All five weekday entries (Monday September 21 through Friday September 25) are present in the research log. Seven items are deduplicated across weeks rather than within this one and are reported as movement or as standing benchmarks rather than as new findings: the PG&E/Rewiring America/Google SHARE VPP was reported in the September 4 digest and is carried here only for the megawatt target and compensation terms this week’s entry adds; Puget Sound Energy’s 2026 DR RFP appears for the fourth time in the series and is carried for its September contract-negotiation phase and December 15 service date; the T&D World / Itron DERMS-alongside-SCADA analysis (June 9 source) appears for the third time and is carried in one line; FERC’s ISO-NE Order 2222 ruling in Docket ER26-1956 was fully reported in the September 11 digest and is carried only for the supply-chain extension it also granted; the PowerHouse Hillwood / ComEd transmission-security dispute was reported in the August 28 digest at the filing stage and is carried here because FERC ruled on it this week — movement, not a repeat; NERC’s computational-load standards have been tracked since September 4 and are carried for the applicability thresholds and registry schedule that emerged after the September 18 comment close; and the electric-school-bus V2G thread was reported in the July 24 digest from the Massachusetts demonstration and is carried here for the national deployment count, with the two conflicting per-bus revenue figures reconciled explicitly. One item — the DOE §202(c) order to Duke Energy Carolinas — is the third order in this series’ emergency-authority thread and is reported for the cadence only. Several items carry pre-week publication dates (the NV Energy tariff order of August 11 reported August 26, the September 9 Sunrun/Tesla dispatch reported September 21, the May 29 ISO-NE order, the June 9 Itron piece, the September 4 SHARE announcement, the September 11 Pennsylvania PUC vote, the September 16 House vote and school-bus feature, the July 31 Virginia SCC order) and are reported on the day they entered the log with original dates shown. Three verification items were chased by targeted search and are noted where they land.
Last week the adequacy gap got measured. This week the denominator got audited.
Start with the denial, because it is the precedent. On September 18, reported September 22, the North Carolina Utilities Commission denied Duke Energy Progress’s request to build a 255-MW hydrogen-capable F-class simple-cycle combustion turbine at the Smith Energy Complex. Public staff called the $584 million price tag “staggering.” The commission held that projected load growth was “much…based upon anticipated data center customer additions” that were “insufficiently reliable for the Commission to act at this point” — and tied the denial to Duke’s commitment under the White House Ratepayer Protection Pledge, folding cost-causation into a siting decision. This sits inside a $103-billion capital plan, the largest on file at any regulated U.S. utility. Staff had reluctantly supported the turbine as “the only resource that is executable on a constrained timeline,” warning of a breach of a 22% planning reserve margin by 2031. The commission denied it anyway.
Then the arithmetic behind the denial, at national scale. Utility Dive’s September 24 roundup assembles the cleanest aggregate evidence this series has acquired that large-load tariffs strip phantom megawatts out of forecasts: 65-plus large-load tariffs proposed or approved across 30-plus states, and — the number to carry — AEP Ohio cut its large-load forecast by more than half, from 30 GW to 13 GW, after Ohio regulators approved a new large-load tariff. That is a ~17-GW reduction at a single utility, caused by nothing but a requirement that customers post money. About a third of these tariffs require minimum payments over a set term, more than half include collateral or credit protections, and roughly half impose exit fees.
Then what an unsecured commitment is actually worth. On September 22 FERC rejected ComEd’s notice of cancellation of a transmission security agreement for a 1.8-GW, $20-billion data center in Joliet, Illinois, leaving the agreement in place. The developer contended it had met the initial credit requirement with a $1 posting. Commissioner LaCerte called the proposition that a dollar “may provide appropriate security” an “embarrassing legal fiction.” Read that beside AEP Ohio’s 17-GW haircut and the mechanism is unmistakable: the forecast and the collateral are the same variable.
And then the object lesson, filed the same week. On September 22 Xcel Energy and Otter Tail Power asked South Dakota regulators to approve roughly 160 miles of new high-voltage line at up to $1.93 billion — including a 90-mile, 765-kV segment estimated at $1.3–1.7 billion — citing data-center and industrial load in a state that hosts no hyperscale campuses, while the Upper Midwest’s actual growth landed in North Dakota and Minnesota. The 765-kV route holds easements on 81 of the 274 parcels it needs. Cost allocation is unsettled. The decision is not expected until late 2027 or early 2028.
Against all of that, the demand side posted its largest measured number on record. On the evening of September 9, reported September 21, Sunrun and Tesla dispatched more than 580 MW into CAISO across a three-hour window — 517 MW from Tesla Powerwalls and 63 MW from other home batteries, drawn from ~110,000 Powerwalls plus 30,000-plus third-party batteries across all three California IOU territories, coordinated through the CEC’s DSGS and CPUC’s ELRP programs and triggered when CAISO’s day-ahead price passed $200/MWh. A September 10 SCE-only follow-on added more than 140 MW. Enough to power all of Sacramento County at peak, from equipment already installed in people’s houses.
And NERC put a floor under the whole conversation. The three draft Computational Load Organization standards — CLO-001-1, -002-1, -003-1 — closed their first comment period and 10-day ballot on September 18, and the applicability test that emerged is a two-part one FERC itself declined to name: a site of 50 MW or more that also connects at 100 kV or above.
The week’s argument, stated plainly: the speculative load forecast lost its presumption of validity in four separate venues in five days — a commission denial, a 17-GW tariff-driven haircut, a dollar of collateral held up to public ridicule, and $1.93 billion sought for load that went elsewhere — while the demand side delivered 580 measured megawatts out of existing customer equipment. Every element of that comparison is citable, and three of the four came from regulators rather than advocates.
⚡ Virtual Power Plants & Demand Flexibility
Aggregated home batteries delivered 580 MW on command into the California grid on September 9, and that single number does more for a DSM business case than any VPP forecast this series has reported in five months — because it was metered, it was dispatched by program rule rather than by pilot arrangement, and it happened at peaker scale. Sunrun and Tesla discharged more than 580 MW during a heat wave across a three-hour evening window, the largest distributed VPP event on record: 517 MW from Tesla Powerwalls and 63 MW from other home batteries, drawn from approximately 110,000 Powerwalls plus 30,000-plus third-party batteries spanning all three California investor-owned utility territories. Dispatch ran through two statewide grid-service programs — the CEC’s Demand Side Grid Support program and the CPUC’s Emergency Load Reduction Program — and triggered when CAISO’s day-ahead LMP exceeded $200/MWh. A September 10 follow-on limited to SCE territory added more than 140 MW, implying combined potential above 720 MW in a single coordinated dispatch.
The figure that ports directly into an avoided-cost filing is Brattle Group’s — commissioned by Sunrun and Tesla — that California VPP programs could deliver up to $206 million in net cost savings by 2028 for all grid-connected customers. That is the same deferred-poles-and-wires logic this series has applied since May, now attached to a demonstrated dispatch rather than a device count.
Two caveats, and they are the procurement-relevant part. First, this is still not accreditation. The standing watch item this series has carried since June — whether the 16.8-GW Sunrun–Renew Home–Tesla aggregation converts from framework into capacity credited in an auction or an IRP at a stated ELCC — is not closed by an event, however large. Two aggregators demonstrating 580 MW on a voluntary program trigger is not a resource-adequacy commitment, and a planner should book it as proof of performance, not proof of capacity. Second, the operators themselves flagged the “shock to the system” of ramping that much capacity at once and said they want a smoother ramping curve next time. That is a DERMS dispatch-design requirement, not a footnote: a platform that can call 580 MW but cannot shape the ramp creates a new disturbance while relieving another, and the ramp-shaping capability belongs in the requirements document with acceptance criteria. (Source: Renewable Energy World / Factor This, September 21 — September 21 entry)
Nevada approved two performance-based DER tariffs that use exactly the avoided-cost method this series argues for — and then declined to let anyone but the utility run them, which makes it the cleanest counterexample on the make-versus-buy question now open in four states. The Public Utilities Commission of Nevada’s August 11 order (reported August 26) clears NV Energy — Nevada Power and Sierra Pacific Power — to develop Energy Grid Services and Capacity Grid Services riders, replacing a demand-response framework 15 years on the books. The compensation design is the textbook version: payment for load reduction at the avoided cost of generation and transmission, plus payment for dispatched energy at hourly market prices.
What the commission refused is the instructive half. Over objections from SEIA, Solar United Neighbors and Advanced Energy United, it declined to let customers assign performance compensation to third-party aggregators or OEMs, declined to require an open data-sharing platform, and declined to add a VPP resource type to NV Energy’s next IRP — citing “insufficient information to determine whether a VPP is a viable resource option for Nevada.” NV Energy’s answer is that the aggregator question belongs in its forthcoming CAISO EDAM participation docket, with EDAM entry expected around late 2028.
Set that against the three other live designs and the map is complete: Virginia ordered dual enrollment (utility-direct or aggregator, reported last week); Minnesota approved a utility-owned, rate-based program and declined a parallel third-party one; New Jersey has not decided; Nevada has now chosen utility-only and deferred the aggregator question by two years. For a utility writing a DERMS specification, the lesson is unchanged and now has four data points behind it: the enrollment-channel decision is made by a commission, it can be reversed by a later docket, and the platform has to survive the reversal. Building for utility-direct settlement only, because that is what your commission approved this year, is the decision that gets expensive at the data model. Advocates’ substantive warning is also worth recording for a cost-effectiveness test: a utility-only structure in a high-solar, low-storage state, after the residential-battery ITC expired at the end of 2025, is likely to suppress behind-the-meter battery attachment rates — which is the enrollment base a later VPP would need. (Source: Utility Dive, August 26 — September 21 entry)
Bidirectional electric school buses now operate as a demand-side resource in 26 utility programs across 19 states, and the segment has crossed from demonstration into a fleet-scale asset class — but this week’s national figures carry a per-bus revenue number roughly half the one this series published in July, and the gap matters more than the growth. At least 26 utilities across 19 states run V2G programs; U.S. districts have more than 230 V2G-enabled buses operating out of 14,000-plus committed electric buses across roughly 1,550 districts, per the WRI Electric School Bus Initiative. PG&E’s territory hosts the largest deployments: Zum’s 74-bus Oakland fleet supports 2.1 GWh of annual discharge, with a planned 238-bus San Francisco fleet targeted at high-need circuits.
The design lessons map cleanly onto DR procurement. The highest guaranteed value comes from emergency and peak programs — PG&E’s Emergency Load Reduction Program is named specifically. A charge-management layer (The Mobility House, Highland Electric Fleets) handles telemetry and dispatch, which means the utility is contracting with a fleet operator and an orchestration vendor rather than with a school district. And vertically integrated utilities without wholesale markets can still capture the value through a capital-offset model — Georgia Power’s version amounts to buying half the battery in the bus in exchange for ten years of guaranteed summer load-shifting, which is the avoided-CapEx framing this series’ business case uses, executed as a procurement rather than an incentive.
The revenue reconciliation is the item to carry forward. This series reported in July, from the MassCEC vehicle-to-everything demonstration, that a Beverly, Massachusetts bus discharged 10.78 MWh and earned $23,500 across two summers via National Grid’s ConnectedSolutions VPP, with school buses earning up to $12,000 per year. This week’s national feature reports the same Beverly bus at ~10.8 MWh per season and ~$12,000 per season — and then has Highland Electric Fleets’ CEO put the realistic Massachusetts norm at about $6,000 per year once discharge patterns and integration costs are accounted for. Those are not contradictory reports; they are a best case and an expected case, and a business case that cites the $12,000 figure as a program average will not survive an intervenor who finds the operator’s own $6,000. Cite the range, attribute both ends, and use $6,000 for portfolio math. The procurement caveats reported from the EPRI/Dominion Virginia demonstration stand unchanged and are the reason the expected case sits below the best case: low 12-volt battery state-of-charge has blocked bidirectional engagement, and feeder-level hosting capacity must be studied fleet by fleet. (Source: Utility Dive, September 16 — September 21 entry; the Beverly/ConnectedSolutions benchmark was previously reported in the July 24 digest from the MassCEC demonstration)
Carried — PG&E SHARE. The Google-funded, Rewiring America-enrolled VPP on PG&E’s DERMS platform was reported in the September 4 digest for its funding structure — third-party capital, zero rate-base exposure, utility keeps the operational record rather than the assets. This week’s entry adds the quantification that was missing then, and it is worth recording: the program targets approximately 12 MW by 2028 from up to 21,000 existing DERs plus new battery-enabled Carrier heat pumps, with encoord supplying distribution and transmission modeling and Demand Side Analytics handling orchestration, sited in Santa Clara and Alameda counties to test locational value against specific regional grid needs. Compensation reference points for the value stack: PG&E ELRP Powerwall owners earn $2.00/kWh delivered during dispatched events, up to roughly $350/year; the heat-pump subsidy is at least $5,000, up to $10,000 for the first 25 households. The program runs through 2027 with findings due late 2026 or early 2027. Twelve megawatts is pilot-grade — its value to an IRP is methodological, as the first utility-run test of whether locational DER siting can be shown to relieve a named constraint, which is the same question Minnesota and Virginia have now ordered answered on longer clocks. (Source: pv magazine USA, September 4 — September 22 entry; previously reported in the September 4 digest)
Carried — Puget Sound Energy. The 2026 Demand Response RFP appears for the fourth time in this series and is carried for its schedule, which has now advanced: proposals closed June 15, the short list was set July 3, contract negotiations are underway this month, and service launches December 15, 2026. PSE holds 129 MW of DR today and is buying net-new, turn-key capacity for both winter and summer peak. Two requirement details from this week’s entry belong in any DR solicitation being drafted now: a single primary contract owner must be accountable for a multi-aggregator bid, and bid capacity must be incremental — it may not duplicate participation in existing PSE programs, which is the provision that prevents a utility from paying twice for the same customer. One number needs reconciling before anyone cites this procurement. This week’s entry states that bidders must demonstrate a pathway to at least 50 MW by 2030; the August 28 digest reported PSE’s portfolio target as approximately 500 MW year-round by 2030, derived from Washington’s DALCU statutory requirement of 10% of historical peak in both seasons. Read together, the 50 MW is a per-bidder qualification floor and the 500 MW is the portfolio obligation — they are different objects, and presenting the 50 MW figure as PSE’s demand-side target would understate the mandate by an order of magnitude. (Source: Puget Sound Energy 2026 DR RFP — September 22 entry; previously reported July 31, August 14 and August 28)
🔌 DERMS & Grid Integration Technology
A survey of 850 utility executives says the industry’s investment priority has swung from decarbonization to resilience and affordability — and it names DERMS and storage among the instruments it is swinging toward, which is the budget argument a DERMS business case now has to be written in. The IFS-commissioned study, conducted by Censuswide across 850 C-level and senior utility executives globally and released September 24, describes a “pendulum shift” away from the sustainability and decarbonization goals that previously dominated utility investment, toward grid hardening, new energy sources and storage, and asset lifecycle management, all underpinned by Industrial AI. Grid flexibility — explicitly including BESS and DERMS — is named among the tools utilities are turning to in order to manage demand volatility and protect against prolonged outages.
The practical significance is about which argument wins a budget cycle, not about market sizing. This series has tracked the DERMS spend case as it moved from a clean-energy-enablement rationale to a reliability-and-affordability one, and that reframing is what put DERMS in the capital plan alongside ADMS and AMI rather than in an innovation line. A survey at this sample size is directional confirmation that the shift is general rather than local, and it broadens the coalition a procurement can appeal to: a DERMS justified on outage duration and bill impact is defensible in front of a commission that has grown skeptical of clean-energy framing, and in front of a CFO who reads affordability as cost control.
The caveat is the ordinary one for this class of evidence and it should be applied strictly: this is a vendor-commissioned executive-sentiment survey, with the full Censuswide report due at the end of September. It is evidence of where capital intent is moving. It is not a spend figure, a deployment count or a benchmark, and it does not belong in a filing as any of those. Use it in an internal business case to explain why the framing changed; use the deployment record — Xcel Colorado’s >15 MW in under six months, PG&E’s 90%-of-hours / <1%-curtailment flexible-interconnection record, and now the 580-MW California dispatch — when a regulator asks for proof. (Source: PR Newswire, September 24 — September 24 entry)
Carried — the DERMS-as-infrastructure case. Itron’s Nick Tumilowicz in T&D World (June 9 source) appears for the third time in this series and is carried in one line, because the argument has not changed and the week’s other items keep validating it: DERMS crossed from the innovation budget into the capital plan in 2026 and is now held to SCADA-grade reliability, response-time, failover and redundancy specifications, extending ADMS rather than replacing it, with the grid-operations-versus-customer-programs organizational split as the real deployment blocker. Note how directly the Sunrun/Tesla ramp problem above lands on that last point: the ramp-shaping fix requires the customer-program team’s dispatch rules and the grid-operations team’s stability limits to be the same document. (Source: T&D World, June 9 — September 22 entry; previously reported August 28 and carried September 4)
🔋 Energy Storage
Massachusetts split $40 million across 36 behind-the-meter storage projects, and the 13-fold spread in what the state paid per kilowatt-hour is the most useful thing in the award — because it is the exact gap a value-of-avoided-outage number is supposed to close, and the state published no cost-effectiveness screen. On September 22 the Massachusetts Department of Energy Resources awarded more than $40 million to 36 projects under its Advancing Massachusetts Power Energy Storage Grant Program, across three categories: Community Resilience, Safety and Education, and Long Duration Energy Storage. The 28 Community Resilience awards total 10.26 MW / 44.14 MWh for $32.4 million, sited at municipal buildings, schools and affordable- and public-housing sites, with the largest single award at $2.5 million and the state noting that a “vast majority” flows to environmental-justice communities.
The economics are explicitly resilience-driven rather than arbitrage-driven, and the range shows it: state dollars run from roughly $56/kWh for a 4-MW/80-MWh long-duration award up to roughly $764/kWh for a small school system. That is a ~13× spread inside one program, with no published cost-effectiveness screen behind the allocation.
For DSM and DER business-case work the spread is the point rather than the criticism. Twenty-eight separate procurements at that range imply a value-of-avoided-outage-hour figure the state has not written down — and the avoided-cost methodology this series works from (deferred generation and T&D capacity plus a quantified reliability value) is precisely the framework that would let a utility or a regulator defend spend at the high end of that range on a stacked-benefit basis. A utility that can show why a school system is worth $764/kWh of public money and a long-duration project is worth $56/kWh has an answer to the affordability question that a grant program by itself does not.
The caveat for planners is the resource-accounting one: these are grant contributions to municipal resilience assets, not dispatchable grid resources. Their capacity value to an IRP is limited unless enrolled into a DR or VPP program — and Massachusetts has a ready-made enrollment path in ConnectedSolutions, plus a 3.5-GW by 2035 demand-management target (reported last week) whose September baseline inventory is running now. Whether these 10.26 MW get counted toward that target, and whether a state-funded battery can also earn performance payments, is a definitional question the inventory will have to answer — and the answer will be copied. (Source: Mgrid.org, September 22 — September 25 entry)
🏗️ Data Centers & Large Load Growth
The most consequential number of the week is a subtraction: AEP Ohio cut its large-load forecast from 30 GW to 13 GW after regulators approved a new large-load tariff — a 17-GW reduction produced by a financial requirement rather than by any change in the underlying economy. Utility Dive’s September 24 national roundup assembles the trend that reduction sits inside: as of late 2025, more than 65 large-load tariffs had been proposed or approved across 30-plus states, aimed at AI data-center growth outpacing generation and grid build. The structural features mirror the risk-shifting logic this series has tracked all year — roughly a third require minimum payments over a set term, more than half include collateral or other credit-risk protections, and about half impose early-exit fees.
Why this is the week’s most usable item for a business case: every avoided-cost argument for demand-side resources rests on discounting speculative load forecasts and refusing to over-build against them, and that argument has until now been made in the conditional. AEP Ohio converts it to the indicative. A two-fold-plus over-statement of “firm” large load at a single utility, corrected only because customers were asked to post money, is the single most concrete illustration available that an unscreened large-load forecast is not a planning input. Any resource plan that compares incremental, cancelable DR/DSM/DER against long-lived generation and transmission CapEx should carry the AEP Ohio revision in its evidence file, because it quantifies the risk premium that the demand-side alternative is being credited for avoiding.
The caveat is definitional and should be stated whenever the figure is used: these are financial-screening tariffs, not resource acquisitions. They add no countable demand-side capacity. Their contribution to a filing is de-risking the load forecast — which is to say, they change the denominator rather than the numerator, and a utility that claims the 17 GW as avoided demand rather than as removed speculation will be corrected. (Source: Utility Dive, September 24 — September 25 entry)
NERC’s three computational-load standards emerged from their first comment period with an applicability test FERC declined to write: 50 MW or more, and connected at 100 kV or above — a continent-wide floor that now sits beneath every state’s large-load definition, though it reaches no operator until a separate registry proceeding lands in December. This is the concrete rulemaking implementing FERC’s July 16, 2026 order directing mandatory reliability rules for computational loads, and it is the watch item this series flagged as the top next-run priority when the comment window closed. The record: CLO-001-1 (interconnection, studies and modeling data), CLO-002-1 (operational data and communications) and CLO-003-1 (protection coordination and disturbance monitoring); the first formal comment period and its 10-day ballot closed September 18. The drafts create two entity types absent from NERC’s current glossary — Computational Load Owner and Computational Load Operator — and CLO-003-1 carries concrete instrumentation duties, including fault recording at 64 or more samples per cycle, at least a 2-cycle pre-trigger and at least a 2.0-second record length.
Two features determine how much this actually binds, and both cut toward “not yet.”
First, the standards reach an operator only once it is on the NERC registry, and the registry change rides a separate Rules of Procedure track: a 20-day comment period in October, a five-day final ballot in November, board adoption in December — the same month as the December 31 FERC filing deadline. If that track slips, the standards exist and apply to nobody.
Second, the enforcement weight is deliberately light: every CLO-003-1 requirement carries the mildest Violation Risk Factor (Lower) and a Long-term Planning time horizon, and the standards authorization request itself calls cost impacts “unknown” but “significant.”
The part that matters most for DER work is the threshold arithmetic. A continent-wide 50-MW floor now sits beneath a patchwork of state and RTO definitions — ERCOT’s NOGRR282 at 75 MW, CAISO at 50 MW, MISO above 50 MW with a separate 25-MW IT-demand computational test — which narrows, for the first time, the asymmetry between how large loads and generators are held to reliability account. That asymmetry is the same one that has allowed interruptible large loads and DR to be treated as reliability resources on different evidentiary standards than generation. Narrowing it cuts both ways: it legitimizes demand-side reliability contributions, and it subjects them to the same registry-and-audit machinery. The standing screen this series established holds — maintain one reconciled large-load customer list, sized by the lowest applicable threshold — and the 50 MW / 100 kV pair is now a column in it. (Source: Mgrid.org, September 19 — September 25 entry; the draft standards were previously reported September 4 and tracked September 11 and September 18)
Texas finalized the rules that turn any load above 75 MW into a curtailable reliability resource — and softened the money while leaving the curtailment obligation intact, which tells you where the bargaining power sits even in the middle of a queue crisis. The PUCT voted September 19 (reported September 21) to adopt final large-load interconnection rules under Senate Bill 6, dropping the proposed non-refundable $50,000/MW interconnection fee in favor of a flat $100,000 study fee regardless of load size, halving financial security to $50,000/MW of requested peak demand, and extending the window for notifying ERCOT of a customer’s non-utilized capacity after a missed energization milestone from as little as six months to a full 24 months, applied to the energization schedule as a whole rather than per milestone.
Those are meaningful concessions, and they arrive while Texas has paused new data-center interconnections so ERCOT can audit a 474-GW queue that is roughly 90% data centers. A commission with that much leverage settling for a flat study fee is a data point about who needs whom.
The provision that matters to this series’ business case is unchanged and is the strongest in the country: large loads must disclose on-site backup generation capable of serving at least 50% of demand, ERCOT may direct them to curtail or self-supply in emergencies, and the PUCT has affirmed that emergency curtailment is not capped by the paired behind-the-meter generator’s capacity. That is conscription of large loads into precisely the firm, dispatchable role this series values DR and VPP capacity against — and it both validates and competes with utility-procured demand-side capacity. If a commission can obtain curtailment from a data center as a condition of service, the marginal value of buying an equivalent megawatt of curtailment from a customer falls; if the obligation proves unenforceable or is never exercised, the procured version is the only one with a settlement record. Texas is now the jurisdiction where that question gets answered empirically. (Source: Utility Dive, September 21 — September 23 entry)
Two utilities asked South Dakota to approve up to $1.93 billion of transmission for data-center load that chose other states — and the filing is the cleanest speculative-CapEx exhibit this series has acquired, because the applicants’ own framing concedes the load is elsewhere. Xcel Energy and Otter Tail Power filed on September 22 for roughly 160 miles of new line before the South Dakota PUC. The larger is the jointly filed PowerOn Midwest Project — about 90 miles of 765-kV line from the Big Stone South Substation toward Minnesota, estimated at $1.3–1.7 billion for the South Dakota segment. Otter Tail’s separate line is 71 miles of 345-kV toward North Dakota at $191–229 million. Both target a 2034 in-service date.
The mismatch is the exhibit. South Dakota hosts no hyperscale data centers; the Upper Midwest growth driving the demand case landed in North Dakota and Minnesota. The state is being asked to host a corridor for load that has not committed to it, and cost allocation is unsettled in the applications. Execution risk is concrete and quantified: the 765-kV route holds easements on only 81 of the 274 parcels it needs — 30% — and Otter Tail’s line does not report its parcel count at all. A commission decision is not expected until late 2027 or early 2028.
This is the transmission analog of the North Carolina generation denial in the Regulatory section below, and the two should be cited as a pair: capital sized to an uncontracted, out-of-state load forecast, on a 2034 in-service date, against cancelable resources that can be sited where the load actually appears. Note also the eight-year gap between filing and service — for a demand-side alternative, the relevant comparison is not cost per kilowatt but cost per kilowatt available in 2028, and nothing in this filing is available before 2034. (Source: Mgrid.org, September 22 — September 25 entry)
FERC ruled on the ComEd / PowerHouse Hillwood security dispute this series reported at the filing stage in August, and left the agreement in place over a credit posting of one dollar — with a sitting commissioner calling that posting an “embarrassing legal fiction.” On September 22 FERC rejected Commonwealth Edison’s “notice of cancellation” of the transmission security agreement for the 1.8-GW, $20-billion data center PowerHouse Hillwood is developing in Joliet, Illinois, leaving the agreement in force while declining jurisdiction over the underlying contract fight, which is pending in the U.S. District Court for the Northern District of Illinois. The dispute turns on credit support: the developer contended it satisfied the agreement’s initial credit requirement via a $1 posting, prompting Commissioner David LaCerte to say the proposition that a dollar “may provide appropriate security” is an embarrassing legal fiction that trivializes the guarantee’s purpose.
This closes the loop on the August 28 report, where this series characterized transmission security agreements as having become “the primary instrument for weeding speculative load and allocating interconnection cost,” and noted that the instrument itself was in litigation. It still is — but now with a ruling that keeps an under-collateralized agreement alive and a public statement from the bench that the collateral is nominal. Read against AEP Ohio’s 17-GW haircut at the top of this section, the pair is a natural experiment: where the financial requirement is real, forecast megawatts disappear; where it is a dollar, they persist on the books and continue to justify capital.
For utilities and their ratepayers the exposure is the familiar one — an under-collateralized large-load commitment is exactly what produces stranded cost when a data center slips or cancels — and it is the risk profile against which incremental, cancelable demand-side resources score well. The concrete action is narrower than the policy point: any load-forecast line item backed by a security agreement whose posted collateral you have not personally verified is an assumption, not a commitment, and the Exelon 40% pipeline haircut this series reported in August is the precedent for what happens when the verification is finally done. (Source: Utility Dive, September 22 — September 24 entry; the Hillwood/ComEd filings were reported in the August 28 digest at the filing stage)
Microsoft moved to keep alive a challenge to Virginia’s mandatory-CIAC order, which is the state-level test of whether a regulator can assign but-for transmission cost directly to the load that causes it — and Dominion’s compliance filing is due this month. On July 31 the Virginia State Corporation Commission ordered Dominion Energy to amend its line-extension policy (Rider T-1) to require a mandatory contribution in aid of construction for defined “direct-connect” transmission facilities — the substations and radial lines that exist only to serve a specific large load — and to file the amended policy within 90 days, putting the compliance filing in October. Microsoft filed an August 28 notice of appeal with the Virginia Supreme Court and told Utility Dive on September 18 that it has not formally appealed but is “preserving its ability” to challenge implementation, with a petition-for-appeal deadline in late November.
The nuance that planners should take from this is the line the hyperscalers are drawing themselves. Google — an intervenor alongside Amazon and Meta — argued it is “comfortable” with mandatory CIAC for direct interconnection costs (radial lines and substations) but wants a separate proceeding for general network upgrades. That distinction determines how much data-center-driven transmission cost is directly assigned versus socialized, and therefore how much avoidable CapEx a demand-side alternative can be credited against in an avoided-cost calculation. A CIAC that makes but-for transmission a direct cost of the triggering load is the mechanism that surfaces deferred-investment value; a network-upgrade carve-out is the mechanism that buries it back in the rate base.
The open question — whether the SCC’s approach extends upstream to network upgrades — is what to watch in Dominion’s October docket, and it is the same question the FERC cost-allocation fight in South Dakota leaves unsettled in a different jurisdiction. (Source: Utility Dive, September 18 — September 21 entry)
📋 Regulatory & Policy
A state commission denied a 255-MW gas turbine because the data-center load behind it was “insufficiently reliable” to act on — and that sentence is the single most valuable acquisition of the week for anyone writing a demand-side business case, because it is a regulator saying in an order what demand-side advocates have been saying in testimony. On September 18, reported September 22, the North Carolina Utilities Commission denied Duke Energy Progress’s request for a 255-MW hydrogen-capable F-class simple-cycle combustion turbine at the Smith Energy Complex in Richmond County. Public staff testimony called the $584 million price tag “staggering.” The commission found that “much of the growth appears to be based upon anticipated data center customer additions” that were “insufficiently reliable for the Commission to act at this point,” and explicitly weighed Duke’s commitment to the White House Ratepayer Protection Pledge — the voluntary promise not to shift data-center costs onto other customers — holding that the record did not establish how the turbine squared with it.
Three features make this stronger precedent than a typical denial.
It happened over staff support. NCUC public staff had reluctantly supported the turbine as “the only resource that is executable on a constrained timeline,” warning that Duke risks violating a 22% planning reserve margin by 2031. The commission denied it anyway, declining to pre-approve generation ahead of its 2025 Carolinas Resource Plan decision, expected late 2026. A denial that survives a reliability warning from the commission’s own staff is not a close call about need; it is a ruling about evidentiary quality.
It happened to the largest capital program in the industry. The turbine sits inside Duke’s $103-billion capital plan, the largest on file at any regulated U.S. utility as of February 2026. If need-and-cost scrutiny reaches that program, it reaches every program.
It folded cost-causation into a siting decision. The pledge is voluntary and has no tariff force, and the commission used it as a test of whether the project was consistent with the utility’s own representations about who pays. That is a procedural move other commissions can copy immediately, because it requires no new statute.
For IRP and avoided-cost work the consequence is direct and should be written into the next filing rather than saved: when a commission discounts a speculative large-load forecast and rejects expensive peaking gas, the comparison the demand-side portfolio has to win changes. It is no longer “is DR cheaper than a CT” — the CT was denied. It is “what fills a reserve-margin gap on a constrained timeline without asking the commission to bank on a forecast it has already refused to bank on,” and incremental, cancelable, fast-to-deploy demand-side capacity is the only class of resource whose answer to that question does not depend on the disputed forecast. Note the timing risk in the other direction as well: the 2025 CPIRP decision is expected late 2026, inside the next quarter, and a utility that has no demand-side alternative on the record when the commission revisits the resource mix will have the gap filled some other way. (Source: Utility Dive, September 22 — September 24 entry)
California wrote large-load cost causation into statute across seven bills, and one of them pays data centers to put clean generation behind their own meter — which is the first state law this series has reported that treats a large load’s own DER as a policy objective rather than a byproduct. On September 21, reported September 22, Gov. Newsom signed seven data-center bills. SB 1168 directs the CPUC to examine data-center energy use and ensure the facilities pay for transmission, distribution and load-increase costs. SB 886 and SB 887 require data centers to bear all infrastructure and generation costs, take a larger share of wildfire mitigation cost and liability, and undergo environmental review. AB 1577 mandates monthly energy-consumption and efficiency reporting to the CEC. AB 2383 incentivizes data centers to use onsite clean-energy resources. AB 2469 and AB 2619 add water-use assessment and reporting. California hosts 296 data centers, third-most in the country, and the package follows Los Angeles County’s temporary ban on large-scale AI data-center development in unincorporated areas.
Two provisions deserve separate treatment because they do different work. The cost-causation bills (SB 1168/886/887) are the lever this series tracks: cost-allocation rules determine how much avoided cost a demand-side or distribution-sited resource can be credited against, and statutory assignment removes the argument about whether the assignment is permissible. AB 1577’s monthly reporting is the underrated one — a monthly energy-use and efficiency series filed with a state agency is the closest thing to a public load-shape dataset for this customer class that any state has created, and load shape is what determines whether a large load is a flexibility candidate or just a big number in a forecast. AB 2383 is the flexibility nudge, and it points somewhere specific: behind-the-meter generation at a data center reduces the load the shared grid must serve, which is the same arithmetic as MISO’s ZGIA fast track reported earlier this month — and it therefore competes with utility-procured flexibility on speed-to-power while removing load from the forecast the utility is planning against.
The caveat is implementation. These are enabling laws whose teeth depend on CPUC and CEC proceedings, so their value to an IRP is directional — cost-allocation certainty and a coming data source — rather than an immediate resource count. Watch which docket the CPUC opens under SB 1168 and whether AB 1577’s reporting is published or held confidential, because a monthly load series that only the agency can see is worth much less than one an intervenor can cite. (Source: Utility Dive, September 22 — September 24 entry)
The U.S. House passed a federal large-load cost-causation standard 417–3, and the provision worth reading twice is the one that reaches power supply rather than wires — because that is what converts “bring your own capacity” from a negotiating posture into a statutory expectation. The Ratepayer Protection Act passed the House on September 16 in a 417–3 vote, co-sponsored by Rep. Gabe Evans (R-Colo.) and Rep. Kathy Castor (D-Fla.). It would require state utility regulators and unregulated utilities to “consider” adopting standards for loads larger than 100 MW — guaranteed recovery of the “full, incremental” cost of any upgrade needed to serve the load, financial assurances before grid upgrades are made, and cost recovery if a large load exits its power-supply contract early — with one year to begin considering and two years to decide.
The consequential detail is scope. Interconnection agreements typically cover transmission and distribution only; this bill’s inclusion of power-supply costs implies an obligation to bring capacity, not just to pay for wires. That expectation can add years to a data-center timeline, and every year it adds raises the option value of fast, firm demand-side capacity that a utility can commit on its own schedule. That is the mechanism by which a cost-causation statute becomes a DR business-case input rather than a ratepayer-equity headline.
Two sober qualifications. ClearView Energy Partners called the bill “somewhat behind the regulatory curve” because it would “largely reinforce” a transition already underway — only 13 states lack any large-load tariff per SEPA’s tracking — and judged Senate passage unlikely before the November midterms. And the operative verb is “consider,” not “adopt”: a duty to consider produces a docket, not a tariff. What the vote supplies is not law but citable direction — a 417–3 bipartisan margin is the strongest available evidence that cost-causation discipline on large loads is now the mainstream position, which is exactly the kind of assertion an intervenor otherwise forces you to prove. (Source: Utility Dive, September 17 — September 21 entry)
Pennsylvania is rewriting the rules that decide who gets curtailed first — and that document, not the incentive level, is what determines whether a utility’s acquired demand response is actually dispatchable. On September 11 the Pennsylvania PUC voted 5-0 to direct its Law Bureau and Bureau of Technical Utility Services to propose updates to the state’s emergency load-control framework — the order, criteria and applicability of curtailment — with a tentative order due at the October 1 meeting, public comment to follow, and a targeted January 28 final vote. This series referenced the September 11 motions last week as context for the Pennsylvania resource-adequacy study; this is its first full treatment as an item, and it deserves one because load-control rules are the regulatory scaffolding beneath every curtailment-based demand resource.
The commission tied the action directly to resource adequacy, citing PJM’s 6,831-MW reliability shortfall for the 2028-29 delivery year from the July 2026 capacity auction, and an August PUC report projecting 18.56% industrial load growth over five years — driven “primarily” by PPL’s large-load forecast — even though total electricity usage declined from 2024 to 2025. That last clause is worth keeping: a state whose overall consumption fell while its industrial forecast rose 18.56% is describing a forecast driven by one customer class, which is the condition under which forecast screening matters most.
The commission is also convening a fall technical conference on how the costs of new large loads are identified and allocated — the same cost-causation question California just legislated and Virginia is litigating.
The concrete reason to engage rather than watch: curtailment priority is where enrolled demand response either counts or does not. If the rewritten framework curtails firm load before it curtails a large load operating without its own supply, then every megawatt of DR a utility has paid for is called ahead of the resource whose growth created the shortage — which inverts the cost-causation principle the same commission is separately advancing. This series has flagged that sequencing check as an action item since August; Pennsylvania has now opened the document where it gets decided, on a schedule that closes in January. (Source: Utility Dive, September 11 — September 23 entry; referenced as context in the September 18 digest)
The Department of Energy issued Duke Energy Carolinas its second Section 202(c) emergency order in three weeks — seven days after an appellate court vacated a structurally identical order in Michigan. Order No. 202-26-48 took effect September 18 and ran through September 21, authorizing Duke Energy Carolinas and its transmission owners to dispatch specified units and activate backup generation as a last resort before or during an Energy Emergency Alert Level 3. It followed Order No. 202-26-43 for the same territory (September 3–8) and a June order referencing anticipated load of roughly 34,589 MW. Last week’s digest recorded the seventh §202(c) order of 2026 to PJM; the Carolinas footprint now has its own cadence.
The legal position is what makes the cadence citable rather than merely notable. On September 11 the D.C. Circuit vacated the first §202(c) order forcing Consumers Energy’s 1,420-MW J.H. Campbell plant to delay retirement, holding that DOE exceeded its authority by blocking market-driven retirements without demonstrating an actual emergency. Seven days later DOE issued this one. Section 202(c) usage now runs at roughly 40 times the prior annual average, and an Earthjustice-cited study estimates that broad stay-open extensions of retiring fossil plants through 2028 could add more than $3 billion in utility-customer costs.
This series’ argument on the point is unchanged and this order reinforces it without modification: every planned megawatt of dispatchable DR or VPP capacity is a megawatt that need not be summoned through an instrument that pays nothing, proves nothing, produces no settlement record, and has now been narrowed by an appellate court on the generation side. The framing for a filing is the accurate one rather than the rhetorical one — emergency authority is being invoked at a frequency that makes it a plannable pattern, not an exception, and a resource plan that treats it as the reliability backstop is relying on an authority whose legal ceiling was tested and found this month. (Source: Discovery Alert, September 19 — September 22 entry; the Campbell vacatur and the PJM order cadence were reported in the September 18 digest)
Carried — FERC / ISO-NE Docket ER26-1956. Fully reported in the September 11 digest: FERC’s May 29 order restored Network Resource Capability and Capacity Network Resource Capability qualification pathways for DERs that an Order No. 2023 compliance change had silently closed, leaving state-interconnected resources with no method at all to establish Forward Capacity Market credentials. Carried here for the one detail the earlier report did not include: the same order granted a one-time 2.5-year extension of the three-year commercial-operation-resumption window for resources on forced outage that face long-lead-time equipment procurement — a supply-chain accommodation that matters specifically to storage developers whose interconnection position is otherwise forfeited by a transformer lead time. The broader clock is unchanged: Order 2222 energy and ancillary-services market rules effective November 1, 2026 across most RTOs, ISO-NE capacity-market participation around February 1, 2027, MISO phased to 2027–2029. (Source: Troutman Pepper Locke, June 2026 — September 23 entry; previously reported in the September 11 digest)
🔬 EPRI Research Spotlight
A direct check confirms no new EPRI publication dated inside this week’s window entered either the research log or EPRI’s own press-release index — the eleventh consecutive dry week. A targeted search of EPRI’s media resources this week surfaced nothing dated September 21–25; the most recent substantive items remain the DCFlex nine-site expansion and the Aurora materials this series has carried since July, alongside the program’s standing data-center-demand studies. As established two months ago, this is reported as a finding about publication cadence, not as a coverage omission: EPRI’s substantive data-center-flexibility output arrives through the DCFlex program and its partners’ channels rather than through standalone research releases. A utility using EPRI publications as its flexibility-research feed is monitoring the wrong surface — the material is in the demonstration coalition and its members’ announcements.
This week’s items make the FlexMosaic gap sharper rather than narrower, and the count is now six. The DCFlex Initiative and its FlexMosaic framework — organized around flexibility classes keyed to notification time, duration, frequency, depth and speed of response — remain the only published vocabulary precise enough to price competing flexibility products against one another. Last week this series counted five distinct flexibility products in play, procured in five venues with no shared grammar. Add this week’s:
- Virginia’s 450-MW dual-enrollment VPP — aggregated, dispatchable, two possible counterparties against one meter.
- PJM’s proposed IRAS — curtail-or-switch-to-backup near emergencies, prioritized by registry.
- Interruptible riders in roughly a quarter of large-load tariffs — bilateral, tariffed curtailment, utility-specific notification and duration.
- The §202(c) backup-generation transition — a 15-minute compulsion, no compensation, no telemetry requirement, no settlement record. Invoked again this week (Duke Carolinas, 202-26-48).
- The PJM DR 24/7 availability obligation effective June 1, 2027 — continuous obligation rather than event response.
- NEW — Texas SB 6 mandatory large-load curtailment, finalized this week: backup generation serving at least 50% of demand must be disclosed, ERCOT may direct curtailment or self-supply in emergencies, and the obligation is not capped by the behind-the-meter generator’s capacity.
Six response characteristics, six venues, one vocabulary that none of them uses. A utility holding committed capacity in more than one of these cannot answer “how much flexibility do we have, and under what notification?” without a mapping exercise — and item 6 is the hardest of the six to map, because an uncapped curtailment obligation has no defined depth, which is one of FlexMosaic’s five keys.
The Aurora timing problem is now fully consequential, and this week closed the argument about it. The coalition — NVIDIA, Emerald AI, EPRI, Digital Realty and PJM — still targets late 2026 for the 96-MW Manassas, Virginia facility, having originally slated it for the first half of the year; this week’s verification search surfaced nothing indicating acceleration or further slip. Set that beside the NERC CLO record reported above: the standards’ first comment period and ballot closed September 18, board adoption is scheduled for December, and the FERC filing deadline is December 31 — so the standards will be filed before the reference design that was supposed to supply demonstration data on useful data-center flexibility is operating. Pair that with the only large-scale empirical record that does exist — the 3,800-MW northern Virginia load trip of July 22, where the data centers disconnected too early — and the position is now unambiguous rather than merely uncomfortable: computational-load reliability standards are being written on modeling, and the one measured event on the record is a mass disconnection. The coalition’s own estimate is that the reference design, adopted nationwide, could unlock roughly 100 GW on the existing system. That estimate is the reason the sequencing matters.
Standing watch: Aurora commissioning data and any further slip; the CLO-001-1/-002-1/-003-1 ballot results this fall and whether the separate Rules of Procedure registry track holds its October comment / November ballot / December adoption schedule; whether FlexMosaic or any response-class vocabulary appears in the ER26-3515 and RM26-4 compliance records; whether the EPRI/Open Compute Project standardized utility–data-center agreement produces publishable contract language; the EPRI/Dominion V2G demonstration’s 12-volt state-of-charge and feeder-hosting findings as school-bus fleets scale; and an EPRI.com direct check again next run — twelfth week.
🚩 Utility-Sector Relevance Flags
⚑ A Commission Denied a $584M Gas Turbine Because the Data-Center Load Behind It Was “Insufficiently Reliable” — Over Its Own Staff’s Support
Topic: Generation-CapEx Denial / Load-Forecast Skepticism / DSM Business Case
Relevance: The NCUC denied (September 18, reported September 22) Duke Energy Progress’s 255-MW hydrogen-capable F-class CT at the Smith Energy Complex, with staff calling the $584M cost “staggering” and the commission finding growth “much…based upon anticipated data center customer additions” that were “insufficiently reliable for the Commission to act at this point.” It tied the denial to Duke’s White House Ratepayer Protection Pledge commitment. The denial came despite public staff reluctantly supporting the CT as “the only resource that is executable on a constrained timeline” and warning of a breach of a 22% planning reserve margin by 2031, and it sits inside Duke’s $103-billion capital plan — the largest at any regulated U.S. utility. The 2025 CPIRP decision is expected late 2026.
Action Signal: Implement — Put this order in the standing evidence file and re-frame the comparison in the next filing. The question is no longer whether DR beats a CT on levelized cost; the CT was denied. It is what fills a reserve-margin gap on a constrained timeline without asking a commission to bank on a forecast it has already refused to bank on — and incremental, cancelable, fast-to-deploy demand-side capacity is the only resource class whose answer does not depend on the disputed forecast. Note the reciprocal risk: with the CPIRP decision inside the next quarter, a utility with no demand-side alternative on the record will have the gap filled some other way. Also copy the procedural move — the commission used a voluntary cost-shift pledge as a test of consistency, which requires no new statute in any other state.
⚑ AEP Ohio Cut 17 GW Out of Its Large-Load Forecast Because Customers Were Asked to Post Money — Use It as the Risk Premium Your Demand-Side Case Avoids
Topic: Load-Forecast Screening / Avoided-Cost Denominator / Speculative CapEx
Relevance: 65-plus large-load tariffs are now proposed or approved across 30-plus states, and AEP Ohio cut its large-load forecast by more than half — 30 GW to 13 GW — after Ohio regulators approved one. Roughly a third require minimum payments over a set term, more than half include collateral or credit protections, about half impose exit fees. The counterexample in the same week: FERC left a 1.8-GW, $20B ComEd agreement in force on a $1 credit posting, with Commissioner LaCerte calling that an “embarrassing legal fiction.” And the object lesson: Xcel + Otter Tail seek up to $1.93B for ~160 miles of South Dakota transmission — including 90 miles of 765-kV at $1.3–1.7B, easements on 81 of 274 parcels — for hyperscale load that landed in North Dakota and Minnesota, in service 2034, decision late 2027/early 2028.
Action Signal: Implement — Carry the AEP Ohio revision as the quantified risk premium a cancelable resource earns. Two disciplines follow. First, state the limit when you cite it: these are financial screens, not acquisitions — they change the denominator, not the numerator, and claiming 17 GW as avoided demand will be corrected. Second, audit your own forecast the way Ohio’s tariff did: any large-load line item backed by a security agreement whose posted collateral you have not personally verified is an assumption. Exelon’s ~40% pipeline haircut is the precedent for what the verification finds. For transmission comparisons, price the demand-side alternative as cost per kilowatt available in 2028 — nothing in the South Dakota filing is available before 2034.
⚑ 580 MW Was Dispatched From Home Batteries in One Evening — Book It as Proof of Performance, Not Proof of Capacity
Topic: VPP Operational M&V / Avoided-Cost Evidence / DERMS Dispatch Design
Relevance: On September 9 (reported September 21) Sunrun and Tesla discharged >580 MW into CAISO over a three-hour window — 517 MW Powerwalls, 63 MW other batteries, from ~110,000 Powerwalls plus 30,000+ third-party units across all three California IOUs — coordinated via CEC DSGS and CPUC ELRP, triggered at a day-ahead LMP above $200/MWh; a September 10 SCE-only event added >140 MW (combined potential >720 MW). Brattle (commissioned by Sunrun/Tesla) puts California VPP net savings at up to $206M by 2028. Two limits: this is a voluntary-program dispatch by two aggregators, not an accredited resource-adequacy commitment — the standing watch item on converting the 16.8-GW framework to credited MW at a stated ELCC is not closed by an event; and the operators flagged the “shock to the system” of ramping that much capacity at once and want a smoother ramping curve.
Action Signal: Implement — Two distinct uses. In filings, cite this where dispatched performance is demanded and keep the capability-versus-capacity distinction this series has enforced all year: metered dispatch evidence for DR, “capability with a timeline” for VPP aggregations. In procurement, write ramp shaping into the DERMS requirements document with acceptance criteria — a platform that can call 580 MW but cannot shape the ramp relieves one disturbance by creating another. That fix also lands squarely on the organizational blocker Itron names: the customer-program team’s dispatch rules and the grid-operations team’s stability limits have to become one document.
⚑ NERC Set a Continent-Wide 50 MW / 100 kV Floor — and It Binds Nobody Until a Separate Registry Track Lands in December
Topic: Large-Load Reliability Standards / Registry Timing / Compliance Screening
Relevance: CLO-001-1, -002-1 and -003-1 closed their first comment period and 10-day ballot on September 18, implementing FERC’s July 16 order (filing deadline December 31). Applicability is a two-part test FERC declined to name: total connected load ≥50 MW and supplying equipment at 100 kV or above. Two new registry entity types are created — Computational Load Owner and Computational Load Operator — and CLO-003-1 mandates fault recording at ≥64 samples/cycle, ≥2-cycle pre-trigger, ≥2.0-second records. But the standards reach an operator only once it is on the NERC registry, and that change rides a separate Rules of Procedure track: 20-day comment in October, five-day final ballot in November, board adoption in December. Enforcement weight is deliberately light — every CLO-003-1 requirement carries VRF Lower and a Long-term Planning horizon — and the authorization request calls cost impacts “unknown” but “significant.”
Action Signal: Watch — Add 50 MW / 100 kV as a column in the single reconciled large-load customer list this series has recommended since August, sized by the lowest applicable threshold: ERCOT NOGRR282 ≥75 MW, CAISO ≥50 MW, MISO >50 MW plus a 25-MW IT-demand test, TVA 5 MW. Then track the Rules of Procedure schedule specifically rather than the standards — if the registry slips, the standards apply to no one and the December 31 filing is a paper milestone. Note the two-sided consequence for demand-side work: narrowing the large-load/generator accountability asymmetry legitimizes demand-side reliability contributions and subjects them to the same registry-and-audit machinery.
⚑ Texas Finalized Mandatory Large-Load Curtailment — It Both Validates and Competes With Procured Demand Response
Topic: State Curtailment Regime / DR Substitution Risk / Interconnection Terms
Relevance: The PUCT adopted final SB 6 large-load rules September 19 (reported September 21), dropping the proposed $50,000/MW non-refundable fee for a flat $100,000 study fee, halving financial security to $50,000/MW of requested peak, and extending the missed-energization notice window to ERCOT from ~6 months to 24 months. The curtailment regime is intact and is the strongest in the country: loads >75 MW must disclose backup generation serving ≥50% of demand, ERCOT may direct curtailment or self-supply in emergencies, and curtailment is not capped by the paired behind-the-meter generator’s capacity. Context: a state-ordered pause on new data-center interconnections while ERCOT audits a 474-GW queue that is ~90% data centers.
Action Signal: Engage — Treat compelled large-load curtailment as a competing supply of the same product you are buying. Where a commission can obtain curtailment as a condition of service, the marginal value of purchasing an equivalent megawatt falls — so quantify your procured DR’s advantage on the axes the mandate cannot deliver: defined depth, defined notification, settlement-grade M&V and a reusable telemetry record. Texas is where this gets settled empirically, so watch whether the obligation is ever exercised; an unexercised mandate leaves the procured version as the only instrument with a performance record, which is precisely the position the §202(c) load-side instrument is in after a year of orders.
⚑ Nevada Got the Avoided-Cost Method Right and the Market Design Narrow — Build the DERMS for the Reversal
Topic: DR/DSM Avoided-Cost Methodology / Make-vs-Buy / Enrollment Channels
Relevance: PUCN’s August 11 order (reported August 26) approves NV Energy’s Energy Grid Services and Capacity Grid Services riders, replacing a 15-year-old DR framework and paying load reduction at the avoided cost of generation and transmission plus dispatched energy at hourly market prices — a textbook value stack. The commission declined to let customers assign compensation to third-party aggregators or OEMs, declined to require an open data-sharing platform, and declined to add a VPP resource type to the next IRP (“insufficient information”); the aggregator question is deferred to NV Energy’s CAISO EDAM docket, with EDAM entry ~late 2028. Advocates (AEU, SEIA, SUN) warn the utility-only design suppresses behind-the-meter battery attachment in a high-solar, low-storage state after the residential-battery ITC expired at end-2025. Four live designs now exist: Virginia dual-enrollment, Minnesota utility-only, New Jersey undecided, Nevada utility-only with a two-year deferral.
Action Signal: Engage — The enrollment-channel decision is made by a commission and can be reversed by a later docket, so specify the platform to survive the reversal: IEEE 2030.5 aggregator interfaces and a settlement path that can attribute performance to either counterparty, as pass/fail requirements rather than scoring bonuses, even where your commission approved utility-only this year. Retrofitting an aggregator channel happens at the data model, which is the most expensive place to find the gap. Separately, borrow Nevada’s tariff language — avoided cost of generation and transmission for load reduction, hourly market price for dispatched energy — because it is now an approved formulation in a vertically integrated jurisdiction, which is harder to attack than a consultant’s method.
⚑ Pennsylvania Is Rewriting Who Gets Curtailed First, With a Tentative Order on October 1 and a Final Vote Targeted January 28
Topic: DR Dispatchability / Emergency Load Control / Cost Allocation
Relevance: The PA PUC voted 5-0 on September 11 to rewrite the state’s emergency load-control framework — order, criteria and applicability of curtailment — with a tentative order due October 1, comment to follow, and a targeted January 28 final vote, plus a fall technical conference on large-load cost identification and allocation. It cited PJM’s 6,831-MW shortfall for 2028-29 and an August report projecting 18.56% industrial load growth over five years led by PPL’s large-load forecast — while total state electricity usage declined from 2024 to 2025.
Action Signal: Engage — File in this docket, and file on one question: curtailment priority. If the rewritten framework curtails firm load before it curtails a large load operating without its own supply, every megawatt of DR the utility has paid for gets called ahead of the resource whose growth created the shortage — inverting the cost-causation principle the same commission is separately advancing in its cost-allocation conference. This series has carried that sequencing check as an action item since August; Pennsylvania has now opened the document where it is decided, and the window closes in January. Utilities outside PJM should read the tentative order anyway on October 1 — a state that rewrites curtailment priority against a quantified RTO shortfall is producing the template.
⚑ California Legislated Large-Load Cost Causation and Ordered Monthly Data-Center Energy Reporting — The Reporting Bill Is the Underrated One
Topic: Cost Allocation / Load-Shape Data / Behind-the-Meter DER Incentive
Relevance: Seven bills signed September 21 (reported September 22). SB 1168 directs the CPUC to ensure data centers pay for T&D upgrades and load increases; SB 886/887 require full infrastructure and generation cost recovery, a larger wildfire cost and liability share, and environmental review; AB 1577 mandates monthly energy-use and efficiency reporting to the CEC; AB 2383 incentivizes onsite clean-energy resources; AB 2469/2619 add water assessment and reporting. California hosts 296 data centers (third-most), and the package follows LA County’s temporary ban in unincorporated areas.
Action Signal: Watch — Track two things, not the headline. First, AB 1577’s implementation: a monthly per-facility energy-use and efficiency series is the closest thing to a public load-shape dataset for this customer class any state has created, and load shape is what separates a flexibility candidate from a large number in a forecast — so watch whether the CEC publishes it or holds it confidential, because an intervenor-citable series is worth far more than an agency-only one. Second, AB 2383’s direction of travel: incentivized behind-the-meter generation at a data center removes load from the shared-grid forecast, which is the same arithmetic as MISO’s ZGIA fast track and therefore competes with utility-procured flexibility on speed-to-power. These are enabling laws; their value to an IRP is cost-allocation certainty and a coming data source, not a resource count.
⚑ FERC Left a 1.8-GW Agreement Alive on a $1 Collateral Posting — Verify the Collateral Behind Every Large-Load Line Item
Topic: Large-Load Credit Support / Stranded-Cost Risk / Forecast Integrity
Relevance: On September 22 FERC rejected ComEd’s notice of cancellation of the transmission security agreement for the 1.8-GW, $20-billion PowerHouse Hillwood data center in Joliet, IL, leaving the agreement in place and declining jurisdiction over the underlying contract fight (pending in the N.D. Ill.). The developer contended it met the initial credit requirement via a $1 posting; Commissioner LaCerte called the idea that $1 “may provide appropriate security” an “embarrassing legal fiction.” This series reported the filings in the August 28 digest and characterized transmission security agreements as the primary instrument for weeding speculative load; the instrument remains in litigation, now with a ruling that keeps an under-collateralized agreement alive.
Action Signal: Watch — Read this beside AEP Ohio as a natural experiment: where the financial requirement is real, forecast megawatts disappear; where it is a dollar, they stay on the books and keep justifying capital. The operational consequence is a diligence step, not a policy position — for each large-load entry in the forecast, record the security instrument, the posted amount and the date verified, and treat unverified entries as assumptions in sensitivity analysis rather than as load. Under-collateralized commitments are the mechanism that converts a canceled data center into a stranded cost, which is the risk profile against which incremental, cancelable demand-side capacity scores well — but only if the exposure is quantified rather than asserted.
⚑ Massachusetts Paid Between $56 and $764 per kWh Inside One Program — That 13× Spread Is the Avoided-Outage Number Nobody Has Written Down
Topic: Resilience Valuation / Cost-Effectiveness Screening / BTM Storage
Relevance: Massachusetts DOER awarded >$40M to 36 projects on September 22 under the Advancing Massachusetts Power Energy Storage Grant Program. The 28 Community Resilience awards total 10.26 MW / 44.14 MWh for $32.4M at municipal buildings, schools and affordable/public housing (largest single award $2.5M), with a “vast majority” to environmental-justice communities. State dollars range from ~$56/kWh for a 4-MW/80-MWh long-duration award to ~$764/kWh for a small school system — a ~13× spread with no published cost-effectiveness screen. These are grant contributions to municipal resilience assets, not dispatchable resources; their IRP capacity value is limited unless enrolled into DR/VPP — and Massachusetts has both an enrollment path (ConnectedSolutions) and a 3.5-GW by 2035 demand-management target whose September baseline inventory is running now.
Action Signal: Implement — Use the spread as the opening for the work this series’ framework is built to do: produce a defensible value-of-avoided-outage-hour figure for your own service territory, because 28 procurements across a 13× range imply a number the state has not published and cannot defend. Then answer the definitional question before a regulator asks it: can a state-grant-funded battery also earn performance payments, and do these 10.26 MW count toward the 3.5-GW target? Whatever Massachusetts decides in the baseline inventory will be copied, and a utility that has already stacked resilience value against deferred generation and T&D has the argument ready.
⚑ Two of This Week’s Carried Items Now Carry Two Different Numbers Each — Reconcile Before Citing
Topic: Filing Accuracy / Benchmark Hygiene / Program Comparables
Relevance: Electric school buses. This series reported in July, from the MassCEC demonstration, that a Beverly, Massachusetts bus discharged 10.78 MWh and earned $23,500 across two summers via National Grid’s ConnectedSolutions, with school buses earning up to $12,000/year. This week’s national feature reports the same bus at ~10.8 MWh/season and ~$12,000/season — and has Highland Electric Fleets’ CEO put the realistic Massachusetts norm at ~$6,000/year once discharge patterns and integration costs are counted. Scale context: 26 utilities across 19 states run V2G programs; 230+ V2G-enabled buses operate out of 14,000+ committed buses across ~1,550 districts; Zum’s 74-bus Oakland fleet supports 2.1 GWh/year, with a 238-bus San Francisco fleet planned on high-need circuits. Puget Sound Energy. This week’s entry states bidders must show a pathway to ≥50 MW by 2030; the August 28 digest reported PSE’s portfolio target as ~500 MW year-round by 2030 under Washington’s DALCU 10%-of-peak requirement, from 129 MW today.
Action Signal: Implement — Fix both in the benchmark library now. For V2G, cite the range with both attributions and use ~$6,000/bus-year for portfolio math — a business case built on $12,000 as an average will not survive an intervenor who finds the operator’s own number, and the EPRI/Dominion caveats (low 12-volt state-of-charge blocking bidirectional engagement; feeder hosting capacity studied fleet-by-fleet) are the reason the expected case sits below the best case. For PSE, record 50 MW as the per-bidder qualification floor and 500 MW as the portfolio obligation; presenting the 50 MW figure as PSE’s demand-side target understates a statutory mandate by an order of magnitude. Also copy PSE’s two structural provisions into any DR solicitation being drafted now: a single primary contract owner accountable for multi-aggregator bids, and incremental capacity that may not duplicate existing-program participation — the clause that prevents paying twice for the same customer.
📌 Sources
September 21, 2026 Entry
– Utility Dive — House Passes Ratepayer Protection Bill to Limit Data Center Cost Shifts (September 17, 2026)
– Utility Dive — Microsoft ‘Preserving Its Ability’ to Challenge Data Center Transmission Costs Decision (September 18, 2026)
– Utility Dive — Utilities, Schools See Growing Value in Electric Buses for Virtual Power Plants (September 16, 2026) (the Beverly/ConnectedSolutions benchmark was previously reported in the July 24 digest from the MassCEC demonstration; carried here for the national deployment count and the $12,000-vs-$6,000 reconciliation)
– Utility Dive — Nevada ‘Misses an Opportunity’ in Performance-Based DER Tariffs: Advocate (August 26, 2026)
September 22, 2026 Entry
– Discovery Alert — DOE Emergency Grid Order Defies Court Ruling, Targets Duke Energy (September 19, 2026) (Order No. 202-26-48; the Campbell vacatur and the PJM §202(c) cadence were reported in the September 18 digest)
– pv magazine USA — PG&E and Rewiring America Announce Google-Funded Virtual Power Plant Program (September 4, 2026) (previously reported in the September 4 digest; carried here for the 12-MW-by-2028 target and the ELRP/heat-pump compensation terms only)
– Puget Sound Energy — 2026 Demand Response Request for Proposals (RFP released April 17, 2026; September negotiation phase; December 15 service launch) (fourth appearance in the series; previously reported July 31, August 14 and August 28 — carried for the schedule advance and the two structural bid provisions)
– T&D World — Why DERMS Is Earning a Place Alongside SCADA (Nick Tumilowicz, Itron; June 9, 2026) (third appearance; previously reported August 28 and carried September 4 — one line only)
September 23, 2026 Entry
– Utility Dive — Texas PUC Adopts Softened Rules on Data Center Interconnection (September 21, 2026)
– Utility Dive — Pennsylvania PUC to Consider Ratemaking, ROE and Curtailment as Data Center Load Grows (September 11, 2026) (referenced as context in the September 18 digest; reported here as an item for the first time)
– Renewable Energy World / Factor This — Look What DERs Can Do: 580 MW Sent to the Grid in Largest-Ever VPP Dispatch (September 21, 2026)
– Troutman Pepper Locke — Washington Energy Report: FERC Approves ISO-NE Updated Market Rules for Distributed Energy Resources (Docket ER26-1956; order May 29, 2026) (fully reported in the September 11 digest; carried here only for the one-time 2.5-year commercial-operation extension)
September 24, 2026 Entry
– Utility Dive — North Carolina Regulators Deny Duke Energy Gas Turbine Request Over ‘Staggering’ Cost, Uncertain Need (September 22, 2026)
– Utility Dive — California Tightens Oversight of Data Center Energy Use, Grid Costs (September 22, 2026)
– Utility Dive — FERC Rejects ComEd’s Cancellation of PowerHouse Hillwood Data Center Contract (September 22, 2026) (the underlying August filings were reported in the August 28 digest; this is the ruling)
– PR Newswire — Research Finds Grid Resilience and Energy Affordability Are the New Defining Priorities for Utilities (IFS / Censuswide, 850 utility executives; September 24, 2026)
September 25, 2026 Entry
– Mgrid.org — NERC Sets Data Center Reliability Rules at 50 MW and 100 kV, a Threshold FERC Declined to Name (September 19, 2026) (CLO-001-1/-002-1/-003-1 previously reported September 4 and tracked September 11 and September 18; carried here for the applicability thresholds and registry schedule)
– Utility Dive — Efforts to Curb Data Center Speculation Gain Ground Across the US (September 24, 2026)
– Mgrid.org — South Dakota Weighs $1.93 Billion of Transmission for Data Centers It Does Not Have (September 22, 2026)
– Mgrid.org — Massachusetts Splits $40 Million Across 36 Storage Projects, With 28 of Them Adding 10.26 MW of Community Backup (September 22, 2026)
Verification Sources (targeted search, not from the daily research log)
– EPRI — Press Releases index / DCFlex Initiative (direct check: no release dated inside September 21–25; eleventh consecutive dry week)
– DCFlex — EPRI Data Center Flexibility Initiative (program site; FlexMosaic framework and demonstration coalition remain the publication channel)
– FERC — PJM Governance and Stakeholder Processes (proceeding page; the end-of-September deadline falls this week and no public resolution surfaced by search — still open)
– Mgrid.org — PJM Opens a 6 GW Backstop Auction September 30, After Its Regular Auction Cleared 525 MW of New Generation (September 9, 2026) (confirms the auction is proceeding on its September 30 open; no FERC approval order located, and no evidence found that large-load DR commitments reduced the procurement target)
– Utility Dive — PJM Files Backstop Auction Plan at FERC to Meet Capacity Shortfall (2026)
Compiled from the DER Research Log, Monday September 21 – Friday September 25, 2026. Twenty daily items consolidated; seven cross-week duplicates carried as movement or standing benchmarks rather than re-reported; three verification lines chased by targeted search. Eleven relevance flags (five Implement / three Engage / three Watch).