This week: the demand side got its largest private purchase order and its clearest institutional snub in the same seventy-two hours. On October 22 it emerged that Google will fund roughly 1,070 MW of additional Voltus virtual-power-plant capacity across PJM’s 13 states, under a bring-your-own-capacity contract running through 2040 — roughly ten times the 100-MW Google/Voltus deal announced in June, and the first gigawatt-class, decade-dated private contract for distributed demand flexibility this series has recorded. Voltus also reported delivering about 105% of committed load reduction during two PJM emergency events, which is the first performance figure of any kind attached to PJM’s October emergencies. Then on October 21, NERC CEO Jim Robb opened FERC’s 2026 Reliability Technical Conference with six top risks and four remedies — and demand response, virtual power plants and load flexibility appear nowhere in the remedy list, while DERs appear only as an integration risk.
The third fact of the week sits between those two and explains both. Reporting on October 21 established that PJM has activated demand response on nine days in 2026 through October 15 — more than the eight days recorded across all of 2015–2024 combined, and fifteen days across 2025 and 2026 together. The resource is being called at an order-of-magnitude higher frequency than the decade that preceded it, a hyperscaler is now buying it on PPA-like terms to 2040, and the continent’s reliability organization still does not list it among the things that fix the problem.
So the week’s argument is no longer about whether flexibility is dispatchable — October answered that nine times. It is about credit. Robb’s proposed energy assurance standard — dependable energy under stress, not nameplate — is the door every DR and VPP accreditation claim now has to pass through, and his explicit skepticism that batteries and solar “contribute reliably in some markets, especially in winter” is the form the challenge will take. Meanwhile the private market has stopped waiting for the accreditation question to be settled: Google’s 2040 contract, Base Power’s first cooperative deal outside Texas, Budderfly’s $500M municipal offer and ev.energy’s 79,000-driver California scale-up are all demand-side capacity being procured outside the IRP process it is still arguing its way into.
⚡ Virtual Power Plants & Demand Flexibility
Google will fund roughly 1,070 MW of Voltus VPP capacity in PJM through 2040 — ten times its June deal, and the longest-dated private contract for distributed flexibility on this series’ record. Google is funding approximately 1,070 MW of additional Voltus virtual-power-plant capacity across PJM’s 13-state footprint plus the District of Columbia, under a bring-your-own-capacity (BYOC) structure in which Voltus enrolls commercial, industrial and residential participants that curtail load or discharge batteries when the grid is strained, and Google pays to offset the capacity its data centers draw. The contract runs through 2040. No price was disclosed, and Voltus says full capacity will take three to four years to reach. Participants can earn up to $350,000/MW-year — that is the participant-facing figure, not Google’s contract price.
Three things make this materially different from the 100-MW June deal rather than a bigger version of it.
First, the term is the product. A contract to 2040 is the same instrument that financed utility-scale wind and solar: it lets a battery owner or a large C&I participant sign without taking merchant capacity-price risk. This series has argued for two months that program duration is a storage-procurement instrument — New York’s three-year DLM minimum, the one-year DR programs that structurally exclude storage. A fourteen-year private offtake is that argument taken to its limit, and it arrives from a customer rather than a commission.
Second, there is finally a performance number attached to a PJM emergency — and it did not come from PJM. Voltus reports delivering approximately 105% of committed load reduction across two PJM emergency events in the week before the announcement, while 46 GW of generation was in planned outage and 15 GW was unexpectedly unavailable. This series has carried an open item since September noting that PJM has published no deployed DR megawatts and no performance results for any of its 2026 shoulder-season events. That gap is still open at the operator level, but an aggregator has now put a verified-performance claim on the record for the same windows. Treat it accordingly: it is company-reported, preliminary, and expressed as a percentage of commitment rather than in megawatts — which is precisely the form that cannot be audited from outside. Use it as evidence that the dispatch happened and cleared, not as a settled accreditation figure.
Third, the scale context is the sell. Voltus holds ~8.5–9 GW under contract in North America against a 20-GW-by-2030 target, following a $225M Series D in October led by Generation IM. North American VPP capacity totals roughly 37.5 GW across about 2,000 programs (Wood Mackenzie), so this single ~1-GW commitment is nearly 3% of the installed continental base. It lands into a PJM forecasting 30–34 GW of new large load by the early 2030s, where DR revenue rose 240% from 2024 to 2025 to $523.5M (Monitoring Analytics), and where FERC in early October struck the rule that would have forced large loads to bring their own capacity or be curtailed.
The caveats that must travel with any citation: the 1,070 MW is a target, not a delivered resource; no price is disclosed, so it cannot be benchmarked against PJM capacity clears; and it is unresolved whether PJM counts the capacity as delivered when called. (Source: MGRID, October 22 — October 23 entry)
PJM called demand response on nine days in 2026 — more than the prior decade combined. Through October 15, 2026, PJM had activated demand response on nine days, against eight days across all of 2015–2024, and fifteen days across 2025 and 2026 together. PJM’s executive director of system operations, David Souder, is the attributed source. Context in the same reporting: roughly 11,000 MW of connected data-center load as of October 15, about 10,000 MW of generation returned to service during the October operating period with more than 61,000 MW still offline that week, and more than two-thirds of those outages planned work or construction. PJM confirmed it did not use the DOE authority to direct data centers onto backup generation.
This is the frequency statistic this series has needed, and it changes the shape of the argument. A resource called nine times in ten months is not a contingency instrument; it is part of normal operations. Three practical consequences: dispatch-cost assumptions built on one or two annual calls understate program cost; customer-facing enrollment terms written for a rare-event resource (opt-out provisions, event caps, notification windows) will bind more often than they were designed to; and the accreditation case gets easier on frequency and harder on depth — more calls mean more observations, and more observations mean the performance question moves from “will it respond” to “how much, verified, each time.” The gap remains exactly where it was: still no published deployed megawatts. Chase PJM’s Operating Committee and Demand Response Subcommittee materials, not press releases. (Source: EconoTimes, October 21 — targeted gap-fill on a standing open item)
Base Power crossed a state line with a cooperative-dispatched residential VPP — and the dispatch right sits with the utility. Base Power will install 50 MW across roughly 2,500 member homes in Central Georgia EMC territory, forming a networked fleet the cooperative controls — charging off-peak and discharging to shave system peaks, with the co-op, not the homeowner, holding charge/discharge authority. Member terms are the enrollment engine: $95 up-front for automatic whole-home backup, no monthly fee, Base installs and maintains the battery for its life, and members pay standard co-op rates for their own usage. First installations are scheduled for 2027. This is Base’s first utility partnership outside Texas, following an August $1B Series D; its overall fleet now exceeds 750 MWh across 30,000+ homes (all deployments, not just this program).
The structural point for an avoided-cost filing is the ownership of the dispatch right. This series has repeatedly flagged that a VPP’s creditability turns less on the hardware than on who can call it: a distribution cooperative holding the dispatch right is the cleanest case for crediting distributed capacity as a planning resource, because the utility is not relying on a third party’s discretion or customer mix. It is also a reminder that public power and cooperatives are moving faster on this than most IOUs — the same pattern as FMPA specifying GIS prerequisites at the joint-action-agency level. (Source: MGRID, October 18 — October 20 entry)
California is subsidizing enrollment rather than hardware: ev.energy takes its managed-charging VPP from 9,000 to a target 79,000 drivers. ev.energy secured $11M from the California Energy Commission for Phase 2 of its ChargeWise program under the CEC’s REDWDS initiative (Responsive, Easy Charging Products with Dynamic Signals), with partner match bringing total budget to about $22M. The mechanism matters more than the dollars: most of the award is match funding covering 50% of user-acquisition, incentive and onboarding costs for utilities, Community Choice Aggregators and other partners, with about $1M for product development including V2X. The program targets growth from a 9,000-driver Phase 1 pilot to 79,000 customers by end-2028, using day-ahead CAISO price and supply signals to shift charging off the 4–9 p.m. peak into midday solar hours. Phase 1 reported up to 98% of charging delivered off-peak and up to 30% shifted to midday. Partners include CCAs MCE, Silicon Valley Clean Energy and Clean Power Alliance.
Read this directly against Brattle’s 4% benchmark from last week, which named OEM-led enrollment — activating capability already installed — as the single lever that takes national residential VPP capability from ~10 GW to 35 GW. The CEC has now funded the other half of that mechanism: by paying down customer-acquisition and onboarding cost through CCAs and IOUs rather than buying devices, it attacks the binding constraint rather than the visible one. That is a program-design template worth copying into any DSM filing that has been losing on cost-per-enrolled-customer. The limitation to state plainly: the release discloses no MW capacity figure, so the firm contribution cannot be sized, and a 98%-off-peak statistic is a load-shape result, not a dispatchable-capacity result. (Source: ev.energy via National Law Review, October 20 — October 22 entry)
Budderfly and Schneider Electric are offering cities $500M to assemble a 150-MW VPP — and have one signed contract. Budderfly, with Schneider Electric supplying the energy-management and automation layer, will make $500M in project capital available to small and mid-sized U.S. cities under a Modernizing American Cities program: solar, battery storage and efficiency upgrades at no upfront cost, repaid out of energy savings rather than grants or capital budgets, with the upgraded sites aggregated into a 150-MW virtual power plant. The target is 100 cities at roughly $4–5M each, with forecast energy savings of 60–80% per city.
The reality check belongs in the same breath as the headline. Budderfly’s only signed municipal deal is a $4.4M, 15-year agreement with Cadillac, Michigan — seven facilities including city hall and the wastewater plant, covering solar, storage, HVAC and lighting, EV charging and microgrid backup, with the city taking a share of demand-response revenue and owning the equipment at term-end (council waived competitive bidding on July 20; ~65% projected cost cut). And critically: no source names a utility, grid operator or market that would buy the 150-MW VPP’s capacity. The capital figures also conflict across sources (>$378M invested over ten years across 9,000+ C&I sites per pv magazine, versus ~$600M raised per Cadillac council minutes), and no baseline, payment schedule or Cadillac solar/battery rating is disclosed.
Why it still belongs in a procurement file: municipal portfolios are a distinct, under-enrolled site class with a decision-maker who is a single counterparty rather than thousands of households — wastewater plants, city halls and pumping stations are large, predictable, controllable loads. Treat the 150 MW as a projection and the Cadillac contract as the only evidence, but take the channel seriously. (Source: MGRID, October 21 — October 23 entry)
Movement — the October 6 California record reconciles at 745 MW, of which Tesla was roughly 601 MW. This series reported the October 6 California residential-battery dispatch as a 601-MW, 85,774-home Tesla figure. Reporting that entered the log this week gives the combined Sunrun-and-Tesla total for the same event as 745 MW, up from 669 MW the prior night and roughly 28% above the September 9 record of 580 MW (prior mark 539 MW, July 2025). Tesla’s 85,000+ Powerwalls supplied 600+ MW, estimated at about 80% of the record; Sunrun reports 120,000+ enrolled batteries but declined to split the total. Calls came through California’s DSGS and ELRP programs, with October 5 the first-ever concurrent call of both across all participating IOU territories, and DSGS triggering only after the CAISO day-ahead LMP exceeded $200/MWh.
This is a reconciliation, not a new record. The two figures are consistent — 601 MW Tesla-only inside a 745 MW combined total — and the series’ standing numbers should now be cited that way, because the fleets overlap and aggregate totals invite double-counting. The caveat is unchanged and still decisive: no MWh, no event duration and no participant payments were disclosed, so peak MW alone cannot establish capacity value, and the resource remains price-triggered rather than firm-scheduled. (Source: MGRID, October 14 — October 19 entry; reconciles the October 6 dispatch reported in a prior digest)
🔌 DERMS & Grid Integration Technology
A utility dispatched a battery inside an occupied apartment building through a named DERMS platform — the first multifamily instance in this series. Viridi and developer Good Carbon Co. installed a 3.75-MWh lithium-ion system across seven battery rooms in Peace Bridge Place, a fully electric, geothermal-heated 105-unit affordable-housing building in Buffalo — about 36 kWh per apartment, described as one of the largest indoor lithium-ion installations in an occupied U.S. residential building.
The control path is the finding. This is a National Grid pilot in which the utility sends dispatch signals through EnergyHub’s Edge DERMS platform to Viridi’s energy-management system. The battery discharges as the building approaches its monthly demand-charge peak and during National Grid peak events, while thermostats in participating apartments pre-cool or pre-heat (residents can opt out). National Grid pays the property both for not importing and for exporting. That is a textbook aggregated-DR value stack — demand-charge management plus peak-event capacity — layered on top of a resilience asset reported to carry several days of outage ride-through.
Two things to take from it and one not to. Take the site class: multifamily buildings are a genuinely new addressable segment for utility DR, and they solve the acquisition-cost problem differently from residential VPPs — one landlord contract reaches 105 meters. Take the architecture: utility → third-party DERMS → building EMS → battery and thermostats is the four-layer control chain a DERMS procurement must actually specify, and it is the chain where latency, telemetry fidelity and command arbitration fail in practice. Do not take the economics: no project cost, no kW power rating, no NYSERDA Future Grid Challenge award figure and no DR revenue were disclosed, so payback cannot be judged, and a 3.75-MWh energy rating without a power rating cannot be converted into a capacity contribution at all. The stronger evidence here is safety — a UL 9540A cell-failure test that did not propagate, and UL 9540 listing — which matters for the siting approvals this site class will need. Supply is the near-term ceiling: Viridi’s Buffalo plant produces roughly 400 MWh/year, about 100 systems this size. (Source: MGRID, October 20 — October 21 entry)
🏗️ Data Centers & Large Load Growth
San Diego shows what happens when a 400-MW load lands somewhere a city cannot zone. Manchester Financial Group is studying a 400-MW data center — potentially California’s largest — as one redevelopment option for the federally owned 70.3-acre NAVWAR / Old Town Campus site, under a lease-for-construction deal giving it a 99-year lease on about 60 acres. Because the land is federal, the project would need no city permits and no congressional appropriation. Rep. Scott Peters, Mayor Todd Gloria and Councilmember Elo-Rivera have pushed back, with a rally and a proposed temporary moratorium on large data centers — but a city moratorium reaches private sites, not the federal parcel, so the hold is political and turns on the Navy’s 2027 environmental review.
For capacity planning the unresolved item is cost: SDG&E says it will upgrade the grid to meet demand wherever it arises but has disclosed no cost or cost-responsibility for 400 MW of continuous load in a service territory with among the highest per-kWh rates in the country — and there is no interconnection request and no load study yet, with the 400-MW figure coming from a single developer email. Treat the number as unverified.
The demand-side hook is forward-looking and worth tracking: San Diego Community Power had a draft large-load-integration guideline on its October 22 agenda. A guideline is not a tariff, and California still handles large loads case-by-case, but a continuous 400-MW load is exactly the setting in which on-site generation, storage and demand flexibility enter an interconnection or special-rate negotiation. The pattern to watch is CCAs writing large-load integration rules in parallel with IOUs — a second venue, with its own timetable, where flexibility obligations can be embedded. Comparison point: a 330-MW Imperial County project is on hold after an August 31 ruling requiring an EIR. (Source: MGRID, October 22 — October 23 entry)
NiSource folded a 600-MW tolled battery into a 20-year Microsoft contract — and Microsoft funds all of it. NIPSCO, through NiSource Generation Company (GenCo), will add about 1,100 MW of new generation under a 20-year term beginning in 2028 to serve a Microsoft data-center campus in northwest Indiana, and has contracted 600 MW of battery storage under a long-term tolling agreement (built by an AGP Group affiliate) that joins GenCo’s pooled resources serving multiple large-load customers. The cost-allocation design is the familiar protection template: GenCo’s large-load customers pay for the generation and infrastructure built to serve them, Microsoft fully funds all generation and transmission for its load, and NiSource states the structure is designed to protect existing NIPSCO customers from incremental costs. NiSource now projects about $1.75B in customer savings over the contracts, up from $1.4B previously disclosed (~$140 per customer annually) — the company’s own figures. Specific tariff terms were deferred to the Q3 earnings call.
Entergy Arkansas’s second Google phase carries 1.4 GWh of batteries — and revives a PPA regulators rejected six weeks earlier. Google will cover about $600M of planned investment plus a PPA for Phase 2 of the West Memphis campus, with service expected 2028 pending state approval: a 1,000-MWh battery at Driver Solar, a 400-MWh battery at Jones Mill, a 440-MW(ac) PPA from the Big Island solar plant, and a Galet substation expansion plus two new 230-kV lines. With the related Arkansas Cypress facility the package adds over 1,000 MW of solar and 2,800 MWh of storage. The regulatory wrinkle is the one to file away: the Arkansas PSC denied the Big Island PPA in early September for failing to show reasonable costs or customer savings against alternatives — and the same 440-MW plant now returns inside the Google deal, “added based on APSC guidance from the initial filing.” Entergy claims ≥$500M in new 20-year customer savings (net $2.1B combined with Phase 1 for 738,000 customers), and the new resources are not charged through the Generating Arkansas Jobs Act Rider. MGRID notes the savings calculation is unexplained.
Read those two together, because they are the same instrument. In both, a hyperscaler’s funding commitment plus pooled or tolled storage is offered as the adequacy answer, and in both the ratepayer-protection claim rests on utility-produced savings math. Two consequences for demand-side practitioners. First, the bar moved: when a large load funds its own 1,100 MW or 1.4 GWh, the marginal-resource question a DR portfolio must win is no longer “cheaper than a peaker” but “cheaper than capacity someone else is paying for.” Second, the batteries are the opening. A 600-MW tolled battery pooled across large-load customers, and 1.4 GWh of utility-side storage, are exactly the flexible assets a DERMS or market-ops layer should be dispatching — and nothing in either disclosure says who orchestrates them, against what signal, or whether their flexibility is credited anywhere. That is a procurement question to ask inside both dockets. The Arkansas re-submission adds a third, sharper point: a customer’s funding commitment can revive a resource that failed a standalone prudence test — which is a prudence standard a demand-side portfolio does not get to borrow. (Sources: MGRID, October 19 and October 20 — October 20 and October 21 entries)
AEP’s $4B fuel-cell plant for Google is the self-supply counterfactual at utility scale. Over October 13–14, AEP agreed to buy up to 1.6 GW of Mainspring linear generators through 2034 — first tranche 190 MW from 2027, roughly 200 MW/year, worth up to ~$4B, Mainspring’s largest order to date, in 2.5-MW modular units running on gas, biogas or hydrogen — and separately plans ~$4B for a 702-MW Bloom Energy fuel-cell plant near Cheyenne, Wyoming, whose output Google takes under a 20-year PPA (via Cheyenne Fuel Cell LLC, AEP ~92% / Basalt Infrastructure ~8%; construction from 2026, full operation targeted end-2028 to catch the investment tax credit). The 702-MW plant is roughly seven times AEP’s prior ~98-MW Ohio fuel-cell deployment and sits inside Google’s planned ~2.7-GW Cheyenne campus, where Black Hills plans up to 590 MW of grid service plus ~2.1 GW of third-party resources in a private microgrid.
Two reasons this is a DER item and not a generation item. These are distributed assets by architecture — modular linear generators and fuel cells, co-located behind or beside the meter — which a utility DERMS must eventually monitor and coordinate even when they operate independently of the distribution grid; this series’ standing argument that distributed flexibility earns avoided-cost value only when dispatched continuously, not when reserved for outages applies to a 702-MW fuel-cell plant as directly as to a microgrid. And the open risk is who pays if the offtake fails: AEP has not said whether the January offtake conditions — a December 2026 milestone disclosed in its 10-Q — are met, nor how costs would be recovered if the single-customer offtake falls through. AEP’s Q3 results and 2027–2031 capital plan were due October 28. (Source: MGRID, October 16 — October 19 entry)
Behind-the-meter gas keeps absorbing load the grid cannot connect: Enerflex’s backlog passes 1.75 GW. Calgary-based Enerflex won an Engineered-Systems award to design, fabricate and assemble more than 1 GW of low-emission, natural-gas-fired behind-the-meter generation for a large, unnamed North American data-center developer. Combined with a ~450-MW award announced October 1, its BTM gas backlog now exceeds 1.75 GW, with deliveries through 2028 and about $100M of authorized Engineered-Systems investment (~$15M in 2026 capex). CEO Paul Mahoney framed the demand as a need for “reliable prime power that doesn’t depend on grid connection.”
The quantity is the argument: 1.75 GW of firm capacity being built entirely outside the regulated planning process, bypassing both the interconnection queue and the demand-side portfolios that would otherwise defer it. For DERMS architects it is also the co-located, grid-independent fleet a control layer must eventually see and, where interconnected, coordinate. Caveat: vendor-disclosed backlog, not operational deployment, with the end customer unnamed — a leading indicator, not an event. (Source: Blackridge Research, October 13 — October 20 entry)
The equipment order book is a demand signal, not yet a delivery story — and that asymmetry raises avoided-T&D cost. ABB reported record Q3 orders of $12,055M (+40% YoY) on revenue of $9,701M (+14%) with a 1.24 book-to-bill, meaning backlog grew. U.S. orders rose 82% (no dollar value given), electrification orders rose 70% to a record $7.67B, Americas electrification was +117%, and the CFO called data centers the quarter’s standout customers with triple-digit order growth. ABB also launched an “Infinitus” DC portfolio for AI data centers including a solid-state transformer slated for commercial availability in H2 2027.
The restraint is the relevant part: ABB is holding capex near $1B with no mid-term increase expected, planning “incremental” capacity plus third-party assembly and a 51% stake in a JV built on Mitsubishi Electric’s low-voltage distribution business (~$495M). If order growth of this magnitude is not matched by manufacturing capacity, lead times for the distribution and substation gear that grid-side capacity additions depend on stay long — which raises the effective avoided-T&D and avoided-capacity cost a DR or VPP program is credited against. The demand-side case gets stronger precisely because the alternative is scarcer and slower. State the limit when citing it: no transformer or switchgear lead times were disclosed, so the inference is directional, not quantified. (Source: MGRID, October 20 — October 21 entry)
📋 Regulatory & Policy
NERC’s reliability remedy list does not include demand response. That is this week’s most consequential fact for anyone building a DER business case. (must-include item; cited from the primary document) NERC President and CEO Jim Robb opened FERC’s 2026 Reliability Technical Conference on October 21 by describing reliability as a “paradox” — still high by historical measure, but with accumulating near-misses he characterized as a “five-alarm fire.” He named six top risks: resource adequacy; unprecedented load growth; the shift to power-electronic generation and load; extreme weather; policies slowing infrastructure development; and cyber and physical security. From the 2025 Long-Term Reliability Assessment: summer peak demand up roughly 225 GW and winter peak up roughly 250 GW over ten years, with more than half the continent at resource-adequacy risk. Most new resources in the five-year horizon are battery and solar, which Robb said “struggle” to contribute reliably in some markets, “especially in winter.” Gas supplies about half of generation fuel, forced-outage rates are rising, and he was explicit that emergencies “should not be normalized.”
His four remedies: better supply-and-demand forecasting, anchored on a new energy assurance standard — dependable energy under stress rather than nameplate capacity; urgent generation and transmission build-out (“being studied doesn’t translate to kWh”); inverter-based-resource integration, with Order 901 work complete but the 2025 Iberian outage signalling more work needed for “large amounts of IBRs and DERs”; and modernization of NERC’s own processes.
Demand response, virtual power plants and load flexibility appear nowhere in that list. DERs appear only as an integration risk.
Do not read that as a reversal of the DER growth thesis — the thesis holds on volume, because the 225-GW/250-GW load growth Robb describes is the adequacy gap that demand-side resources address. Read it as a precise statement of where the fight now is: credit, not need. Three operational consequences.
1. The energy assurance standard is the door. Every accreditation claim a DR or VPP portfolio will make over the next two years has to pass through a standard built around dependable energy under stress. Start now on what your portfolio can demonstrate under that framing — sustained delivery across a multi-hour stress window, not peak-hour response — and track the standard’s development as a primary docket rather than a secondary one.
2. Winter is the exposure. Robb’s skepticism is specifically that batteries and solar struggle to contribute reliably in winter. That bears directly on winter-peaking systems and on any four-hour-duration resource claiming capacity credit in a December morning ramp. Portfolios that have only ever been measured against summer afternoons — and, as this series established last week, not even reliably against September and October evenings — have no winter record to offer. Build one before the standard asks for it.
3. Absence from a remedy list is a participation failure, not a technical verdict. The remedy list reflects who was in the room and what evidence was on the table. The evidence gap is the same one this series has tracked all quarter: grid operators are generating the operational record — nine DR days in PJM, three shoulder-season events, a 745-MW California dispatch — and no one is publishing it in a form a standards body can cite. (Source: NERC — Jim Robb opening statement, FERC 2026 Reliability Technical Conference, October 21 — in-window)
Resolved — NERC’s computational-load standards cleared final ballot, and the Rules of Procedure registry track this series chased for three runs has landed. NERC stakeholders approved new reliability standards for AI data centers, crypto-mining facilities and other computational loads on a final ballot announced October 15, developed under Project 2026-02 and carried in standards referenced as CLO-001-1 (interconnection studies and modeling, with verification and reporting for planning-horizon cases) and CLO-002-1 (operational data and communications, so transmission operators, balancing authorities and reliability coordinators receive load-site data under predefined protocols), plus load-protection coordination and disturbance monitoring. Applicability is the two-part test FERC declined to name: computational loads of 50 MW or more connected at 100 kV or higher. The trigger was a Level 3 NERC alert in May 2026 after data centers unexpectedly dropped or rapidly oscillated load.
And the registry track is now confirmed rather than open. The NERC Board approved computational-load registry criteria and two new functional entity types — Computational Load Owner and Computational Load Operator — and is filing with FERC; the three commissioning standards passed both initial and final ballots and go to the Board on Monday, October 26, with a FERC filing in early November — ahead of the December 31 deadline this series has been tracking. Revise the standing December 5 and December 31 deadline lines accordingly. NERC’s 2027 work program includes ride-through, event reporting, root cause, model validation and security. FERC approved NERC’s 2027 budget.
Why this is a demand-side item. The modeling and telemetry being mandated for 50-MW-plus loads is the same observability a DERMS or market-ops layer needs to treat those loads as dispatchable flexibility — the standards build the measurement substrate for a resource they do not name. Against the Robb statement the pairing is stark: NERC is mandating that large loads be observable while declining to list their flexibility among the remedies. The methodological risk this series has flagged since August is now locked in by the calendar: the standards will be filed before the measured operational data exists, with the only large-scale measured event still the 3,800-MW July 22 trip in which data centers disconnected too early. (Sources: Utility Dive, October 15 — October 22 entry; NERC, October 20–21 — October 23 context)
Movement — FERC’s October 9 rejection also struck the capacity-market demand-curve provision, and that is the part with the largest dollar figure attached. This series reported FERC’s rejection of IRAS and BYOC in full last week. A consumer-advocate analysis that entered the log this week adds a provision the news coverage omitted: FERC also rejected a measure that would have removed data centers lacking their own energy supply from the capacity market’s demand curve — a change the Citizens Utility Board had fought to include. The demand curve sets how much capacity all ratepayers pay to procure, so with both measures struck, CUB argues the burden of data-center-driven capacity costs stays on ordinary consumers, citing the ~$29.4B figure from recent PJM auctions (a carried benchmark in this series, not a new finding). Because PJM’s plan always relied on states to enforce its terms, CUB reads FERC’s decision as making state legislation “the last hope,” and urges Illinois to pass the POWER Act.
The framing is advocacy and should be cited as such. The substantive addition is that the demand-curve treatment of large loads — not just emergency curtailment priority — is a live lever, and it sits in the same place everything else went on October 9: the states. For a multi-jurisdictional utility, state-by-state divergence in DR and DSM program rules is now a planning assumption rather than a risk. (Source: Citizens Utility Board, October 18 — October 19 entry)
California issued the final 2026 Avoided Cost Calculator — and refused to let SDG&E hand back energy-efficiency administration. At its October 8 voting meeting the CPUC issued, in Resolution E-5476, the final 2026 Avoided Cost Calculator — the model that monetizes avoided generation energy and capacity, ancillary services, T&D capacity and decarbonization-policy-compliance value, and that feeds the Standard Practice Manual cost-effectiveness tests (TRC, PAC, RIM, PCT) used to screen every demand-side resource. This series reported the ACC at the D.26-09-007 stage in September; the October 8 resolution is the final issuance, and it is the version to cite. Anyone whose avoided-cost derivation rests on a pre-2026 vintage should re-run it against this one — California’s values remain the most widely borrowed reference for the ~$66/kW-year capacity method and the ~$20.70/MWh energy-efficiency cost this series benchmarks.
In the same meeting, D.26-10-006 rejected SDG&E’s proposal to withdraw from administering regional energy-efficiency programs, requiring the utility to continue in that role while permitting it to modify or cancel underperforming programs. That is the counterweight worth noting: regulators are refusing to let utilities abandon lower-glamour EE administration even amid the data-center capacity scramble. The Commission also approved, with added ratepayer protections (Resolution E-5455, held to October 29), PG&E’s exceptional-case agreement to serve a new 250-MW, 230-kV Google load in San Jose. (Source: Stoel Rives Energy Regulatory Update, October 14 — October 19 entry)
The curtailable-interconnection tariff wave is now an inventory, and load flexibility is the fastest item on it. Analysts argue that planners racing to bring data centers online against five-to-seven-year gas-turbine and equipment lead times are overlooking fast near-term capacity options — above all load flexibility and advanced transmission technologies, deployable in 36 months or less and in some cases capable of doubling a line’s capacity (six or more states have mandated considering advanced transmission since 2024). The concrete trend is a wave of connect-and-manage-style tariffs that credit curtailable load: SPP’s CHILL and pending PALS; PJM’s proposed IRAS (now rejected at the federal level); and ERCOT’s WLPUN and PCLR — on-site generation offsets, and faster interconnection in exchange for curtailable controllable load. All trade load flexibility for speed, which is the same flexibility a DR or VPP value stack monetizes.
The planning frame to cite is the July ESIG/Brattle report Transmission Planning with Large Loads (Pfeifenberger), which calls for scenario-based, least-regrets planning that explicitly includes load flexibility. The regulatory clock: FERC’s June 18 show-cause orders to six system operators on speeding large-load interconnection, with PJM granted an extension to November 16 — a standing item in this series, carried here in one line. With PJM projected to host roughly 70% of U.S. data-center development (White & Case) and most transmission investment still driven by reliability and asset renewal rather than load growth (FERC 2025 State of the Market), the throughline is that demand-side flexibility is the cheapest and fastest marginal resource available to planners, and the one most at risk of being left out of the plan — which is precisely what Robb’s remedy list did four days later. (Source: Utility Dive, October 8 — October 22 entry)
Movement — the 2026 Section 202(c) docket passed fifty orders, and the large-load curtailment thread runs through four of them. DOE’s running Federal Power Act Section 202(c) docket for 2026 now exceeds 50 emergency orders. The most recent, No. 202-26-50 — issued October 12, effective October 13–16, directing PJM to dispatch units and deploy backup generation — was the centerpiece of last week’s digest and is not a new item. What is new is the count and the pattern: multiple 2026 orders have explicitly authorized backup generation at data centers and large loads as a last resort before rolling blackouts — No. 202-26-01 (ERCOT, January 24), No. 202-26-06 (PJM, January 26), No. 202-26-07 (Duke Energy Carolinas & Progress, January 26) and No. 202-26-23 (PJM, May 18) — treating the very large loads reshaping every IRP as a controllable, curtailable reliability lever rather than only a threat.
The standing argument holds and gets stronger with the count: this emergency backstop is the expensive, last-resort version of what a planned DR or VPP program delivers in an orderly, compensated way. Fifty-plus orders across PJM, Duke, SPP and MISO is the adequacy evidence behind the ~$66/kW-year avoided-capacity benchmark. The caveat is unchanged: these are entity- and window-specific authorizations, not a standing program, and they confer no accreditation. (Source: U.S. DOE/CESER, accessed October 23 — October 23 entry)
Still unverified, third run — the Pennsylvania Bulletin publication date. The Bulletin citation that starts the 30-day comment clock plus 15 days reply on the PA PUC’s October 1 tentative order (Docket M-2026-3064961 — EDCs must shed new large loads without secured capacity before existing residential and commercial customers; data centers barred from “critical load” status) remains unlanded after three runs. A targeted search this week surfaced the PUC’s October 1 press release and the Bulletin‘s own volume-56 issues, but the site refuses programmatic retrieval, and no secondary source states the publication date. This is still the most time-sensitive action item on the board — the comment window is the strongest filing opportunity available to a demand-side intervener, and Pennsylvania has the only finished state product FERC pointed toward. Recommendation: check it manually at pacodeandbulletin.gov, or call the PUC’s Secretary’s Bureau. The cost-allocation technical conference on November 17 and the January 28, 2027 final-order target are unchanged.
🏢 Utility Programs & Resource Plans
Dominion’s 2026 IRP plans 3.5 GW of storage against a 16,000-MW statutory target, and prices the full-storage case as a $6.3B penalty. In its 2026 IRP — filed October 15, a full plan in North Carolina and an update in Virginia — Dominion’s Company Preferred Plan adds 3.5 GW of new storage over 20 years within 39.2 GW of total new resources (21+ GW solar, 8,965 MW gas, 3.4 GW offshore wind, SMRs from 2042), even though a new Virginia law directs the company to petition for 16,000 MW of short-duration storage by 2045, including 4,000 MW by end-2030. Dominion explains the gap as a modeling rule: it pre-programmed only the storage not subject to further Commission review — the 2030 petition amount — and let the model select the rest on least cost. The Southern Environmental Law Center calls that a shortfall. A “Forced Full Storage” sensitivity builds 14.5 GW and costs about $6.3B more on 20-year NPV, roughly $21B more in construction capital and about $15–16/month more on a residential bill.
The avoided-cost relevance is sharp and quotable. Dominion projects ~4% annual load growth (500–700 MW/year, demand doubling by 2045, data centers the named driver) and discloses that the 2028/2029 PJM capacity auction left committed capacity more than 6,800 MW below the reliability requirement — the same 6,831-MW shortfall this series has tracked since September, here restated inside the utility’s own resource plan. That is the adequacy gap demand-side resources are valued against, stated by the utility, in the same document that treats incremental storage as a cost rather than a hedge. Use the pairing: a filing that quantifies a 6,800-MW shortfall and simultaneously models additional flexible capacity as a $6.3B penalty has made the demand side’s argument available in its own exhibits.
Two things to track. The decisive event is the SCC ruling required by March 1, 2031 on whether the short-duration storage target stands — it resets the planning baseline for every demand- and supply-side resource in the Dominion zone. And this IRP arrives four weeks before Dominion’s November 15 VPP tariff filing and its December 1 pilot filing to the SCC for the up-to-450-MW, dual-enrollment Virginia VPP. Read the three together: if the IRP’s preferred plan does not carry the VPP as a resource, the tariff will be filed into a resource plan that does not count it. Also still open from prior runs: Dominion’s October CIAC compliance filing on mandatory contributions for direct-connect transmission — unverified, and the question of whether direct assignment extends upstream to network upgrades remains the one that decides how much data-center-driven transmission cost is directly assigned versus socialized. (Source: MGRID, October 20 — October 21 entry)
Three RTOs set new summer peak records in 2026, and EIA’s figure for PJM does not match the one in circulation. Drawing on Form EIA-930 hourly data, EIA reports that PJM reached 162.6 GW on July 2 (1.3% above its June 23, 2025 record of 160.6 GW), ERCOT hit 91.1 GW on July 22 (6.5% above its August 20, 2024 record of 85.5 GW) and SPP peaked at 58.0 GW on July 27 (3.5% above its August 21, 2023 mark of 56.0 GW) — most during regional heat waves in what EIA calls the hottest summer on record nationally. EIA attributes the peaks to cooling demand rather than load growth or data centers directly.
One reconciliation note, because both numbers are now in circulation for the same event. Trade reporting in this week’s log cites a PJM July 2 record peak of 168.2 GW, while EIA’s Form EIA-930 figure for July 2 is 162.6 GW — and independent market commentary on the October 13 event also used 162 GW as the summer peak. Use 162.6 GW and attribute it to EIA-930 unless a PJM primary source establishes the higher figure on a stated basis; the two likely differ in metric definition rather than in fact, but a capacity-value derivation built on the wrong one is off by 5.6 GW.
The relevance is unchanged: the avoided-capacity value justifying a DR or VPP program is set by the system peak it helps serve, so rising all-time peaks across PJM, ERCOT and SPP raise the marginal cost of capacity and therefore the ~$66/kW-year benchmark. ERCOT’s +6.5% is the standout — the sharpest single-year record break of the three. For IRP developers the takeaway is that peak growth is broad-based and weather-amplified, which is where capacity-focused DR accrues value rather than kWh-only efficiency. (Sources: U.S. EIA Today in Energy, October 15 — October 20 entry; Yes Energy, October 14 — targeted reconciliation)
🔬 EPRI Research Spotlight
No new EPRI publication fell inside this week’s window — the fifteenth consecutive week, and a targeted direct check again returned only prior-vintage material. Searching for EPRI research dated October 19–23 surfaced no new release; the available material was earlier work (the 2024 technical update on Grid Services Provided by Distributed Energy Resources, the DERMS Control Architecture for Grid Services in the Distribution and Bulk Power Systems product, the fast-and-flexible demand-response work and the DER-in-resource-planning product) alongside EPRI’s standing NERC Large Load Actions material. Report this as cadence, not omission: EPRI’s research releases arrive on a slower rhythm than a weekly digest, and the institute continues to reach the trade and regulatory record through citation and personnel rather than publication.
This week supplies an unusually sharp instance of exactly that — and of its limits. EPRI’s large-load material is the substrate beneath the NERC computational-load standards that cleared final ballot on October 15 and go to the Board on October 26 for a FERC filing in early November. So EPRI’s work is inside the regulatory instrument. But the instrument it is inside mandates observability for large loads without crediting their flexibility, and the NERC remedy list delivered at FERC on October 21 omits demand response and VPPs altogether. The institution whose research underpins the standards is not the institution deciding what the standards are for.
The methodological consequence is the one this series has restated for a quarter, and the calendar has now closed on it: the computational-load standards will be filed in early November — earlier than the December 31 deadline previously tracked — before the measured operational data exists. The only large-scale measured event remains the 3,800-MW July 22 trip in which data centers disconnected too early, and the 96-MW Aurora reference installation in Manassas, Virginia still targets late 2026 with no acceleration or slip detected. Meanwhile the operational record keeps accumulating outside the research channel: nine PJM demand-response days in 2026, a 745-MW California residential dispatch, and an aggregator’s claim of ~105% of committed reduction across two PJM emergency events. The measurement record the standards need is being generated by grid operators and aggregators in real time; the research institution’s cadence is slower than the regulatory calendar it informs, and the regulatory calendar just accelerated.
Carry forward from EPRI’s standing contribution, because it bears directly on this week’s items: EPRI’s Baum — microgrids and distributed assets must move from emergency-only to everyday dispatch, because “the majority of the time, the power is on.” Fifty-plus Section 202(c) orders in 2026 and a NERC remedy list that omits flexibility are two halves of the same failure that sentence describes.
🚩 Utility-Sector Relevance Flags
⚑ Demand Response Is Absent From NERC’s Remedy List — Get Into the Energy-Assurance-Standard Docket Now
Topic: Capacity Accreditation / Reliability Standards Participation
Relevance: Robb’s October 21 FERC statement named six risks (adequacy, unprecedented load growth, power-electronic generation and load, extreme weather, infrastructure-slowing policy, cyber/physical) and four remedies (forecasting anchored on a new energy assurance standard, urgent gen + transmission, IBR integration, NERC process modernization). DR, VPPs and load flexibility appear nowhere; DERs appear only as an integration risk. 2025 LTRA: summer peak +~225 GW, winter +~250 GW over ten years, >half the continent at adequacy risk.
Action Signal: Engage — The energy assurance standard is the door every DR/VPP accreditation claim must pass through, and it is being defined now. Track it as a primary docket, file into it, and frame your evidence as dependable energy under stress across a multi-hour window, not peak-hour response. The absence is a participation failure, not a technical verdict: the remedy list reflects the evidence that was on the table, and the operational record — nine PJM DR days, three shoulder-season events, 745 MW in California — is not being published in a form a standards body can cite.
⚑ Robb’s Winter Skepticism Is the Specific Attack — Build a Winter Record Before the Standard Asks For One
Topic: Seasonal Capacity Credit / M&V Scope
Relevance: Robb stated that most new five-year resources are battery and solar, which “struggle” to contribute reliably in some markets, “especially in winter,” against winter peak growth of ~250 GW over ten years. This series has already established that dispatch windows and accreditation built on summer afternoons undercount the September–October 5–8 p.m. net-load ramp; winter is the next and larger version of the same measurement gap.
Action Signal: Implement — Instrument and report winter performance explicitly: morning-ramp delivery, multi-hour sustained output, state-of-charge at call time, and cold-weather derate. A four-hour-duration resource claiming capacity credit in a December morning ramp needs a measured record, not a modeled one. Winter-peaking systems are the most exposed — and the most able to produce the evidence that answers the objection.
⚑ A Hyperscaler Just Bought Distributed Flexibility to 2040 — Duration Is the Product, and You Can Match It
Topic: Contract Term as a Procurement Instrument / BYOC
Relevance: Google is funding ~1,070 MW of Voltus VPP capacity across PJM through 2040 — about 10× the 100-MW June deal — with no price disclosed and full capacity in 3–4 years; participants can earn up to $350,000/MW-year. Voltus holds ~8.5–9 GW under contract against a 20-GW-by-2030 target after a $225M Series D; North American VPP capacity is ~37.5 GW across ~2,000 programs, so this is nearly 3% of the installed base.
Action Signal: Implement — Review the term length on every DR and DER program you run, and treat duration as the instrument it is. A 2040 contract removes merchant capacity-price risk, which is why it can finance batteries that a one-year program structurally excludes. Note that the counterparty here is a customer, not a commission — the easier sale this series has flagged since Cornell’s credit market. Cite with the caveats: target not resource, no disclosed price, and no confirmation PJM counts the capacity as delivered when called.
⚑ PJM Called DR on Nine Days in 2026 — More Than 2015–2024 Combined. Re-Price Your Program Accordingly
Topic: Dispatch Frequency / Program Cost and Enrollment Terms
Relevance: Nine DR activation days through October 15, 2026, versus eight across all of 2015–2024, and fifteen across 2025–26 (PJM’s David Souder). Context: ~11,000 MW connected data-center load, ~10,000 MW of generation returned during the October period with >61,000 MW still offline, two-thirds of outages planned work. PJM did not use the DOE authority to direct data centers onto backup.
Action Signal: Implement — Re-run program cost on nine calls a year, not one or two: incentive payments, customer fatigue, opt-out rates and M&V overhead all scale with frequency. Audit customer-facing enrollment terms written for a rare-event resource — event caps, notification windows, consecutive-day limits — because they will now bind. The accreditation case gets easier on frequency and harder on depth: more observations move the question from “will it respond” to “how much, verified, each time.” Still missing: deployed MW. Chase Operating Committee and Demand Response Subcommittee materials.
⚑ An Aggregator Reported ~105% Delivery in Two PJM Emergencies — The First Performance Claim, and Not Auditable
Topic: DR Performance Evidence / M&V Rigor
Relevance: Voltus reports delivering approximately 105% of committed load reduction across two PJM emergency events, while 46 GW of generation was in planned outage and 15 GW unexpectedly unavailable. PJM itself has still published no deployed megawatts or performance results for any 2026 shoulder-season event (September 1, September 15–18, October 13–16).
Action Signal: Watch — Use it as evidence that dispatch happened and cleared; do not cite it as an accreditation figure. It is company-reported, preliminary, and stated as a percentage of commitment rather than in megawatts, which is exactly the form an outside party cannot audit. The useful move is to make your own reporting the counterexample: publish delivered MW against committed MW, per event, with the baseline method stated. The side that publishes auditable performance will define what “dependable under stress” means in the energy assurance standard.
⚑ California Is Paying Down Enrollment Cost, Not Hardware Cost — Copy the Structure Into Your Next DSM Filing
Topic: Customer-Acquisition Economics / Program Design
Relevance: The CEC’s $11M (≈$22M with match) ChargeWise Phase 2 award is structured mostly as match funding covering 50% of user-acquisition, incentive and onboarding costs for utilities and CCAs, plus ~$1M for product development including V2X. Target: 9,000 → 79,000 drivers by end-2028, shifting charging off the 4–9 p.m. peak using day-ahead CAISO signals; Phase 1 reported up to 98% off-peak and up to 30% midday-shifted.
Action Signal: Implement — This is the funding counterpart to Brattle’s OEM-led-enrollment lever (4% of peak ≈ 40 MW per GW of system peak; ≈$2.6M/yr per GW at ~$66/kW-yr). If your DSM program keeps failing on cost per enrolled customer, propose acquisition-cost sharing rather than a larger device incentive — it attacks the binding constraint instead of the visible one. Caveat when citing: no MW disclosed, and a 98%-off-peak figure is a load-shape result, not a dispatchable-capacity result.
⚑ Dominion’s Own IRP States a 6,800-MW Shortfall and Prices Flexibility as a $6.3B Penalty — Use Its Exhibits
Topic: IRP Intervention / Storage Mandate vs. Least-Cost Modeling
Relevance: 2026 IRP (filed October 15): 3.5 GW of storage within 39.2 GW of new resources against a statutory 16,000 MW by 2045 / 4,000 MW by 2030; the “Forced Full Storage” case builds 14.5 GW at ~$6.3B higher 20-year NPV (~$21B more construction capital, ~$15–16/month residential). ~4% annual load growth; 2028/29 PJM auction left committed capacity 6,800+ MW below the reliability requirement. SCC must rule by March 1, 2031.
Action Signal: Engage — Intervene on the modeling rule, not the megawatt number: pre-programming only the storage exempt from further review and letting least-cost pick the rest is the choice that produces the gap, and it is contestable. Pair the 6,800-MW shortfall with the $6.3B figure from the same filing — a plan that quantifies the adequacy gap and treats additional flexible capacity as a penalty has made the demand-side case in its own exhibits. Then read it against the November 15 VPP tariff and the December 1 pilot filing: if the preferred plan does not carry the VPP as a resource, the tariff lands in a plan that does not count it.
⚑ Two Gigawatt-Scale Large-Load Deals Carry Storage With No Named Orchestrator — Ask Who Dispatches It
Topic: Large-Load Contract Design / Flexibility Crediting
Relevance: NiSource/Microsoft: ~1,100 MW new generation on a 20-year term from 2028, plus 600 MW of battery storage under a tolling agreement pooled across GenCo’s large-load customers; Microsoft funds all generation and transmission; NiSource projects ~$1.75B customer savings (its own figure, up from $1.4B). Entergy/Google Phase 2: ~$600M Google-funded, 1,000 MWh at Driver Solar + 400 MWh at Jones Mill (1.4 GWh), a 440-MW(ac) Big Island PPA the APSC denied in early September now riding inside the deal, two new 230-kV lines; claimed ≥$500M new 20-year savings, calculation unexplained.
Action Signal: Engage — Nothing in either disclosure says who orchestrates ~2 GWh of storage, against what signal, or whether its flexibility is credited anywhere. Put that question in both dockets, and ask for the dispatch right and the telemetry spec as contract terms. Note the prudence asymmetry to raise on the record: a customer’s funding commitment revived a PPA that failed a standalone prudence test — a standard a demand-side portfolio does not get to borrow. And note the moved bar: when a large load funds its own capacity, the marginal-resource test becomes “cheaper than capacity someone else is paying for.”
⚑ Multifamily Is a New DR Site Class — and Peace Bridge Place Is the Four-Layer Control Chain to Spec Against
Topic: DERMS Architecture / Site-Class Expansion
Relevance: 3.75 MWh across seven battery rooms in a 105-unit fully electric Buffalo building (~36 kWh/apartment), run as a National Grid pilot with utility dispatch through EnergyHub Edge DERMS → Viridi EMS, shaving monthly demand-charge peaks and discharging on peak events while participating thermostats pre-condition (opt-out). National Grid pays for non-import and export. No cost, no kW rating, no NYSERDA award and no DR revenue disclosed; safety evidence (UL 9540A non-propagation, UL 9540 listing) is stronger than the economics. Viridi’s Buffalo plant makes ~400 MWh/yr (~100 systems this size).
Action Signal: Implement — Two uses. Site class: one landlord contract reaches 105 meters, which is a fundamentally better acquisition economics than single-family VPP enrollment — build a multifamily channel. Architecture: utility → third-party DERMS → building EMS → battery + thermostats is the chain to write into a DERMS spec, with latency, telemetry fidelity and command arbitration as pass/fail acceptance criteria at each hop. Do not carry the economics: a 3.75-MWh energy rating with no power rating cannot be converted into a capacity contribution at all — require kW alongside kWh in every disclosure you rely on.
⚑ Use 162.6 GW for PJM’s 2026 Peak — Two Figures Are in Circulation and They Differ by 5.6 GW
Topic: Avoided-Cost Inputs / Benchmark Hygiene
Relevance: EIA Form EIA-930: PJM 162.6 GW on July 2 (+1.3% over 160.6 GW, June 23, 2025); ERCOT 91.1 GW on July 22 (+6.5% over 85.5 GW); SPP 58.0 GW on July 27 (+3.5% over 56.0 GW). Trade reporting in this week’s log cites a PJM July 2 peak of 168.2 GW; independent market commentary on the October 13 event used 162 GW.
Action Signal: Implement — Anchor on 162.6 GW attributed to EIA-930 unless a PJM primary source establishes the higher figure on a stated basis; the difference is probably metric definition, but a capacity-value derivation built on the wrong one is off by 5.6 GW. Separately, the records themselves lift the marginal cost of capacity behind the ~$66/kW-year benchmark, and ERCOT’s +6.5% is the sharpest single-year break — which is where capacity-focused DR, not kWh-only efficiency, accrues value.
⚑ The Curtailable-Interconnection Tariffs Are Now an Inventory — Screen Your Territory Against All of Them
Topic: Load-Flexibility-for-Speed Tariffs / Program Positioning
Relevance: SPP CHILL plus pending PALS; PJM IRAS (proposed, rejected federally October 9); ERCOT WLPUN and PCLR — all trading curtailable load for faster access to existing transmission, against 5–7-year gas-turbine lead times, while load flexibility and advanced transmission deploy in ≤36 months and can in some cases double a line’s capacity. Planning frame: July ESIG/Brattle Transmission Planning with Large Loads (Pfeifenberger) — scenario-based, least-regrets, flexibility included. FERC’s June 18 show-cause orders to six operators, PJM extended to November 16 (carried item, one line). PJM ~70% of U.S. data-center development (White & Case).
Action Signal: Watch — Maintain these as a single screened inventory rather than separate news items, because they are the same product under five names and the differences in curtailment depth, notice and compensation are what decide whether your portfolio is competing with them or selling into them. Add the venue column this series flagged after October 9: who writes the rule — FERC, a state, an RTO tariff, or a bilateral market — now varies more than the resource characteristic does.
⚑ The Equipment Order Book Raises Your Avoided-T&D Number — Directionally, Not Yet Quantifiably
Topic: Supply Chain / Avoided-T&D Cost
Relevance: ABB record Q3 orders $12,055M (+40%), revenue $9,701M (+14%), 1.24 book-to-bill; U.S. orders +82%; electrification orders +70% to a record $7.67B, Americas electrification +117%; data-center orders triple-digit. Yet capex is held near $1B with no mid-term increase — “incremental” capacity, third-party assembly, a 51% JV on Mitsubishi Electric’s LV distribution business (~$495M), solid-state transformer commercial H2 2027. No transformer or switchgear lead times disclosed.
Action Signal: Watch — If order growth of this magnitude is not matched by manufacturing capacity, lead times for distribution and substation gear stay long, which raises the effective avoided-T&D and avoided-capacity cost a DR or VPP program is credited against. Treat it as a sensitivity input rather than a benchmark — it is an order story, not a delivery story — and keep pressing suppliers for lead times in writing, because that is the figure that converts this into a number you can file.
⚑ CCAs Are Writing Large-Load Integration Rules in Parallel With the IOUs — That Is a Second Venue
Topic: Large-Load Integration / Flexibility Obligations
Relevance: A 400-MW data center is under study for the federal 70.3-acre NAVWAR site in San Diego under a 99-year lease needing no city permits; a city moratorium push reaches private sites only. SDG&E will build to meet the load but has disclosed no cost or cost-allocation, and there is no interconnection request or load study — the 400-MW figure comes from one developer email. San Diego Community Power had a draft large-load-integration guideline on its October 22 agenda. Comparison: a 330-MW Imperial County project is on hold after an August 31 EIR ruling.
Action Signal: Watch — Track CCA and municipal large-load guidelines as a separate docket stream from IOU tariffs: they move faster, they set expectations the IOU proceeding later inherits, and a continuous 400-MW load is exactly the negotiation in which on-site generation, storage and demand flexibility get priced. Treat the 400 MW as unverified until an interconnection request or load study exists.
📌 Sources
October 19, 2026 Entry
– MGRID — Sunrun and Tesla Push a Record 745 MW of Home Batteries Into California’s Grid, But Sunrun Gives No Company Split (October 14, 2026) (movement/reconciliation — the October 6 dispatch was reported in a prior digest as a 601-MW, 85,774-home Tesla figure; this gives the combined Sunrun-and-Tesla total for the same event as 745 MW, not a new record)
– MGRID — AEP Puts $4 Billion Into a 702 MW Fuel Cell Plant for Google, But Has Not Said Its Offtake Conditions Are Met (October 16, 2026) (first appearance)
– Citizens Utility Board — FERC Rejection of PJM Data Center Plan Makes State Action Last Hope to Protect Consumers (October 18, 2026) (movement on the October 9 FERC order reported last week; carried only for the capacity-market demand-curve provision. Advocacy framing; the ~$29.4B figure is a carried benchmark, not a new finding)
– Stoel Rives — Energy Regulatory Updates: October 14, 2026 (covering CPUC actions of October 8) (movement — the 2026 ACC was reported at the D.26-09-007 stage in September; Resolution E-5476 of October 8 is the final issuance. D.26-10-006 and Resolution E-5455 are first appearances)
October 20, 2026 Entry
– MGRID — Base Power Takes Its Home Battery Fleet Outside Texas With a 50 MW, 2,500-Battery Deal for Central Georgia EMC (October 18, 2026) (first appearance)
– MGRID — NiSource Signs 20-Year Microsoft Deal That Adds 1,100 MW in Indiana and Contracts a 600 MW Battery (October 19, 2026) (first appearance)
– U.S. Energy Information Administration — Today in Energy: PJM, ERCOT, and SPP All Set New Summer Peak Electricity Demand Records in 2026 (October 15, 2026) (first appearance; Form EIA-930 hourly data)
– Blackridge Research — Enerflex Books Over One Gigawatt Data Center Power Contract, Pushing Behind-the-Meter Backlog Past 1.75 Gigawatts (October 13, 2026) (first appearance; vendor-disclosed backlog, end customer unnamed)
October 21, 2026 Entry
– MGRID — Viridi Puts a 3.75 MWh Battery Inside an Occupied 105-Unit Buffalo Apartment Building, but Discloses No Cost (October 20, 2026) (first appearance; the week’s DERMS anchor)
– MGRID — Dominion Plans 3.5 GW of Storage Against Virginia’s 16,000 MW Target, Says the Full-Storage Case Costs $6.3 Billion More (October 20, 2026) (first appearance; the 6,831-MW 2028/29 shortfall is a carried benchmark, here restated inside the utility’s own IRP)
– MGRID — Google Takes On the Big Island Solar PPA Arkansas Regulators Rejected, in a $600 Million Second Phase With Entergy (October 20, 2026) (first appearance)
– MGRID — ABB Books a Record $12 Billion Quarter With US Orders Up 82%, But Keeps Capital Spending Near $1 Billion (October 20, 2026) (first appearance)
October 22, 2026 Entry
– Utility Dive — What Google’s Virtual Power Plant in PJM Says About the Race for Electricity (October 19, 2026) (carried as context for the 1,070-MW item below — this covers the 100-MW June 2026 Google/Voltus deal and the BYOC structure, which the October 22 announcement is roughly ten times larger than)
– ev.energy — ev.energy to Scale EV VPP to 79,000 Drivers in California with $22M from the California Energy Commission and Partners (October 20, 2026) (first appearance)
– Utility Dive — Reliability Standards for AI Data Centers, Crypto Miners Pass Final NERC Ballot (October 15, 2026) (movement — the CLO-001/-002/-003 comment period and 10-day ballot closed September 18 and were reported then; the final ballot is new)
– Utility Dive — Power System Plans to Meet Large-Load Demand Miss Near-Term Solutions: Analysts (October 8, 2026) (first appearance for the curtailable-interconnection inventory and the ESIG/Brattle planning frame; FERC’s June 18 show-cause orders are a carried item)
October 23, 2026 Entry
– MGRID — Google Pays Voltus for a 1,070 MW Virtual Power Plant in PJM Through 2040, Ten Times June’s 100 MW Deal (October 22, 2026) (first appearance; distinct from the 100-MW June deal above — a new announcement roughly ten times larger)
– MGRID — Budderfly Pitches a 150 MW City VPP and $500 Million, but Its One Municipal Contract Is $4.4 Million in Cadillac (October 21, 2026) (first appearance)
– MGRID — Manchester Financial Weighs a 400 MW Data Center on San Diego’s 70-Acre NAVWAR Site as City Leaders Push a Moratorium (October 22, 2026) (first appearance)
– U.S. Department of Energy (CESER) — 2026 DOE 202(c) Orders (accessed October 23, 2026; most recent order No. 202-26-50 issued October 12, 2026) (movement only — Order No. 202-26-50 was last week’s centerpiece and is not re-reported; carried for the >50-order 2026 count and the four large-load/backup-generation orders)
In-Window Primary Document and Targeted Gap-Fills
– NERC — Jim Robb Opening Statement, FERC 2026 Reliability Technical Conference (October 21, 2026; press release October 20) (first appearance; in-window primary document, cited directly rather than through secondary coverage)
– EconoTimes — PJM Uses Demand Response on Nine Days Amid Rising Data Center Load (October 21, 2026) (first appearance; in-window targeted gap-fill on the standing open item for PJM DR dispatch data — supplies frequency, not deployed megawatts)
– Yes Energy — Total (Planned Outage) Recall: A Look Inside PJM’s October 13th Price Spike (October 14, 2026) (targeted reconciliation only — 46 GW planned / 60 GW total outages, prices above $1,500/MWh, 108 GW peak load, and a 162-GW July 2 summer peak corroborating the EIA figure against the 168.2 GW in circulation)
– PA PUC — PUC Advances Consumer Safeguards as Data Center Growth Adds New Demands to Electric Grid (October 1, 2026) (carried; located again this week, but the Pennsylvania Bulletin citation that starts the comment clock remains unverified for a third run)
Method: All five weekday entries (Monday October 19 through Friday October 23) are present in the research log; no days are missing. Items are reported on the day they entered the log, with their original publication dates shown. Several carry pre-week publication dates: the AEP/Bloom and Sunrun-Tesla items (October 14 and 16), the Base Power deal (October 18), the CUB analysis (October 18), the Stoel Rives digest of CPUC actions taken October 8, the EIA peak-records brief and the NERC final-ballot report (October 15), the Enerflex award (October 13), the Utility Dive Google/Voltus 100-MW piece (October 19) and the large-load near-term-solutions analysis (October 8). Within-log deduplication: the two Google/Voltus items are distinct deals, not one event reported twice — the October 19 Utility Dive piece covers the 100-MW June 2026 contract and the October 22 MGRID piece covers a new, roughly tenfold-larger ~1,070-MW commitment through 2040; the former is carried as context for the latter rather than as a separate item. The NiSource/Microsoft and Entergy/Google packages are reported as one consolidated item because they are the same cost-allocation instrument. Cross-week check: the October 6 California residential-battery dispatch was reported in a prior digest as a 601-MW, 85,774-home Tesla figure and appears here only as a reconciliation to the 745-MW combined Sunrun-and-Tesla total for the same event — not as a new record; DOE Order No. 202-26-50 was last week’s centerpiece and appears only as movement (the >50-order 2026 count and the four large-load/backup-generation orders); FERC’s October 9 rejection of IRAS and BYOC was reported in full last week and appears only for the capacity-market demand-curve provision the news coverage omitted; the NERC CLO standards’ September 18 ballot close was reported then, so the October 15 final ballot and the October 26 Board date are movement; the 2026 Avoided Cost Calculator was reported at the D.26-09-007 stage, so Resolution E-5476 is the final issuance rather than a new model; FERC’s June 18 show-cause orders and the ~$29.4B PJM capacity figure are carried benchmarks, cited in one line each. Zero source URLs overlap last week’s Sources section. Targeted search was used only to verify or fill gaps: it resolved the PJM DR dispatch-frequency figure (nine days in 2026 versus eight across 2015–2024) and corroborated EIA’s 162.6-GW PJM peak against the 168.2 GW in circulation, and it did not resolve PJM’s deployed demand-response megawatts for any 2026 shoulder-season event, which remains the single number an avoided-cost filing needs. The Pennsylvania Bulletin publication date that starts the 30-day comment clock on the PA PUC’s October 1 tentative order (Docket M-2026-3064961) remains unverified for a third run — the Bulletin site refuses programmatic retrieval and no secondary source states it; it should be checked manually, and it is still the most time-sensitive action item on the board. EPRI direct check: no publication dated inside the window — the fifteenth consecutive week — with EPRI appearing instead as the substrate beneath NERC’s computational-load standards. One item was included at Marc’s request and is cited from the primary document rather than secondary coverage: Jim Robb’s October 21 opening statement at FERC’s 2026 Reliability Technical Conference.