedule. What the project quantifies for the record: how long a large load now waits for utility upgrades, and what storage is worth per month of that wait. (Source: MGRID, October 31 — November 4 entry)

A $27M federal microgrid award is a clean specification of the controller function at feeder scale. The U.S. Army Corps of Engineers awarded nearly $27M to AACON-RavenVolt JV (RavenVolt is an ABM subsidiary) to build a secure feeder-level microgrid at White Sands Missile Range, New Mexico, under the Department of Defense’s Energy Resilience and Conservation Investment Program, with operation expected in 2028. The system pairs a 600 kW carport solar array, a 500 kW / 2,000 kWh battery and a 670 kW dispatchable propane generator, coordinated by an intelligent microgrid control system enabling islanding from the local utility grid during outages and extreme weather, plus new distribution lines, fiber-optic communications and security infrastructure, delivered Design-Bid-Build.

Carry it for the architecture, not the business case. The driver is defense resilience rather than a ratepayer program, so there is no DR economics to borrow. But the controller is performing multi-asset dispatch, state-of-charge management and seamless grid-to-island transition — the same functions a utility DERMS performs at feeder scale — which is why federal and campus microgrid specifications are a useful source of controls requirements for a utility RFP: they are written by an owner who must prove the island works, and they are public. Note also the asset stack as a non-wires-alternative building block: solar + 2 MWh battery + firm genset + controller, which replicated across many sites is exactly the aggregated flexible capacity a VPP monetizes — and the 500 kW / 2,000 kWh disclosure is the correct form, kW stated alongside kWh, which this series has had to ask for repeatedly this quarter. (Source: MGRID, November 5 — November 6 entry)


🔋 Storage, Distributed Solar & DER Procurement

RMI measured California batteries carrying peak load at 64% of installed capacity against gas peakers’ 67% — and that three-point gap is the most useful accreditation number of the week. RMI estimates California utility-scale batteries avoided roughly $29 million in peak energy costs in July 2024, supplying 717 GWh during peak hours, and — the figure that matters for capacity accreditation — operated at 64% of installed capacity during the most stressed hours of 2024’s seven tightest days, against 67% for gas peakers. During the September 2022 heatwave batteries delivered 6% of demand at the record-peak moment. In ERCOT, batteries supplied 6–9% of electricity in 2025’s most grid-stressed hours, up to 7 GW of fast-responding power. Looking forward, RMI estimates the ~66 GW of storage expected online by 2030 could save more than $400 million a year in health costs if it displaces existing gas peakers, with nationwide battery-related tax collections already exceeding $290 million.

The regulatory-design finding buried in the same analysis is the one to file. A December 5, 2025 ERCOT market-rule change that led many batteries to curtail ancillary-service participation under otherwise-normal conditions added an estimated $3.6 million to wholesale costs over 24 hours. That is a measured price for a rule, not a model — market rules, not physics, stranded flexible capacity, and the cost showed up in a day. Keep it as the counter-citation whenever an eligibility screen, a participation restriction or a narrowed offset definition is proposed on the grounds that it merely tidies up a market: the last time a rule change quietly reduced what batteries could offer, it cost $3.6M in 24 hours.

For the avoided-cost case, the 64%-versus-67% comparison is the strongest available answer to seasonal and performance derates. Storage is demonstrably carrying peak load at peaker-comparable capacity factors in exactly the high-stress hours where DR capacity value concentrates — which reinforces the narrow-peak-hours logic behind the ~$66/kW-year anchor rather than undermining it. The caveats travel with the citation: the health-savings and peaker-displacement figures are contingent on batteries actually displacing gas generation, which the analysis does not verify, and the $29M is a single-month California figure, not an annual national one. (Source: MGRID, November 5 — November 5 entry)

Co-located storage crossed 16% of new utility-scale solar and is planned above 30% — the hybrid fleet is becoming the default, which changes what gets credited. EIA data through September 2026 show over 16% of new U.S. utility-scale solar added in 2026 is co-located with operable battery storage, up from 2% in 2017, with Q4-2026 plans pointing to ~19% and 2027–28 plans above 30%. Solar is now 13% of total U.S. utility-scale capacity, and 18% of that solar is battery-paired. Co-location runs more than double the national average across the Southwest — California, Arizona, New Mexico and Nevada — where hybrid operators direct output to the grid or the battery on price, shifting midday surplus into the evening ramp.

The planning consequence is that the evening-ramp flexibility MISO says is scarce is being built, on the supply side, as a price-responsive fleet. That is the physical substrate beneath the ramp risk flagged in the same week’s winter outlook (below). The valuation consequence is the one to track in dockets: interconnection queues and capacity markets increasingly model hybrids as single resources, which changes how both the solar and the storage components are credited — and a hybrid accredited as one resource can be credited below the sum of its parts. The caveats EIA’s own framing raises: share-of-capacity says nothing about how long the batteries run or whether they charge from the grid rather than the array, the price signal driving co-location may be weaker outside the four Southwest states, and the 2027–28 figures are plans operators revise monthly — the next EIA-860M release on November 24, 2026 tests the 19% near-term projection. (Source: MGRID, October 31 — November 4 entry)

Exelon framed up to ~$8B of regulated storage, and a 1999 restructuring statute may bar the first 500 MW of it — the ownership-model question that decides whether a demand-side alternative can even be offered. On its Q3 earnings call Exelon framed up to roughly $8B of potential regulated energy storage, with officials putting the firmer range at $3–7B for 1.5–3.5 GW, on top of a separate $1B, 500 MW New Jersey proposal. That project — Atlantic City Electric’s four-hour lithium-iron-phosphate “Pittsgrove” battery in Salem County, developed with Invenergy — is pitched as PJM’s largest battery, projecting $1.36 of customer benefits per $1 invested, more than $700M in net customer benefit, and no incremental bill impact for roughly its first five years. The fight is the instructive part: New Jersey’s Division of Rate Counsel, backed by PJM’s independent market monitor, moved to dismiss the petition on the grounds that the state’s 1999 restructuring law bars rate-regulated utilities from owning generation. ACE’s opposition was filed October 30; a Board of Public Utilities decision is expected late in Q1 2027. Exelon’s underlying driver is familiar: high-probability large-load additions holding at 11 GW (9 GW in ComEd, 2 GW Mid-Atlantic) against a pipeline eased to 23 GW from 25 GW, and a $16.3B transmission plan through 2029, with management arguing PJM is too slow to add generation.

This is the venue-analysis lesson from Wisconsin in its positive form, and it should change how a non-wires case is routed. Last week this series argued that where a certificate proceeding is legally barred from weighing alternatives, demand-side resources cannot be offered as a substitute and the case belongs in a different docket. New Jersey is the inverse: a restructuring statute that bars utility generation ownership is a structural opening for third-party and customer-owned DERs, because the capacity has to come from somewhere and the utility is not permitted to build it. In a restructured state, the distributed-capacity procurement pathway is not the harder route — it is the only route available to the wires company, which is exactly the argument the SEIA/Current Energy Group framework was written to make in a docket.

Note the benefit-cost vocabulary, because it is being used against the demand side’s own test. A $1.36-per-$1 benefit-cost ratio and a >$700M net customer benefit are the same avoided-cost arithmetic a DR value stack produces — applied here to a supply-side asset the utility wants to rate-base. When a utility can run that calculation for a 500 MW battery it owns, it can run it for a 500 MW demand-side portfolio it does not, and the asymmetry in which one gets a cost-effectiveness screening is a fair question to put in a filing. (Source: MGRID, November 4 — November 6 entry)


🏗️ Data Centers & Large Load Growth

OpenAI gave Georgia Power the right to curtail 1,000 MW for twenty-five years — and Georgia Power will not say when it can call it or how it will be measured. OpenAI announced “Grid Protectors,” under which Georgia Power holds authorization to curtail up to 1,000 MW at OpenAI’s Effingham County, Georgia campus during peak demand or local emergencies — a right embedded in a 25-year service agreement approved by the Georgia PSC in August after a July announcement, with power delivery expected in phases from 2028 to 2032. The curtailable 1,000 MW is roughly 31% of the project’s expected ~3,200 MW requirement. OpenAI pays the full cost of the grid upgrades and transmission lines needed to serve the site, and pledges to ride through curtailments using on-site generation, batteries and power controls. OpenAI equates 1,000 MW to the average demand of roughly 700,000 households and says plainly that flexibility “cannot replace investment in new energy supply and transmission.” OpenAI says it wants to work with utilities, regulators and policymakers to set how much a campus can reduce, how fast it responds and how long it can sustain the reduction, with terms varying by location; no additional agreements have been announced. OpenAI cited Georgia Power’s estimate of ~$950 million in annual customer savings beginning in 2029 — a figure that covers the utility’s broader large-load contract portfolio, not the OpenAI contract alone.

The scale is real and the instrument is the right one. The disclosure is the problem, and it is not incidental. Georgia Power declined to disclose when it may call the resource or how the 1,000 MW will be measured and verified. No response time was specified. No compensation to OpenAI for curtailment is described — the consideration appears to be service itself, plus the infrastructure cost OpenAI is absorbing. So the record contains a quantity, a term and a counterparty, and none of the three terms this series has insisted belong in every large-load flexibility agreement: when the ceiling applies, how response is measured at the grid connection including repeated calls and post-event rebound, and what fallback protects equipment during an event.

The independent analysts quoted on it converge on exactly that gap, and their list is the better procurement checklist. Neil Osnato of Persistence Analytics Group names the attributes a utility must actually assess — response time, duration, availability, repeatability and independent verification — and makes three technical points worth writing down: AI training is easier to pause than inference; cooling and networking loads limit what can be reduced at the meter; and workload rebound after curtailment must be accounted for. His recommendation is the one to copy into a 25-year agreement: periodic testing and revalidation across the term, because stated capacity “is an assumption until its operating performance is demonstrated.” Tricia Martinez of Dapple calls the pledge constructive but conditions its value on the mechanisms, and asks who pays for upgrades, how costs are allocated, and whether commitments can be independently verified — adding that demand flexibility may work poorly for enterprise workloads under long-term service commitments.

For a utility negotiating its own version, the Georgia agreement is a template with one clause missing and one worth copying exactly. Worth copying: the customer pays the full cost of upgrades and transmission, and future deals are to include long-term contracts and financial assurances protecting other customers if a project is delayed, reduced or cancelled. That is cost causation and stranded-cost protection volunteered by the load, which is the posture every large-load tariff proceeding in the country is currently fighting to compel. Missing: a published measurement and verification method. Until one exists, 1,000 MW is a contractual right, not accredited capacity — and the distinction matters because a utility that counts it as capacity in a resource plan and cannot verify it has moved the risk onto its own reliability case. Ask for the M&V appendix before the megawatts enter a load forecast. (Sources: TechTarget, November 4 — November 6 entry; MGRID, November 6 — November 6 entry. The two are consolidated as one item; the TechTarget reporting supplies the ~3,200 MW denominator, the 2028–2032 phasing, the verification refusal and the analyst assessments that the MGRID summary does not carry.)

AEP raised its capital plan 28% to $100B and its contracted large load 22% to 84 GW — with no demand response, DER or storage named in the plan. American Electric Power raised its five-year capital plan 28% to $100B through 2031 from $78B, split roughly $37B transmission, $36B generation, $22B distribution, and lifted contracted large-load commitments 22% to 84 GW by 2031 from 69 GW the prior quarter, with 15 GW signed in the quarter, counted only where a fully executed service agreement or letter of agreement exists. Gas generation spending rises to $20B from $13B for about 9 GW of gas-fired capacity through 2032, dwarfing $8B of renewables, while the transmission rate base nearly doubles to ~$70B within a $162B total rate base. Drivers include SB Energy’s 10.4 GW Piketon, Ohio campus, served by a ~$5B 765-kV line SoftBank affiliates will fund, and a $4B, 702 MW Bloom Energy fuel-cell project under a 20-year Google PPA. AEP projects up to $21B in take-or-pay fixed-cost offsets for existing residential customers, up from $16B. Q3 operating EPS was $2.23, retail sales +11%.

This closes a tracked item and the answer is a clean negative. This series flagged AEP’s October 28 Q3 and capital-plan release as an item to check; the check is now done, and AEP is meeting load growth almost entirely through utility-owned supply-side capital with no demand response, DER or storage cited in the plan. The $21B of take-or-pay residential offsets is the ratepayer-protection logic that large-load tariffs and demand-side alternatives compete to deliver — which makes it the right number to argue against, not with: if take-or-pay terms can produce $21B of fixed-cost offsets, verified flexibility that reduces the build in the first place produces the same protection without the asset. The caveat that limits both the build and the offsets is the same one: 84 GW counts non-binding letters of agreement, so each figure depends on how much of that pipeline signs firm, take-or-pay terms. (Source: MGRID, October 29 — November 3 entry)

FirstEnergy put a $250M-of-transmission-per-GW multiplier on the record, and most of its 26 GW is unsigned. FirstEnergy’s Q3 presentation, filed October 27, projects 2036 data-center demand at ~26,950 MW — 7,880 MW contracted, 19,070 MW pipeline, the pipeline +4.8% from Q2 — with roughly 4 GW more in active negotiation and contracted load now 7.9 GW, nearly double since the start of 2026, where “contracted” means a construction contract or executed electric service agreement. Growth concentrates in Ohio (9,360 MW by 2036) and West Virginia (6,045 MW), though West Virginia is almost entirely speculative pipeline — just 175 MW contracted. The $36B Energize365 capital plan is expected to rise at least 25% (~$4.75B above the current $19B transmission component) on its 2027–2031 roll, guided by a rule of thumb of about $250M of transmission upgrades per GW of data-center load. The $2.7B Maidsville project — ~1,200 MW gas plus 70 MW solar across three projects, ~35% equity-funded, Siemens turbines reserved for 2030 — awaits a West Virginia PSC order expected in Q4 2026 and has a data-center off-taker under negotiation but not signed, alongside a 3 GW combined-cycle RFP. No demand response or DSM programs are discussed; the only distributed-type options named are reciprocating engines, fuel cells and batteries as West Virginia “bridge generation.”

Take the multiplier, because it converts a load number into an avoided-cost number in one step. At ~$250M of transmission per GW, a utility’s own stated arithmetic says a 500 MW large load implies roughly $125M of transmission upgrades — which is the figure a demand-side or non-wires alternative competes against, sourced from the utility rather than from an advocate. Pair it with the disclosure pattern: most of the 2036 growth is unsigned pipeline, and plant is being built ahead of firm off-take. A 1,200 MW plant with an unsigned off-taker and a 19 GW pipeline with no contracts is precisely the forecast-risk profile that makes flexible, cancellable demand-side capacity the lower-regret purchase — and West Virginia’s 175-MW-contracted-against-6,045-MW-forecast ratio is the sharpest version of that argument available this quarter. (Source: MGRID, October 29 — November 3 entry)

Xcel’s $70B plan excludes the large-load service most likely to drive it, and the exclusion is the disclosure. Xcel’s base 2027–2031 capital plan rises 17% to $70B, with roughly $10B of additional spending already identified outside it, split across 9.2 GW of renewables (~$21.7B), 2.3 GW of storage, 3.3 GW of gas, 2,400 miles of new transmission, $4B of wildfire mitigation and about $7B of contracted data-center projects — funded with $26B of new debt and $8B of equity. Crucially, the plan excludes generation from open RFPs and service to future large-load customers, so the headline understates the data-center-driven build to come. The pipeline: 4 GW of new data-center contracts expected before the end of 2027, a 25 GW generation pipeline and 15+ GW of pending RFPs, plus a Google Pine Island, Minnesota data center whose “Clean Energy Accelerator Charge” would fund 1,900 MW of clean resources and, Xcel says, yield ~$1.1B in benefits to NSP-Minnesota customers (decision early 2027). On storage, the first Blue Lake units — 135.5 MW / ~542 MWh, four-hour, 132 units, in service 2027 — anchor a push to more than 735 MW of Minnesota battery storage by end-2027, with Xcel electing to own capacity rather than contract it. Q3 net income rose 83% to $960M, EPS $1.52, weather-adjusted sales +2.5% YTD.

The Clean Energy Accelerator Charge is the item worth tracking by name. A large-load-funded charge that procures 1,900 MW of clean resources and books a benefit to existing customers is a cost-causation instrument designed by a utility rather than imposed by a commission — the same structural move as SRP’s 60% pre-payment requirement and OpenAI’s agreement to fund its own upgrades, and the third such instrument this series has logged in five weeks. Ask whether the charge can fund flexibility and storage rather than only generation, because the design choice is being made now and a charge restricted to supply-side resources writes the demand side out of a mechanism built on demand-side cost causation. Meanwhile the structural caveat stands: Xcel is meeting load growth primarily through utility-owned supply-side assets — storage, gas, transmission — rather than orchestrated demand-side flexibility, which reinforces that DR and DSM must compete on cost-effectiveness against a utility strongly incented toward rate-based capital. (Source: MGRID, October 30 — November 2 entry)

APS projects retail sales doubling by 2041 and behind-the-meter storage rising from 69 MW to 1.2 GW — while rooftop solar adoption runs 30% below forecast. Arizona Public Service’s new resource plan projects peak demand growing ~3.4% annually over 15 years — net of demand-side management — with retail sales doubling from ~34,200 GWh in 2025 to ~80,600 GWh in 2041 and peak exceeding 14.5 GW by 2041, nearly 60% above the current 9.2 GW record, driven by industrial and data-center load. The preferred portfolio leans supply-side: ~2,000 MW of utility-scale solar and ~2,500 MW of storage by 2029 (roughly 6 GW of storage on the system by 2030), ~4,000 MW of gas through 2036, and a seven-year extension of the Four Corners coal plant to 2038 that APS says could save customers close to $1B. Vote Solar and the Sierra Club criticized the plan for not adequately considering distributed resources such as rooftop solar and batteries. APS flags a “risk of unrealized loads,” mitigated via minimum-demand and minimum-energy contract terms, collateral and early-exit fees.

The behind-the-meter forecast is the signal, and it describes a different distributed fleet than the one most DER programs were designed for. APS sees BTM storage rising from 69 MW today to 1.2 GW by 2045 — its 2032 figure is roughly 95% above its 2023 view — while BTM solar growth slows, with adoption about 30% below prior forecasts (2.1 → 3.7 GW), and EVs rise from 65,000 to 1.1 million (5.2 TWh). Read together: the distributed mix is tilting from export-oriented solar toward dispatchable customer storage and controllable vehicle load. That is a straightforwardly better resource — it is schedulable, it is bidirectional, and it is exactly what a DERMS exists to orchestrate — but it arrives through a channel most utility DER programs are not built for, because a battery and a vehicle are enrolled, dispatched and settled differently from an exporting array. A utility whose DER roadmap still assumes solar interconnection as the dominant growth path is planning for the resource that is decelerating.

And note where the DSM assumption sits: inside the growth rate. A 3.4%/yr peak growth figure stated net of demand-side management means the DSM contribution is embedded and invisible — ask for the gross number and the DSM delta separately, because that difference is the only place in the filing where the demand-side program’s value is quantified, and a plan that netted it out cannot be challenged on it. (Source: Utility Dive, November 4 — November 6 entry)


📋 Regulatory & Policy

Florida’s PSC rejected the state’s first SB 484 large-load tariff 5-0, and the operative precedent is procedural: what a cost-of-service showing has to contain. The Florida Public Service Commission voted 5-0 on November 3 to reject Duke Energy Florida’s proposed large-load tariff — the first compliance filing tested under SB 484, the May 2026 law directing utilities to “reasonably ensure” that hyperscale and other large loads cover their full cost of service without shifting those costs, or the risk of nonpayment, onto other ratepayers, and to create a separate large-load rate class. Commissioners found Duke had not demonstrated compliance. PSC staff had recommended denial about two weeks earlier. At September hearings Duke declined to make public the full cost figures for residential and small-business ratepayers, while disclosing that data-center developers had approached it about warehouse loads requiring up to 1,400 MW. Chair Gabriella Passidomo Smith framed the decision as reinforcing ratepayer protections amid Florida’s load growth and said she is confident Duke can craft a compliant tariff; Duke said it remains focused on supporting economic growth and ensuring data centers pay their full cost of service, and plans a rate-increase petition in the new year.

This closes the single most time-sensitive item on this series’ board, and it closed in the demand side’s favor. The November 3 vote in Docket 20260064-EI was carried as an imminent deadline for three consecutive weeks. The outcome is the stronger of the two possibilities: not a narrow remand on one term, but a unanimous finding that the cost-of-service showing itself was inadequate. What makes it usable elsewhere is the specific failure — a utility that would not put the ratepayer cost figures on the public record did not clear the bar. That is a transparency holding dressed as a cost-allocation holding, and it is portable to any jurisdiction with a cost-causation statute: the obligation to “reasonably ensure” full cost recovery implies an obligation to show the arithmetic.

Two things to carry forward. First, the capacity arithmetic: an unmitigated 1,400 MW interconnection implies well over $100M of network-upgrade exposure on FirstEnergy’s own $250M/GW multiplier — which is the deferral opportunity, quantified from the utility’s disclosure rather than an advocate’s model. Second, the open question staff raised and the Commission did not resolve: the tariff improperly excluded non-firm loads above 50 MW. Whether Duke’s refiling brings the interruptible class inside the framework is the thing to watch, because a large-load tariff that covers only firm service writes demand flexibility out of the rate design entirely — and Tampa Electric’s own tariff is next in line behind it. (Source: Florida Phoenix, November 3 — November 4 entry)

PJM’s qualifying-offset criteria finally surfaced — and they answer this series’ sharpest open question in the negative. Since October 29 this series has carried, as its most consequential unresolved item, the question of what PJM’s amended Reliability Backstop Procurement actually requires for demand response and DERs to count as qualifying offsets — the filing was reported to narrow eligibility, specifying when resources count “rather than treating every behind-the-meter battery as automatically eligible,” and no public source stated the criteria. A targeted search this week located them in PJM’s own draft manual language for Reliability Backstop Procurement, produced through the CIFP-RBP stakeholder process. They are specific, and they are restrictive.

The offset route (Qualifying RBP Offset UCAP MW). Eligible resources are those reasonably expected to be an Annual Demand Resource or DER Capacity Aggregation Resource, registered with PJM and “under a contract to perform based on PJM dispatch instructions,” no later than June 1, 2032; planned DR and planned DER resources under RAA Schedule 6 and 6.2 also qualify. The evidence route is the binding constraint: a Curtailment Service Provider or DER Aggregator must hold a signed commitment with new load that itself has an executed electric or transmission service agreement, the new load’s MW must equal the MW committed, and the commitment must either enroll that load as an ADR or DER Capacity Aggregation Resource from the 2028/2029 Second Incremental Auction through 2032/2033, or purchase replacement eligible generation or storage UCAP. Offsets are capped at each zone’s Load Adjustment MW (2028 minus 2026) for zones not opted out, converted to UCAP at the 0.9401 Forecast Pool Requirement, and valued using 2028/2029 BRA ELCC Class Ratings or an estimated unit-specific value. Evidence was due September 30 to October 21, 2026, with decisions by November 4, 2026 — inside this week.

The supply-offer route, and the answer to the dormant-site question. For Demand Resources offering into the procurement, the draft language states the offer must not rely on locations that supported a Demand Resource in a prior Delivery Year, unless the increase comes from new generation or storage investment at that location. Gating criteria (Section 4.1) add: a site-by-site location table by zone and EDC territory with ELCC class; EDC coordination — for DER, an interconnection application or pre-application coordination plus a distribution study plan; for DR, an executed EDC agreement covering metering, telemetry and baseline measurement; a minimum three years of registration history in PJM or an equivalent market, with material shortfalls in prior Performance Assessment Intervals counted adverse in Stage 2; a deployment schedule consistent with commercial operation date; binding site-level commitments for the full 15-year term, with a substantial majority of offered MW backed by executed customer agreements at offer submission; and an equipment procurement plan in which letters of intent are not accepted.

So the two questions this series has carried since October 1 now have answers, and they diverge. Dormant enrolled DR sites: no. A location that supported a Demand Resource in a prior delivery year is excluded unless new generation or storage was installed there — the October 1 concession allowing previously-enrolled sites that sat out a year to count as “new” does not survive in this language. Incremental battery additions at existing DR sites: yes — that is precisely the exception the exclusion carves out. The practical consequence is a reallocation of effort: reactivating idle enrollments is not a path into this procurement; installing storage at enrolled sites is.

Three caveats, stated plainly, because this is draft manual language rather than the amended tariff. It comes from the August 10 stakeholder draft, not the October 29 amended filing, and terms may have moved; FERC has not approved the amended cost-allocation formula, so the next FERC order remains the gating event; and the document sets no detailed M&V methodology — it references metering, telemetry and baseline-measurement coordination with the EDC, and PAI performance history, but baseline methods and verification procedures are not specified. That last gap is where a well-instrumented portfolio has room to argue, and where a poorly-instrumented one will lose. Read the amended ER26-3380 tariff sheets against this language before relying on either. (Source: PJM draft manual language for Reliability Backstop Procurement, CIFP-RBP materials, August 10 — in-window targeted gap-fill)

Movement — PJM’s amended backstop rewrote cost allocation to follow large-load growth, removed the municipal and cooperative opt-out, and will shrink the procurement by qualifying bilateral megawatts. The October 29 refiling itself was reported in full in last week’s digest; what this week’s coverage adds is the allocation mechanics. The amended filing keeps the one-time structure — a 6,831 MW shortfall from the 2028/2029 Base Residual Auction, contracts up to 15 years, a $555/MW-day MW-weighted maximum willingness to pay (~$202,575/MW-year) — but allocates cost to each zone in proportion to its firm large-load growth between the actual 2026 summer peak and the 2028/2029 delivery-year forecast, with electric distribution companies reporting that growth to PJM. It removes the municipal and cooperative opt-out, requires a five-year plan for any Peak Shaving Adjustment opt-out, and provides that the procurement target will be recalculated downward by qualifying megawatts from executed bilateral contracts between large loads and suppliers. Timeline: March 1 launch, March 28 bids close, April 28 final target and supply set, May 10 results.

The sales argument inside the allocation formula is the part to put in front of large customers this quarter. Charges run on large-load growth to the 2028/2029 forecast. A verified flexibility commitment that lowers a large load’s contribution to that forecast lowers its zone’s backstop charge — which converts demand flexibility from a grid-service product into a rate-management instrument the customer buys for its own account, the posture this series has argued is the demand side’s strongest commercial position. And the bilateral channel is explicitly favored: executed large-load/supplier contracts reduce the auction target directly, while the auction itself demands fifteen-year site-level commitments an aggregator’s customer mix cannot honor. Sell bilateral. (Source: MGRID, October 31 — November 3 entry; the October 29 refiling, the $555/MW-day cap, the 6,831-MW shortfall and the auction schedule were all reported in the October 30 digest — carried here as movement for the allocation mechanics, the opt-out removal and the bilateral target reduction)

Movement — the House’s 100 MW cost-allocation bill stalled in the Senate, and public power is opposing the consideration mandate. The House passage of H.R. 9340, the Ratepayer Protection Act, 417-3 was reported in this series at the time of the September 16 floor vote. What is new in this week’s reporting is the aftermath and the mechanics. The bill adds “Standards for Large-Load Customers” to PURPA that utilities must consider — not adopt — covering rates that recover the full incremental generation, transmission and distribution cost of serving campuses with peak demand of at least 100 MW, financial contributions or assurances before those upgrades are built, and payment for stranded costs if a large customer departs. Because PURPA imposes only a consideration duty, final rate decisions stay with state regulators. The Senate effort stalled the day after the House vote over whether federal requirements should go further. The American Public Power Association, representing roughly 200 public power utilities, opposes the mandate, noting many already have large-load rates and terms — the reporting cites Michigan’s DTE requiring large customers to pay for at least 80% of contracted demand. The same coverage carries an MSI Economics finding that only ~5 GW of data-center capacity is actually under construction against 12–16 GW projected for 2026 delivery.

Treat it as a directional signal and use the APPA objection rather than arguing past it. A 417-3 vote establishes that cost causation for large loads is not contested on partisan lines; a stalled Senate companion establishes it is not becoming federal law this Congress. The operative venue remains the state commission — which is where Florida, Pennsylvania and North Carolina all moved this autumn. And APPA’s objection is the more useful fact for a utility already doing this work: if roughly 200 public power utilities consider the standard redundant because they already impose large-load terms, then “we already do this” is the benchmark position, and a utility without such terms is the outlier. The 80%-of-contracted-demand provision is the concrete, copyable version, and it now has a federal bill and a public-power trade association both pointing at it. Carry the 5 GW versus 12–16 GW construction gap as the standing caution: the forecasts driving these cost fights remain the contested object. (Source: MGRID, November 5 — November 5 entry; House passage reported in the September 25 digest — carried here as movement for the Senate stall, the APPA position and the PURPA consideration mechanics)


🏢 Benchmark Hygiene & Carried Context

MISO’s winter outlook shows a 21 GW paper surplus and says plainly that the surplus is not the risk. MISO’s 2026–27 Winter Planning Resource Auction, released October 28, secured about 134.7 GW of capacity offers against a projected ~113.4 GW winter peak — a roughly 21 GW paper surplus — built on about 3.6 GW of new capacity led by natural gas (with coal, wind, solar and battery storage contributing), roughly 1 GW more imports, and 2.3 GW removed by suspensions and retirements. The projected peak sits well above the ~103 GW expected for 2025–26 and above the 108 GW reached during Winter Storm Enzo in January 2025. MISO cautions that offered capacity is not the same as accredited, fueled and weatherized capacity, and flags tighter reserve margins, steep evening ramps — when solar fades and gas, coal and storage must carry load — and multi-day region-wide cold as the real adequacy risks.

Use the framing, not the surplus. The value of flexible load and dispatchable storage concentrates in exactly the evening-ramp and extreme-cold hours MISO names, not in the comfortable annual margin — which is the narrow-peak-hours logic behind the ~$66/kW-year avoided-cost anchor, restated here by the operator. Note what the outlook does not provide: no reserve-margin percentage, no accredited-capacity breakdown and no explicit demand-response contribution, so the surplus’s robustness on the coldest day cannot be judged from the headline figures. And note the correlated risk inside the additions: heavy reliance on gas and imports, which neighbouring regions may also need in a region-wide cold snap. Read it beside the MISO Load Modifying Resource penalty order effective June 1 — the same market is tightening the penalty on demand-side non-performance while naming winter as the hour its own supply stack is least sure of. (Source: MGRID, October 31 — November 4 entry)

Carried benchmarks, cited rather than re-reported. The $325/MW-day 2028/2029 PJM clear (~$118,625/MW-year) against PJM’s own $555/MW-day backstop cap; the 6,831 MW shortfall; FERC’s October 23 MISO LMR penalty order effective June 1; the CPUC’s ~$608M two-year DR bridge funding and its Societal Cost Test exemption; R.21-06-017’s orchestration framework, CAISO-visible telemetry mandate and 2027–2028 Load Flexibility Scenario; SB 913’s June 2028 resource-adequacy deadline and its device-level telemetry precondition; NERC’s October 26 Board adoption of the computational-load standards; the ~8% winter capacity accreditation in Missouri’s Ameren docket; Google–Voltus >1 GW to 2040; SRP’s 60% pre-payment requirement; and the ~$66/kW-year and ~$20.70/MWh avoided-cost anchors.


🔬 EPRI Research Spotlight

No EPRI publication dated inside the window — the seventeenth consecutive week — but EPRI’s own framework is the instrument this week’s largest story is missing. A direct check returned no EPRI publication dated November 2–6. What the check did surface is worth restating precisely, because it is the classification scheme that would resolve the central ambiguity in OpenAI’s commitment. EPRI’s FlexMosaic framework — announced at CERAWeek in March 2026, built with 65+ utilities, system operators, regulators and hyperscalers, and reported in this series through the spring and summer — defines five flexibility classes for large loads: Class A, support during infrequent extreme system stress; Class B, management of daily or weekly demand peaks; Class C, response to long-lasting energy shortages; Class D, response to sudden supply or demand swings; Class E, frequency stabilization. Class membership is determined by notification time, response duration and frequency, and the depth and speed of the response, across three pillars: managed workloads, reduced AI plant energy consumption, and back-up power. EPRI demonstrations have delivered flexibility of up to 40%, and EPRI projects data centers could use up to 17% of U.S. electricity by 2030.

Apply it to the week’s lead item and the gap becomes a specification question rather than a complaint. Georgia Power may curtail OpenAI’s 1,000 MW “during peak grid demand or local emergencies.” Those are two different EPRI classes. Peak demand management is Class B — frequent, scheduled, shallow-to-moderate, with predictable notification. Local emergencies are Class A — rare, deep, fast, unscheduled. A resource designed for one is not a resource for the other: Class B is a load-shifting product priced against energy and demand charges; Class A is an insurance product priced against loss of load. The 25-year agreement apparently covers both, and nobody has said which class governs the notification time, the response duration or the frequency of calls — the three parameters EPRI identifies as determining the class. That is the same disclosure gap the independent analysts named from the other direction when they asked for response time, duration, availability and repeatability.

So the deliverable is a one-page classification, and it is a smaller ask than a verification regime. Require any large-load flexibility commitment to name its FlexMosaic class, and derive the terms from the class rather than negotiating them separately: notification window, sustained duration, maximum call frequency, depth as a percentage of site load, and the measurement point at which depth is assessed. A commitment that spans Class A and Class B should be written as two products with two sets of terms, because a utility that counts a Class A insurance resource in its Class B peak-management plan has double-counted the same megawatts. The framework exists, it was built with sixty-five utilities and hyperscalers in the room, and it is public — which makes “which class is this?” the cheapest question a utility can ask at the negotiating table, and the one that produces the rest of the terms for free. (Source: Utility Dive, June 26, 2026 — EPRI FlexMosaic, a carried framework reported in this series through the spring and summer, applied here to this week’s items; no in-window EPRI publication)


🚩 Utility-Sector Relevance Flags

⚑ 1,000 MW Was Pledged for 25 Years With No Published Measurement Method — Get the M&V Appendix Before the Megawatts Enter a Forecast
Topic: Large-Load Demand Flexibility / Verification & Accreditation
Relevance: Georgia Power holds the right to curtail up to 1,000 MW at OpenAI’s Effingham County campus — roughly 31% of a projected ~3,200 MW site — under a 25-year agreement approved by the Georgia PSC in August, with power phasing in 2028 to 2032 and OpenAI paying the full cost of grid upgrades and transmission. Georgia Power declined to disclose when it may call the resource or how the 1,000 MW will be measured and verified; no response time was specified; no curtailment compensation is described. OpenAI states flexibility “cannot replace investment in new energy supply and transmission” and cites Georgia Power’s ~$950M annual customer savings from 2029 — a portfolio-wide figure, not this contract’s. Independent analysts name the missing terms: response time, duration, availability, repeatability, independent verification, plus cooling and networking loads limiting meter-level reduction and post-event workload rebound.
Action Signal: Implement — This is the largest flexibility commitment on the North American record and it is currently unaccreditable, which makes it the best available template for what to demand in your own agreement. Copy the two clauses that are right: the customer pays the full cost of upgrades and transmission, and long-term contracts plus financial assurances protect other customers if the project is delayed, reduced or cancelled. Then write in what is absent: the call window, the measurement point at the grid connection including repeated calls and rebound, the sustained duration, and the equipment-protecting fallback during an event. Add the term no 25-year agreement should omit — periodic testing and revalidation across the full term — because stated capacity “is an assumption until its operating performance is demonstrated.” Do not let 1,000 MW of contractual right enter a resource plan as capacity until an M&V appendix exists; a utility that counts it and cannot verify it has moved the risk onto its own reliability case.

⚑ PJM’s Qualifying-Offset Criteria Are Now Quotable — Dormant DR Sites Are Out, Incremental Storage Is In
Topic: Capacity Procurement Eligibility / Portfolio Qualification
Relevance: PJM’s draft manual language for Reliability Backstop Procurement (CIFP-RBP, August 10) states the criteria the October 29 amended filing was reported to narrow. Offsets: resources reasonably expected to be an Annual Demand Resource or DER Capacity Aggregation Resource, registered with PJM and under contract to perform on PJM dispatch instructions by June 1, 2032; the CSP or DER Aggregator must hold a signed commitment with new load that already has an executed electric or transmission service agreement, with MW matching and enrollment from the 2028/2029 Second Incremental Auction through 2032/2033 (or purchase of replacement generation/storage UCAP). Offsets are capped at each zone’s Load Adjustment MW (2028 minus 2026), converted at the 0.9401 Forecast Pool Requirement and valued at 2028/2029 BRA ELCC Class Ratings; evidence was due September 30–October 21, 2026 with decisions by November 4, 2026. Supply offers: DR may not rely on locations that supported a Demand Resource in a prior Delivery Year unless the increase comes from new generation or storage investment there; plus a site-by-site location table, an executed EDC agreement covering metering, telemetry and baseline measurement, a minimum three years of registration history, binding site-level commitments for the full 15-year term, a substantial majority of offered MW backed by executed customer agreements at submission, and letters of intent not accepted. No detailed M&V methodology is specified.
Action Signal: Implement — Inventory your PJM portfolio against these three screens today: prior-delivery-year locations, registration history, and executed-versus-intended customer agreements. The dormant-site reactivation path this series reported as an October 1 concession does not survive in this language — so stop building a qualification case on idle enrollments and redirect that effort to storage additions at already-enrolled sites, which the exclusion explicitly carves out. Where you hold DR sites with three-plus years of clean PAI history, that history is now a quantifiable qualification asset worth protecting. And because baseline methods and verification procedures are not specified, this is the open term: file your baseline methodology into the record rather than waiting to be assigned one. Caveat the citation properly — draft manual language, not the amended tariff; read the ER26-3380 sheets against it before relying on either.

⚑ Florida Rejected the State’s First SB 484 Tariff 5-0 Because the Utility Would Not Show Its Arithmetic
Topic: Large-Load Cost Allocation / Regulatory Precedent
Relevance: The Florida PSC voted 5-0 on November 3 to reject Duke Energy Florida’s large-load tariff — the first compliance filing under SB 484 (Docket 20260064-EI) — finding Duke had not shown the tariff would “reasonably ensure” each large-load customer bears its full cost of service without shifting costs or nonpayment risk to residential and small-business ratepayers. Staff recommended denial two weeks earlier. At September hearings Duke declined to make public the full ratepayer cost figures, while disclosing data-center interest in loads up to 1,400 MW. Chair Passidomo Smith framed it as reinforcing ratepayer protections; Duke plans a rate-increase petition in the new year. Staff had separately faulted the tariff for improperly excluding non-firm loads above 50 MW. Tampa Electric’s tariff is next.
Action Signal: Engage — The portable holding is evidentiary, not substantive: an obligation to “reasonably ensure” full cost recovery implies an obligation to put the cost arithmetic on the public record. Cite it in any jurisdiction with a cost-causation statute, and cite it against confidentiality designations that hide ratepayer impact. Then watch the one term the Commission did not resolve — whether Duke’s refiling brings non-firm loads above 50 MW inside the framework, because a large-load tariff covering only firm service writes demand flexibility out of the rate design. Carry the capacity arithmetic as the standing deferral argument: 1,400 MW unmitigated implies well over $100M of network-upgrade exposure at FirstEnergy’s own ~$250M/GW multiplier.

⚑ Entergy Is Paying ~$125/kW-Year for Residential Battery Flexibility — and Dropped the Solar Requirement
Topic: VPP Program Design / Incentive Structure
Relevance: FranklinWH launched (October 27) an Entergy Louisiana residential-battery DR program paying $1,000 at enrollment plus $125 per average kW delivered over the year (Visa prepaid card by January 31), with events callable any day, year-round, standing down only for forecast severe storms, and — the eligibility change — no solar required if the battery charges from the grid under its interconnection service agreement. The parallel Alliant Energy (Iowa/Wisconsin) program pays $100/kWh up to $1,500 at enrollment plus $100 per average kW discharged. Entergy projects 155 MW of dispatchable VPP capacity by 2030. Hardware is the 15 kWh aPower 2 LFP system, scalable to 225 kWh; enrollment runs through Derapi into Entergy’s Smart Hours platform.
Action Signal: Implement — Copy the two design choices and interrogate the third. Copy grid-charging eligibility: dropping the solar requirement is the single largest expansion of residential VPP eligibility in any program this series has tracked, and it makes the product a pure load-shifting asset available to any customer willing to buy a battery. Copy the storm stand-down: preserving customer backup when severe weather is forecast removes the objection that kills enrollment and costs nothing in the hours the utility actually needs the resource. Interrogate the per-average-kW-over-the-year payment structure — it rewards availability across the season rather than depth in any single event, which is the opposite of how a capacity market accredits, so model payment and accreditation separately before assuming they move together. And note the price: ~$125/kW-year against the ~$66/kW-year anchor is roughly 1.9×, so if you are defending the lower number, be ready to say whether the gap is deliberate acquisition buy-down or a genuinely higher local avoided cost.

⚑ Only ~3% of Data-Center Capacity Can Shift Load — Sell the Storage, Not the Scheduler
Topic: Large-Load DR Potential / Derate Discipline
Relevance: Ember finds only about 3% of global data-center capacity can realistically shift electricity use by 2030, because ~87% runs cloud, corporate IT and AI inference that is hard to pause, leaving ~13% AI training as the shiftable slice — so load-shifting alone trims U.S. peak demand just ~0.1–0.6% by 2030 (Texas ~0.4%). The worked case: Stargate Abilene and Comanche Circle add ~5.6 GW (averaging ~2.7 GW at 48% utilization), of which load-shifting relieves ~0.55 GW versus the full 2.7 GW (~2% of ERCOT’s 2030 peak) via on-site generation. Currence projects non-firm interconnection rules could create up to 285 GWh of U.S. battery demand by 2035, PJM more than half, with curtailment duration ranging ~1.3 hours (uniform ERCOT) to 5.3–7 hours (last-in-first-curtailed, PJM/SPP).
Action Signal: Implement — Reallocate the large-load flexibility product rather than retreating from it: the 0.55-versus-2.7-GW split says on-site storage and generation deliver roughly five times what workload shifting does at the same site. Derate every claim that does not state which of the two produces the megawatts, and require the split in writing in any large-load flexibility agreement. Then treat the Currence duration range as a sizing specification: a curtailment regime that can hold a load down 5 to 7 hours needs a fundamentally different asset than one calling 1.3 hours, and the rule that sets the duration is written before the battery is sized — so intervene in the interconnection-rule docket, not the procurement.

⚑ CAISO’s Only Flexible Large-Load Product Requires PMU-Grade Monitoring and Remote Disconnect — and Expires
Topic: Flexible Interconnection / DERMS & Telemetry Requirements
Relevance: The CAISO Board approved (October 28) a large-load framework defining a “large load” as ≥50 MW at a single site (kept over PG&E’s proposed 20 MW) with a single flexible path — Flexible Interim Load Interconnection (FILI) — under which a load accepts enforceable, grid-condition-based curtailability until its network upgrades are complete, backed by hourly forecasts, telemetry, PMU-grade monitoring in the TO tariff and CAISO remote disconnect. The permanent flexible option (FLIP) was dropped. Tariff language was due at FERC by November 16 (Docket EL26-71) in response to FERC’s June show-cause order to six operators; study windows compress to 60–90 days, questioned by Microsoft and others. CAISO cites a CEC forecast of just 1.8 GW of in-grid data-center growth by 2030 and 4.7 GW by 2035, against ERCOT’s 191.8 GW of large-load studies.
Action Signal: Engage — Put all four FILI instrumentation requirements into your large-load and DERMS evaluation criteria: hourly forecast submission, continuous telemetry, PMU-grade monitoring and an operator-controlled remote-disconnect path. PMU-grade monitoring at a load connection is a materially higher bar than SCADA polling and it is the instrument that settles whether a curtailment actually happened, at the resolution a disturbance analysis requires. Then read the dropping of FLIP as the strategic fact: with no permanent flexible class, flexibility in CAISO buys speed-to-power and then expires, accruing no accreditation record — so if you want a durable demand-side resource class, it has to be created somewhere other than an interim interconnection product. And note that most cost and operating detail sits in TO tariffs not yet written, which is the venue where the terms a customer experiences get set.

⚑ A Restructuring Statute That Bars Utility Generation Ownership Is an Opening for Distributed Capacity, Not an Obstacle
Topic: Ownership Model / Procurement Pathway
Relevance: Exelon framed up to ~$8B of potential regulated storage (officials’ range $3–7B for 1.5–3.5 GW) plus a separate $1B / 500 MW New Jersey proposal — Atlantic City Electric’s four-hour LFP “Pittsgrove” battery with Invenergy, pitched as PJM’s largest, at $1.36 of benefits per $1 invested and >$700M net customer benefit with no incremental bill impact for ~five years. New Jersey’s Division of Rate Counsel, backed by PJM’s market monitor, moved to dismiss on the grounds that the state’s 1999 restructuring law bars rate-regulated utilities from owning generation. ACE’s opposition was filed October 30; a BPU decision is expected late Q1 2027. Context: high-probability large-load additions holding at 11 GW (9 GW ComEd, 2 GW Mid-Atlantic), pipeline eased to 23 GW, $16.3B transmission through 2029.
Action Signal: Engage — This is last week’s venue-analysis lesson inverted, and it is the more useful half. Where a certificate proceeding cannot weigh alternatives, demand-side resources cannot be offered; where a restructuring statute bars the wires company from owning generation, third-party and customer-owned distributed capacity is not the harder route — it is the only route available to it. Run the ownership screen jurisdiction by jurisdiction and lead with the distributed-capacity pathway in every restructured state, which is precisely what the SEIA/Current Energy Group framework was drafted to support in a docket. And use the arithmetic symmetry: a utility that can produce a $1.36-per-$1 benefit-cost ratio for a 500 MW battery it owns can produce one for a 500 MW demand-side portfolio it does not — ask, in the filing, why only one was screened.

⚑ APS Sees Behind-the-Meter Storage Going 69 MW to 1.2 GW While Rooftop Solar Runs 30% Below Forecast — Your DER Roadmap May Be Planning for the Decelerating Resource
Topic: Behind-the-Meter DER Mix / Program Design
Relevance: APS’s new resource plan projects peak demand ~3.4%/yr over 15 years, net of demand-side management, retail sales doubling 34,200 → 80,600 GWh by 2041, and peak >14.5 GW (~60% above the 9.2 GW record). Preferred additions are supply-side: ~2,000 MW solar and ~2,500 MW storage by 2029 (~6 GW storage by 2030), ~4,000 MW gas through 2036, and Four Corners coal extended to 2038 (~$1B claimed savings). Behind the meter: storage 69 MW → 1.2 GW by 2045 (the 2032 figure ~95% above its 2023 view), solar 2.1 → 3.7 GW with adoption ~30% below prior forecasts, EVs 65,000 → 1.1 million (5.2 TWh). APS flags a “risk of unrealized loads,” mitigated by minimum-demand and minimum-energy terms, collateral and early-exit fees. Vote Solar and the Sierra Club say the plan under-weights distributed resources.
Action Signal: Implement — Re-aim the DER roadmap: the distributed mix is tilting from export-oriented solar toward dispatchable customer storage and controllable vehicle load, and a battery and an EV are enrolled, dispatched and settled differently from an exporting array. A program architecture built around solar interconnection is optimized for the resource that is decelerating. Then do one thing in the filing: ask for the gross peak-growth number and the DSM delta separately. A 3.4%/yr growth rate stated net of DSM embeds the demand-side contribution where it cannot be examined or challenged — and that delta is the only place in the document where the program’s value is quantified.

⚑ A Market Rule Stranded Battery Capacity and Cost $3.6M in 24 Hours — Keep the Receipt
Topic: Market Design / Eligibility Screens
Relevance: RMI estimates California utility-scale batteries avoided ~$29M in July 2024 peak energy costs, supplying 717 GWh in peak hours and operating at 64% of installed capacity across 2024’s seven tightest days against 67% for gas peakers; ERCOT batteries supplied 6–9% of energy in 2025’s most-stressed hours (up to 7 GW). And: a December 5, 2025 ERCOT market-rule change that led many batteries to curtail ancillary-service participation under otherwise-normal conditions added an estimated $3.6M to wholesale costs over 24 hours. Forward: ~66 GW of storage by 2030 could save >$400M/yr in health costs if it displaces gas peakers (contingent, unverified).
Action Signal: Watch — Carry two numbers from this. The 64%-versus-67% comparison is the strongest available empirical answer to a performance or seasonal derate: storage is carrying peak load at peaker-comparable capacity factors in exactly the hours where DR capacity value concentrates, which supports rather than undercuts the narrow-peak ~$66/kW-year logic. And the $3.6M in 24 hours is the counter-citation for every eligibility screen, participation restriction or narrowed offset definition proposed as mere market housekeeping: the last time a rule change quietly reduced what batteries could offer, the cost showed up in a day and someone measured it. Cite it in the PJM offset-criteria comment round.

⚑ New York Is Consolidating Residential Load Control Into One State-Funded Platform — Including a Multifamily Pathway
Topic: DR Program Administration / Acquisition Cost
Relevance: The New York PSC approved modifications across National Grid’s three entities (Brooklyn Union Gas, KeySpan Gas East/KEDLI, Niagara Mohawk/NMPC) — fewer annual program tests, lower enrollment thresholds — and advanced a statewide Excelsior Power Program centralizing enrollment of residential electric and gas direct-load-control devices for both summer and winter load relief, with $25/month first-year bill credits for new smart-thermostat customers (pre-2026–2027-budget enrollees excluded from the enhanced incentive), $33M in state-budget funding, and a Phase 1 plan open for public comment. Multifamily customers are recommended to migrate out of National Grid’s thermostat program into Excelsior. No MW target, enrollment count or cost-effectiveness finding is disclosed.
Action Signal: Watch — Two transferable features and one disqualifying omission. Transferable: a statewide enrollment platform spreads customer-acquisition and DLC administration costs across the state’s load and standardizes M&V in one place, which is the precondition for a defensible aggregate capacity claim; and the winter-plus-summer scope is the half a dual-peaking system usually fails to accredit. Also transferable, and the part to show your own commission: $33M of state budget money rather than a rate-case ask. Note the multifamily pathway — this series flagged a multifamily enrollment channel as a standing requirement, and a statewide platform with one built in is the cheapest version available. Disqualifying for citation: no MW. Treat Excelsior as a design and funding template until Phase 1 produces a capacity number.

⚑ A Managed C&I Portfolio Is Already Dispatchable Before the Batteries Arrive — Ask Who Wins When Objectives Conflict
Topic: C&I Aggregation / DERMS Multi-Objective Dispatch
Relevance: GridPoint will add on-site batteries to its commercial platform, targeting 2 GW across 20,000+ commercial sites (Panda Express, Walgreens, Wendy’s among customers), sold as a subscription like its existing HVAC, refrigeration and lighting management. Value stack: bill optimization, peak-demand reduction, grid services, resilience, with stored energy earning utility and capacity-program payments. Customers save ~20% today, qualifying sites up to +15% from batteries, against commercial rates up 42% over five years across its sites. CEO Adrian De Luca cites falling battery costs and transferable ITC financing; off-the-shelf batteries initially. No pricing, timeline, named utility programs or partners; the 2 GW is a target, not installed capacity.
Action Signal: Watch — The batteries are the announcement; the existing control of thermal, refrigeration and lighting load at 20,000 sites is the resource — and refrigeration in particular is thermal storage the customer already paid for. Adding storage produces multi-objective dispatch across two asset classes behind one meter, simultaneously chasing demand-charge reduction, energy arbitrage and a capacity commitment. That is the hardest function in a DERMS specification and the one most often assumed rather than tested: when bill optimization and a grid-service call conflict, which wins, who decides, and what does the utility see? Require the arbitration logic in writing from any aggregator offering a managed C&I portfolio, and treat a 2 GW target with no named utility program as a pipeline, not a resource.

⚑ Storage Is Enforcing a Utility Import Limit for Eighteen Months Because the Feeders Are Not Ready — and the Recharge Window Is the Binding Constraint
Topic: Flexible Interconnection / Non-Wires Economics
Relevance: Critical Loop and SRG Commercial announced (October 27) a 25 MW temporary power system for a six-building advanced-manufacturing campus near Long Beach: 12 MW / 60 MWh of batteries (two CLB-5100 units per building, Cygnus controllers enforcing the SCE import limit) with ~7 MW firm + 2 MW flexible SCE service and 3.5 MW of normally-disconnected standby generation, running ≥18 months until SCE completes two new feeders. Critical Loop owns and operates the batteries as a managed service that relocates to SRG’s next site; tenant reported as Anduril (~$1B, 1.18M sq ft; first buildings ~February 2027). It runs on a non-export Rule 21 application still under SCE review.
Action Signal: Watch — The control stack is the same one CAISO just mandated, built privately: import-cap enforcement, charge/discharge arbitration against available feeder capacity, and islanding to generation on outage — all OpenADR / IEEE 2030.5 dispatch-and-constraint functions worth naming in an RFP. Carry two engineering facts into any import-limited design: the 60 MWh must recharge over the same ~7 MW feed, so the recharge window, not the discharge rating, sets what the system can repeat day after day; and who bears the cost if the feeders slip past 18 months is unspecified, which is the real risk allocation in every bridge-power deal and belongs in the contract rather than the schedule. What the project establishes for the record: how long a large load now waits for utility upgrades, and what storage is worth per month of that wait.

⚑ AEP Added $22B of Capital and 15 GW of Contracted Load in One Quarter With No DR, DER or Storage in the Plan
Topic: Avoided-Cost Denominator / Capital Deployment
Relevance: AEP raised its five-year capital plan 28% to $100B through 2031 (~$37B transmission, $36B generation, $22B distribution; total rate base to $162B, transmission rate base nearly doubling to ~$70B) and lifted contracted large load 22% to 84 GW by 2031 (15 GW signed in the quarter), counted only with an executed service agreement or letter of agreement. Gas spending rises to $20B from $13B for ~9 GW through 2032 against $8B of renewables. Drivers include SB Energy’s 10.4 GW Piketon campus (~$5B 765-kV line, SoftBank-funded) and a $4B / 702 MW Bloom fuel-cell plant under a 20-year Google PPA. Up to $21B in take-or-pay fixed-cost offsets for existing residential customers, from $16B. No demand response, DER or storage cited. Separately, FirstEnergy projects ~26,950 MW of 2036 data-center demand (7,880 MW contracted, 19,070 MW pipeline) on a rule of thumb of ~$250M of transmission per GW, with West Virginia at 6,045 MW forecast against 175 MW contracted.
Action Signal: Engage — Two numbers do the work here. Use FirstEnergy’s ~$250M/GW multiplier to convert any large load into an avoided-transmission figure sourced from a utility rather than an advocate — 500 MW implies ~$125M. And turn AEP’s $21B of take-or-pay residential offsets around: if take-or-pay terms can produce $21B of ratepayer protection, verified flexibility that reduces the build produces the same protection without the asset. Then press the forecast risk, which both filings disclose: 84 GW counts non-binding letters of agreement; West Virginia’s 6,045 MW rests on 175 MW contracted; and MSI Economics counts only ~5 GW of data-center capacity actually under construction against 12–16 GW projected for 2026 delivery. Plant built ahead of firm off-take is the argument for flexible, cancellable demand-side capacity as the lower-regret purchase — and it is the argument a commission is currently most receptive to.

⚑ MISO Has a 21 GW Paper Surplus and Says the Risk Is the Evening Ramp and Multi-Day Cold
Topic: Capacity Adequacy / Seasonal Value Concentration
Relevance: MISO’s 2026–27 Winter Planning Resource Auction (released October 28) secured ~134.7 GW of capacity offers against a projected ~113.4 GW winter peak — a ~21 GW paper surplus — on ~3.6 GW of new capacity (gas-led), ~1 GW more imports and 2.3 GW suspended or retired. The projected peak sits above ~103 GW (2025–26) and 108 GW (Winter Storm Enzo, January 2025). MISO cautions that offered capacity is not accredited, fueled and weatherized capacity, and names tighter reserve margins, steep evening ramps and multi-day region-wide cold as the real risks. No reserve-margin percentage, accredited-capacity breakdown or demand-response contribution is reported.
Action Signal: Watch — Quote the operator’s own risk list rather than its surplus: flexible load and dispatchable storage earn their value in the evening-ramp and extreme-cold hours MISO names, not in the annual margin — which is the narrow-peak logic behind the ~$66/kW-year anchor, restated by the market. Note the correlated risk inside the additions — gas and imports that neighbouring regions will also want in a region-wide cold snap — and read the whole outlook beside MISO’s LMR non-performance penalties effective June 1: the same market is raising the price of demand-side non-performance while naming winter as the hour its own supply stack is least sure of. That asymmetry is worth putting on the record in the LMR compliance docket.


📌 Sources

November 2, 2026 Entry
– Latitude Media — Report: Only 3% of Data Centers Will Shift Power Demand by 2030 (October 29, 2026) (first appearance; the Ember analysis with Currence battery-demand projections — a distinct source and dataset from the October 28 Forbes data-center-flexibility piece carried in last week’s digest)
– MGRID — Xcel Energy Lays Out a $70 Billion 2027-2031 Capital Plan as the First Blue Lake Battery Units Arrive in Minnesota (October 30, 2026) (first appearance)
– EIN Presswire — PSC Approves National Grid Modifications to Demand Response Programs (October 16, 2026) (first appearance; the statewide Excelsior Power Program)

November 3, 2026 Entry
– MGRID — PJM Refiles Its 6,831 MW Backstop Procurement for a March 1 Launch and Bills Large-Load Growth Since Summer 2026 (October 31, 2026) (movement — the October 29 refiling, the $555/MW-day cap, the 6,831-MW shortfall, the auction schedule and the DR/DER offset clarification were all reported in the October 30 digest from three separate sources; carried here only for the allocation mechanics, the muni/co-op opt-out removal and the bilateral target reduction)
– MGRID — CAISO’s Board Approves a 50 MW Large-Load Framework With One Flexible Interim Service, Due at FERC by November 16 (October 30, 2026) (first appearance for the October 28 board approval; the November 16 FERC compliance deadline arising from FERC’s June show-cause orders has been a standing tracked item, cited in one line in prior digests)
– MGRID — AEP Lifts Its Capital Plan to $100 Billion and Contracted Load to 84 GW, With Gas Spending Rising to $20 Billion (October 29, 2026) (first appearance; closes the tracked item flagging AEP’s October 28 Q3 and capital-plan release — distinct from the October 16 MGRID piece on AEP’s 702 MW Bloom fuel-cell plant carried in the October 23 digest)
– MGRID — FirstEnergy Sees 26 GW of Data Center Demand by 2036, With 7.9 GW Contracted and a $2.7 Billion Plant Awaiting an Order (October 29, 2026) (first appearance for the Q3 presentation; the 26 GW figure appeared as a one-line comparison inside the CAISO article and in the October 16 digest’s DOE filing item — here sourced to FirstEnergy’s own filing with the $250M/GW multiplier)

November 4, 2026 Entry
– Florida Phoenix — Florida Public Service Commission Rejects Duke Energy’s Data Center Plan (November 3, 2026) (first appearance; resolves the November 3 vote in Docket 20260064-EI carried as an imminent deadline for three consecutive weeks — the October 22 staff recommendation was reported in the October 30 digest, this is the decision)
– MGRID — Critical Loop Will Bridge a Long Beach Manufacturing Campus With 12 MW and 60 MWh of Batteries on a 7 MW Grid Feed (October 31, 2026) (first appearance)
– MGRID — MISO Reports 134.7 GW of Winter Capacity Offers Against a 113.4 GW Peak, With 3.6 GW Added and 2.3 GW Retired (October 31, 2026) (first appearance; distinct from the October 23 FERC order on MISO LMR penalties carried in last week’s digest — different proceeding, different source)
– MGRID — EIA Says 16% of New Utility-Scale Solar This Year Came With Co-Located Batteries, Up From 2% in 2017, and Plans Top 30% (October 31, 2026) (first appearance)

November 5, 2026 Entry
– MGRID — FranklinWH Expands Virtual Power Plant Reach With Entergy Louisiana and Alliant Energy (November 5, 2026) (first appearance; the week’s VPP-deployment anchor)
– MGRID — House Passes 100 MW Data Center Cost Bill, Leaving Final Rate Decisions to States (November 5, 2026) (movement — the September 16 House passage of H.R. 9340 was reported in the September 25 digest; carried here for the Senate stall, the APPA opposition, the PURPA consideration mechanics and the MSI Economics construction figure)
– MGRID — RMI Estimates Battery Dispatch Avoided $29 Million in California Peak Energy Costs (November 5, 2026) (first appearance)
– MGRID — GridPoint Targets 2 GW of Commercial Battery Storage With a Subscription Offering (November 5, 2026) (first appearance)

November 6, 2026 Entry
– TechTarget — OpenAI Takes Data Center Power Flexibility Pledge National (November 4, 2026) (first appearance; in-window gap-fill consolidated with the MGRID item below as one story — supplies the ~3,200 MW denominator, the 2028–2032 phasing, Georgia Power’s refusal to disclose call and verification terms, the portfolio-wide scope of the $950M savings figure, and the Osnato and Martinez assessments)
– MGRID — OpenAI Agrees to 1,000 Megawatt Power Curtailment in Georgia Under New Grid Protectors Program (November 6, 2026) (first appearance; consolidated with the TechTarget piece above — same development, two sources, reported as one item)
– MGRID — Exelon Sees Up to $8 Billion in Utility Storage as Its New Jersey Project Faces a Dismissal Bid (November 4, 2026) (first appearance; the Exelon load-forecast haircut cited in the October 16 digest is a different item from a different source)
– Utility Dive — APS Expects Sales to Double by 2041, Plans Extending Coal Use to 2038 (November 4, 2026) (first appearance for the IRP filing; distinct from the October 26 KJZZ Arizona load-growth report carried in last week’s digest, which covered SRP and APS reserve margins and the refund dispute rather than the resource plan)
– MGRID — USACE Awards Nearly $27 Million for White Sands Solar and Storage Microgrid (November 5, 2026) (first appearance; carried for the microgrid-controller architecture rather than a demand-side business case)

In-Window Targeted Gap-Fills
– PJM — DRAFT Manual Language for Reliability Backstop Procurement, CIFP-RBP materials (August 10, 2026) (first appearance; resolves the sharpest open item in the series — the qualifying-offset eligibility criteria, the prior-delivery-year DR location exclusion and its new-generation/storage carve-out, the three-year registration-history minimum, the 15-year binding site-level commitment, and the absence of a specified M&V methodology. DRAFT stakeholder language, not the October 29 amended tariff)
– Utility Dive — Data Centers’ Flexibility Could Speed Utilities’ Time to Power (June 26, 2026) (EPRI FlexMosaic framework — a carried framework reported in this series through the spring and summer, retrieved this week to supply the five flexibility classes applied to the OpenAI commitment in the EPRI Spotlight; not an in-window publication)


Method: All five weekday entries — Monday November 2 through Friday November 6 — are present in the research log; no days are missing. The window is strict Monday–Friday and no item from a prior week is reported as new. Items are reported on the day they entered the log with their original publication dates shown; several carry pre-week publication dates inside the daily entries’ trailing-48-hour windows (the Cornell and National Grid items of October 15–16, the Latitude Media Ember analysis and the AEP and FirstEnergy pieces of October 29, the Xcel and CAISO pieces of October 30, and the Critical Loop, MISO, EIA and PJM-refile pieces of October 31). Cross-week deduplication, which is where this week’s hard cases were: Cornell’s Curtailment Credit Market appeared in the November 2 daily entry but was reported in full in the October 16 digest, with its own relevance flag, so it is carried as a benchmark in the Data Centers framing and NOT re-reported — the daily log’s within-log dedupe notes are window-local and do not catch this; PJM’s October 29 backstop refiling was reported in the October 30 digest from three separate sources, so the MGRID piece in the November 3 entry is movement for the allocation mechanics, the opt-out removal and the bilateral target reduction only; House passage of H.R. 9340 was reported in the September 25 digest, so the November 5 coverage is movement for the Senate stall, the APPA opposition and the MSI construction figure; the Florida PSC item is the DECISION on a vote whose staff recommendation was reported last week, which is movement in the strongest form — the resolved outcome of a tracked deadline; FirstEnergy’s 26 GW figure appeared as a one-line comparison inside last week’s reporting and in the October 16 digest’s DOE item, and is sourced here to FirstEnergy’s own Q3 filing for the $250M/GW multiplier; the APS item is distinct from the October 26 KJZZ Arizona report carried last week, which covered SRP and APS reserve margins and the refund dispute rather than this resource-plan filing; and Exelon’s load-forecast haircut cited in the October 16 digest is a different item from a different source than this week’s storage and New Jersey ownership item. Carried benchmarks cited in one line each rather than re-reported: the $325/MW-day PJM clear, the 6,831-MW shortfall, the October 23 MISO LMR penalty order, the CPUC’s ~$608M DR bridge funding and Societal-Cost-Test exemption, R.21-06-017’s orchestration framework and Load Flexibility Scenario, SB 913’s June 2028 deadline and device-level telemetry precondition, NERC’s October 26 Board adoption of the computational-load standards, the ~8% Missouri winter accreditation, Google–Voltus >1 GW, SRP’s 60% pre-payment, and the ~$66/kW-year and ~$20.70/MWh avoided-cost anchors. Zero source URLs overlap last week’s Sources section. Within-log deduplication: the OpenAI Grid Protectors story is consolidated from two sources into one item (deduplicated), with the TechTarget reporting carrying the material the MGRID summary omits; no other development appeared on more than one day. The DERMS and ADMS searches in all five daily entries again returned only DistribuTECH and IEEE PES T&D 2026 session listings and 2025–2036 market-sizing reports — including a November 5 GlobeNewswire release pegging the DERMS market at $5.55B, excluded as marketing — none newly published, the sixth consecutive week with no standalone DERMS vendor or deployment item, so the DERMS section is built on CAISO’s FILI instrumentation requirements, the Critical Loop import-limit controls and the White Sands microgrid controller instead. MGRID’s homepage and dated archive pages were again unreachable at run time on several days, though the RSS feed and individual posts resolved. Targeted search was used only to verify or fill clear gaps, and it resolved two items of consequence. First, and most important: PJM’s qualifying-offset criteria, the sharpest unanswered question in this series since October 29, were located in PJM’s own draft manual language from the CIFP-RBP stakeholder process — the offset evidence route requiring a signed commitment with new load holding an executed service agreement, the Load-Adjustment-MW cap at the 0.9401 Forecast Pool Requirement, the 2028/2029 BRA ELCC crediting basis, the September 30–October 21 evidence window with decisions due November 4, and, answering the two questions carried since October 1, the exclusion of locations that supported a Demand Resource in a prior delivery year unless the increase comes from new generation or storage investment there, which means dormant enrolled DR sites do NOT qualify while incremental battery additions at existing sites DO. The caveat is material and is stated in the body: this is August 10 draft stakeholder language rather than the October 29 amended tariff, FERC has not approved the amended cost-allocation formula, and no M&V methodology is specified. Second, the EPRI direct check found no publication dated inside the window — the seventeenth consecutive week — but retrieved the FlexMosaic five-class framework, which the EPRI Spotlight applies to the OpenAI commitment to show that “peak grid demand or local emergencies” spans two different flexibility classes with different notification, duration and frequency parameters, and that no party has said which governs. What targeted search did not resolve: NERC’s energy assurance standard remains unestablished after a second run, with searches returning only the computational-load standards track and no dedicated project, docket or timeline, so whether demand-side resources are in scope is still unknown — next attempt should be NERC Standards Committee agendas and the project tracker rather than open search; NERC’s early-November FERC filing of the computational-load standards and registry criteria was not confirmed as filed; no coverage of the November 4 CEC interoperability workshop output was found; and PJM’s deployed demand-response megawatts for the 2026 shoulder-season events remain unpublished. The Pennsylvania Bulletin publication date that starts the comment clock on the PA PUC’s October 1 tentative order in Docket M-2026-3064961 was not pursued this week and remains unverified after four runs; pacodeandbulletin.gov refuses programmatic retrieval and no secondary source states it, so it needs a manual check or a call to the PUC Secretary’s Bureau.