Solrova · Personal R&D

The Virtual Power Plant Learning Hub

A deep, self-paced field guide to where solar is going next, how VPPs actually work, who's winning the market, and the policy machinery underneath. Built to learn from, return to, and expand.

~33–38 GW US VPP capacity today 80–160 GW DOE 2030 target 16.8 GW Tesla, Sunrun & Renew Home deal (Jun 2026)

Start here: the mental models that make everything click

You already know the consumer story (what a battery earns, when to enroll). This hub goes one level deeper, into the machinery. Four ideas below carry 80% of the subject. Get these and the rest is detail.

The one-sentence definition

A Virtual Power Plant is a fleet of small, customer-owned energy devices, home batteries, rooftop solar, EV chargers, smart thermostats, water heaters, stitched together by software so they behave like one mid-size power plant. When the grid is stressed, the operator sends one signal and thousands of devices respond in seconds to minutes, either injecting power or cutting demand.

The scale, in numbers worth memorizing

Today

~33–38 GW
US/North America VPP capacity (DOE ~33 GW; Wood Mac 37.5 GW, 2025)

2030 Target

80–160 GW
DOE goal, ~tripling, = 10–20% of US peak demand

The prize

~$10 B/yr
Estimated annual grid-cost savings at that scale (DOE)
Today
~35 GW
2030 low
80 GW
2030 high
160 GW
Media-literacy flag, lead with GW, not dollars

You'll see "$2B → $40B VPP market" headlines everywhere. Ignore the precision. Those USD market-size reports disagree ~3× on the base year alone, they're modeled SEO products, not measured. The defensible numbers are capacity (GW) from DOE, Wood Mackenzie, and Brattle. When someone quotes a dollar TAM, that's your tell they're selling a report.

Model 1: Two axes, not one (the confusion-killer)

Almost every muddle in this space dissolves once you hold two separate distinctions apart. They are different questions, not synonyms:

Axis A: what it DOES

Supply-side = adds power to the grid (battery discharging).
Demand-side = reduces load (thermostat easing off AC). Descendant of demand response.

Axis B: WHERE it sits

Behind-the-meter (BTM) = customer's side of the utility meter (your home).
Front-of-meter (FTM) = the grid's side (a utility battery farm).

A Sunrun home battery exporting to the grid is BTM + supply-side. A thermostat program is BTM + demand-side. "Behind-the-meter" does not mean "demand-side", that's the classic conflation. Consumer-solar VPPs live almost entirely in the BTM box and span both A-axis modes.

Model 2: Firm capacity is a fraction of nameplate

A utility can bank on a gas peaker for ~100% of its nameplate. A fleet of 50,000 customer-owned batteries is messier, some are drained, some opted out, some didn't charge on a cloudy day, some are offline. "Firming" is the whole discipline of turning that noisy, probabilistic fleet into a single number a utility can count on. The result, the firm (accredited) capacity, is typically ~50–65% of nameplate for home batteries and ~70–95% for demand response. Maximizing that fraction is the entire game.

Why aggregation works at all

Any single battery is unreliable on a given day. But across tens of thousands of homes, those failures are largely independent, so the aggregate is far more predictable than any one device (the law of large numbers). The catch: a grid-wide heat wave is a correlated stress that hits exactly when you need the fleet, draining batteries and triggering opt-outs at once. Good VPP design (reserve floors, forecasting, staggered dispatch) exists to manage that correlated risk.

Model 3: Why data centers make this urgent

The hottest 2026 storyline is VPPs-as-an-answer-to-AI-data-center load. On June 24, 2026, Tesla, Sunrun, and Renew Home announced a 16.8 GW VPP framework explicitly aimed at serving data-center / hyperscaler demand (PJM and "Data Center Alley" in Virginia). Google became Voltus's first "bring-your-own-capacity" customer the same month. This reframes residential VPPs from a green nice-to-have into a grid-supply play with deep-pocketed buyers, but also creates competitive pressure, because utilities are simultaneously tempted by big central generation. Track this tension; it's the strategic backdrop for everything else here.

How this hub is organized

01 · How It Works

The machinery: the device-to-grid software stack, DERMS, dispatch & latency, the comms standards, the grid services, and firming.

02 · Business & Market

Capacity numbers, the players table, how aggregators actually make money, value stacking, and the 2024–26 M&A wave.

03 · Policy & Markets

FERC Order 2222 by region, wholesale vs. retail doors, the state-program map, the tax-credit cliff, and the real barriers.

04–06 · Practice

Two interactive calculators (homeowner + aggregator economics), a flashcard & quiz deck, and every source.
How to use it

Tap through the tabs in order, or jump around. Expand the accordions in 01–03 for the deep dives. Play with the calculators in 04 until the economics feel intuitive, then test yourself in 05. This is a living file, when you learn something new, add a card or a row.

How a VPP actually works

The machinery between a home battery and a wholesale market. Read top to bottom, or expand the sections that matter. The single most useful artifact here is the stack diagram, keep it in your head.

See a dispatch event happen

Press Run event. It's a calm afternoon, your battery charged from today's solar and is sitting full, held in reserve.
Grid demand vs. the peak limit
peak limit
Signal & power flow
home
home
home
home
Homes dispatched 0Delivered to grid 0 MWPer battery ~5 kWh

The stack: device → grid, in five layers

The aggregator almost never talks to your battery directly. It calls the manufacturer's cloud over an API (a doorway the OEM exposes). Commands flow down; telemetry flows back up the same chain.

LAYER 1 · DEVICE + FIRMWARE
The physical asset
Local control logic; actually charges/discharges.
Powerwall 3, Enphase IQ Battery, SPAN panel, EV charger, smart water heater
▼ telemetry ▲ commands
LAYER 2 · OEM CLOUD / DEVICE API
The manufacturer's servers
Exposes each device's available "flex envelope" to partners.
Tesla cloud, Enphase Enlighten + VPP API, SPAN Fleet Manager
▼ ▲ (cloud-to-cloud)
LAYER 3 · AGGREGATOR / VPP PLATFORM
Pools thousands of devices
Optimizes for market value; decides who dispatches, when, how much.
Tesla Autobidder, Uplight, EnergyHub Mercury, Virtual Peaker
▼ ▲
LAYER 4 · UTILITY / GRID-EDGE DERMS
The grid-reliability layer
Checks the physics: voltage, feeder limits, hosting capacity.
EnergyHub, GE Vernova GridOS, Itron, Schneider
▼ ▲
LAYER 5 · ISO / RTO MARKET OPERATOR
The wholesale market
Where FERC Order 2222 participation happens.
CAISO, PJM, NYISO, ISO-NE, MISO, SPP
DER typeTypical specNote
Home battery, Powerwall 313.5 kWh usable, 11.5 kW continuousPW3's 11.5 kW is more than double the old PW2's 5 kW, don't carry the old number.
Home battery, Enphase IQ 5P5.0 kWh, 3.84 kW continuousModular; homes stack 2–4.
Battery dispatch / event~3–8 kWhSunrun CalReady delivers ~3.3 kWh/battery over a 2-hr event. Low end reflects reserve floors.
Smart thermostat (AC shed)~1–1.5 kW/homeDecays each hour as AC cycles back on & customers override, not a flat number.
EV charger (Level 2)7.2–11.5 kWThe car's onboard charger caps the rate regardless of charger size.
Electric water heater4.5 kW elementA thermal battery, "load up" early, shed later. Controlled via CTA-2045 port.
C&I loadsMW blocksHVAC, refrigeration, pumps. CPower aggregates ~6 GW of these.
The core distinction

Dispatchable ("turn up") resources actively push or shift energy, like turning on extra faucets. Curtailable ("turn down") resources only reduce their own use, like turning down faucets already running. Both relieve grid pressure, but only dispatchable resources supply. The home battery is special: it does both.

V1G vs V2G today: V1G (smart one-way charging) is real and scaled now, a curtailable load. V2G (the EV discharging back) is technically proven but still mostly US pilots in 2026 (Ford F-150 Lightning is the most mature), bottlenecked by vehicle/charger availability and interconnection rules.

A DERMS (Distributed Energy Resource Management System) monitors, forecasts, coordinates and controls DERs. The teaching trap: two different layers both get called DERMS.

Grid / utility DERMS

Run by the utility. Optimizes physics, voltage regulation, feeder/transformer loading, hosting capacity. Lives near the control room. "Will the local wires physically handle this?"

Grid-edge / aggregation DERMS

Run by an aggregator/OEM. Optimizes economics, arbitrage, capacity commitments, ancillary revenue, enrollment. "How do we make money and serve the program?"

Clean one-liner: VPP software optimizes for market value; DERMS optimizes for grid reliability, same devices, different objective function. On platforms like Uplight the two increasingly blur into one product.

The signal path: the operator decides it needs power/load-cut → the aggregator translates "drop 5 MW" into thousands of device commands → the OEM cloud pushes each command to the unit over home Wi-Fi/cellular → the device acts → telemetry flows back up.

Why firming exists

Every link is a cloud or a home internet connection, so dispatches fail partially, industry rule-of-thumb is that 30–50% of enrolled devices may not respond in a given event (dropped connections, firmware, opt-outs). This is precisely why you can't count a fleet at 100%.

Latency spans six orders of magnitude depending on the service. Most "advanced" VPPs today operate at ~15-minute granularity (Maturity Level 2). Sub-second control is the frontier, not the norm.

ServiceCadenceWhat it means
Capacity / emergency eventhours of noticeLike a text an hour ahead: "use less at 5pm." What residential VPPs do today.
Spinning / non-spin reserverespond ≤10 minEmergency backup when a generator trips.
Frequency regulationevery 2–4 secA metronome the fleet dances to (PJM 2s, CAISO 4s). Batteries excel.
Primary frequency responsesub-secondToo fast for the internet, the inverter reacts autonomously, locally, not via the cloud.

Baselining & the "Aggregated Accuracy" insight: a VPP gets paid for load it didn't use, so it must estimate "what you'd have used anyway" (the baseline / CBL, e.g. CAISO's 10-in-10 average). Baselines are gameable, which is why directly-metered batteries are easier to settle than behavioral thermostat DR. And utility VPPs deliberately skip revenue-grade meters on every device, because cheap, noisy device meters average out across a big fleet ("aggregated accuracy"). That's what makes residential VPPs economically viable.

These protocols live at different altitudes of the stack and mostly complement rather than compete:

ProtocolLayer / directionJobMandated?
OpenADR 2.0bCloud-to-cloud, operator→aggregatorDR events, prices, signals (XML/SOAP)Widely deployed
OpenADR 3.0/3.1Cloud-to-cloud + to loadsSame, modernized (JSON/REST/OAuth2)Early adoption since 2025; coexists w/ 2.0b
IEEE 2030.5 / CSIPOperator→inverterSmart-inverter grid support (Volt-VAR, ride-through)Yes, CA Rule 21, HI Rule 14H
CTA-2045 / EcoPortLocal appliance port"USB for appliances", Shed / Load UpYes, WA, OR, CO, NY
SunSpec ModbusDevice-to-local-controllerStandard data models for inverters/batteriesNamed in IEEE 1547-2018

One-sentence synthesis: OpenADR carries the grid's wishes between clouds; IEEE 2030.5/CSIP delivers and governs those wishes at the smart inverter (and it's the law in CA & HI); CTA-2045 is the little plug that lets a water heater listen; SunSpec Modbus is the shared vocabulary for reading what a battery is actually doing. They stack far more than they fight.

Think of it as a maturity ladder: a basic VPP only shaves the system peak; a sophisticated one sells the "full stack." Response time is the organizing axis.

sub-secondsecondsminuteshoursseasonal
Primary freq.
Freq. regulation
Voltage / VAR
Reserves
Energy arbitrage
Capacity / peak
Distribution defer
  • Peak shaving / capacity: get paid to be available during the ~63 worst peak hours/year so the utility avoids a peaker. The biggest, most predictable stream.
  • Energy arbitrage: charge on cheap midday solar, discharge into the evening price spike. Daily "normal operations."
  • Frequency regulation: the fastest market service, follow an AGC signal every 2–4 sec. Nearly energy-neutral, so even a small battery qualifies.
  • Reserves: be-ready-to-respond-once within ~10 min for a contingency. Called rarely.
  • Voltage / VAR support: continuous, autonomous reactive-power injection by smart inverters (IEEE 1547-2018). Local, value is right where the DER sits.
  • Distribution deferral (NWA): shave a specific feeder's peak to delay a wires upgrade. ConEd's Brooklyn-Queens program deferred a $1.2B substation upgrade for ~$200M of DERs.
Battery vs load-only VPPs

Battery VPPs can play in all tiers (including fast regulation and voltage). Load-only VPPs (thermostats, water heaters) play the slower tiers, reserves, capacity, energy.

The output of firming is firm (accredited) capacity = a fraction of nameplate. The formal metric is ELCC (Effective Load Carrying Capability), the share of nameplate that actually reduces loss-of-load risk during the hours that matter.

ResourceFirm % of nameplateSource
Residential battery (4-hr)~58%PJM 2025 ELCC
Residential battery (in VPP)~60%Brattle assumption
Battery, 8–10 hr (longer = firmer)~70–78%PJM 2025 ELCC
Demand response~70–95%PJM (varies by planning year)

The five firming levers: forecasting (weather + solar + load), derating (ELCC), the diversity benefit (independent failures average out), reserve floors (e.g. CalReady holds back ≥20% for the homeowner, reduces firm capacity but essential for buy-in), and managing opt-outs (call events too often and people quit, permanently shrinking the fleet, which is why opt-out/default-in enrollment beats opt-in).

Treat ELCC as a snapshot

Every specific ELCC % is re-derived each planning year and varies by ISO, PJM's are shifting fast as loss-of-load risk moves into winter. Use ranges, not constants.

The business & the market

Who's winning, how the money actually flows, and the consolidation wave reshaping the field. The defensible framing leads with capacity and with verified deal facts, not with dollar TAMs.

Market size: the numbers you can defend

Today

~33–38 GW
N. America VPP capacity (DOE / Wood Mac 2025), growing ~14%/yr

2030 target

80–160 GW
DOE goal; current pace is well short of it

Cost edge

40–60% cheaper
VPP peak capacity vs. utility battery / gas peaker (Brattle)
Brattle's headline (use this framing)

60 GW of VPPs could meet US resource-adequacy needs for $15–35 billion less than conventional alternatives over a decade, plus ~$20B in societal benefits. Avoid the widely-circulated "$43 / $69 / $99 per kW-year" triplet, it's not supported by the source as a per-kW figure. Stick to "40–60% cheaper; $15–35B saved over 10 years."

CompanyRole in VPPScale / status (2025–26)
TeslaPowerwall VPPs via the app; Tesla Electric retail provider (TX)1M+ Powerwalls, ~6.7 GW global fleet; ~539 MW in one 2025 CA dispatch (a combined Sunrun+Tesla event)
SunrunLargest US residential VPP operator; leases solar+storage, aggregates (CalReady)~4.3 GWh networked, 18 programs; CalReady ~75k batteries, ~250 MW/event
Sonnen (Shell)Most sophisticated utility-dispatched home-battery VPPUtah Wattsmart: 4,000+ batteries. (Shell's sale unresolved, still Shell-owned)
Enphase / SolarEdgeDevice-enablement, plug into 3rd-party aggregatorsEnphase ~210 MWh enrolled. SolarEdge closed its storage division in the 2024 crisis.
Generac (ecobee)Load-flexibility VPP, thermostats + Concerto softwareecobee shed 108 MW summer 2025 across 143k+ devices
Base PowerTX home-battery + retail electricity startupMarquee raise: $1B Series C at $4B valuation (Oct 2025)
CompanyWhat they doStatus
Renew HomeN. America's largest residential demand-flex VPP (thermostats)OhmConnect + Google Nest Renew; backed by Sidewalk Infrastructure Partners ($100M), not Google
CPowerLargest independent C&I DR/VPP aggregator (~6 GW)Now an NRG asset, deal closed Jan 30, 2026
Uplight (+AutoGrid)Utility SaaS, DERMS + VPP; absorbed AutoGrid from SchneiderOctopus Energy agreed to take majority stake (Mar 2026, close pending); 8.5 GW across 85+ utilities
EnergyHubGrid-edge DERMS for 70+ utilitiesOwned by Alarm.com; 1.7+ GW; acquired Resideo Grid Services (Dec 2025)
VoltusC&I DR/VPP aggregator, all 9 NA markets (~7–8 GW)Private (2021 SPAC terminated); $18M FERC/MISO settlement (Jan 2025)
Swell EnergyResidential solar+storage VPPsShut down Aug 2024, over-leveraged financing its own installs
Green Mountain PowerUtility-run residential battery VPP (Vermont)~75 MW, Vermont's largest dispatchable resource; the $55/mo Powerwall lease
Common errors worth un-learning

The NRG/Renew Home Texas deal is a ~1 GW VPP, not a "$10B" deal (the $10B is DOE's national savings estimate). The big VPP M&A event is NRG buying CPower. Virtual Peaker is independent (NOT NRG-owned, that's CPower). EnergyHub is owned by Alarm.com, not Generac.

Four buyers pay for VPP value: (1) ISOs/RTOs in wholesale markets, capacity, energy, ancillary; (2) utilities, via programs/incentives; (3) retail providers, via the rate spread; (4) distribution utilities, for deferring local upgrades. The ultimate funding source is ratepayers, and the pitch is that VPPs lower total system cost (~$10B/yr at scale).

Value stacking is the crux: layering capacity + energy + ancillary + utility incentive onto the same asset without double-counting. A 400 MW VPP can provide $21–42M/yr of stackable services, but "too often grid operators only compensate VPPs for resource adequacy," leaving most of the stack on the table (RMI/Brattle).

Why a leased battery pays the homeowner less

If you own the battery (BYOB), you collect the incentive directly, ConnectedSolutions ≈ $1,200–$1,500/yr. If a third party owns it (Sunrun lease), the system owner holds the grid-services revenue, so the homeowner gets a smaller credit (CalReady ≈ up to $150). That's why Sunrun's ~$78/home average is far below a BYOB homeowner's. Same hardware, different ownership = different payee.

Capacity price anchor: PJM's 2027/28 capacity cleared at the cap, ~$333/MW-day ≈ ~$122k/MW-year, the most a dependable MW is worth in the priciest US market right now. But residential aggregations can't fully access it until PJM's Order 2222 rollout (2028/29).

The cleanest public numbers come from Voltus's 2021 investor deck, still the reference for what a healthy aggregator looks like:

Margins

~40% gross margin. Captures ~40% of value created ≈ ~$20k gross margin / MW / year. OpEx ~$7,500/MW/yr.

Acquisition

CAC ~$20k/MW; LTV ~$200k/MW → ~10:1 (software-grade). 73-day sales cycle; 100%+ net revenue retention.

Revenue splits are the least-settled number in VPP economics: an integrated residential player who owns customer + hardware may keep ~60–80% of gross payments; a pure C&I curtailment aggregator may keep only ~15–30% (a "risk-transfer premium"). They describe different businesses, diligence per program.

The structural moat

VPPs often see only 2–5% enrollment of an eligible population, so enrollment is the binding constraint. Players with a direct customer + hardware relationship (Sunrun, Tesla, Base Power) have the advantage. "Running grid services at scale is ultimately a software, operations, and policy problem." → Go play with calculator ② to feel this.

DateEventWhy it matters
May 2024Renew Home launches (OhmConnect + Nest Renew)~3 GW at launch; consolidates residential demand-flex
Aug 2024Swell Energy shuts downCautionary tale: over-leveraged owning its own installs
Oct 2025Base Power $1B Series C ($4B val)The marquee VPP-adjacent raise of the period
Jan 30, 2026NRG closes CPower (via ~$12B LS Power deal)The headline VPP M&A; NRG now bundles ~6 GW C&I VPP + Vivint hardware
Mar 2026Octopus Energy → majority of UplightIndependents being absorbed by energy majors
Jun 24, 2026Tesla + Sunrun + Renew Home: 16.8 GWThe current headline, VPPs aimed at data-center load (PJM, Virginia)

Read of the wave: the top is consolidating fast (NRG, Octopus, Renew Home), direct customer+hardware ownership is the durable moat, and the hot 2026 demand signal is hyperscaler data centers, deep-pocketed buyers that reframe residential VPPs as a serious supply play.

Who holds the 33–38 GW, and where

The honest answer to "where's the capacity?" The top-line is solid; the breakdown is fuzzy. Most of it is demand response, the geography concentrates in a few grid regions, and the company figures overlap so badly they can't be added together.

Read this first

The defensible number is the ~33–37.5 GW North America total (DOE Jan 2025; Wood Mackenzie Sept 2025) plus the type and area splits below. The per-company figures overlap and do NOT sum to 33–38 GW, an aggregator's "managed" capacity is often the same battery a utility and an ISO also count (FERC Order 2222 literally lets one asset appear in both books). Treat company numbers as a sense of scale, never as a total.

By type: it's mostly demand response

The ~33–38 GW is "mainly demand response" (DOE), a mix of commercial & industrial load curtailment and residential smart thermostats and water heaters. Batteries and solar grab the headlines but are still a small slice of the installed total.

The bulk

Demand response
C&I load curtailment + residential thermostats / water heaters. Thermostats are the single most common technology.

Fastest-growing

Batteries & EVs
Now appear in ~61% as many new installs as thermostats (a deployment count, not a capacity share).

Residential share

~10% of wholesale
10.2% of VPP wholesale-market capacity in 2025, up from 8.8% in 2024, small but rising fast.
No exact pie exists

Neither DOE nor Wood Mackenzie publishes a clean GW split of the total into C&I DR / batteries / thermostats / EVs. Anyone showing you a precise four-slice pie is inventing the precision. The honest read is directional: demand-response-dominant, thermostats #1, batteries & EVs rising fastest.

By area: where the capacity sits

On a registered-demand-response basis, the capacity concentrates in a few grid regions. One myth to drop: "PJM is half of US demand response" is outdated, MISO is now #1.

Grid region (ISO/RTO)Demand response, 2024% of regional peak
MISO (Midwest/South)~12,950 MW10.6%
PJM (Mid-Atlantic)~8,530 MW5.7%
CAISO (California)~4,370 MW9.0%
ERCOT (Texas)~4,100 MW4.8%
NYISO (New York)~1,920 MW6.1%
SPP (Central)~970 MW1.8%
ISO-NE (New England)~430 MW*1.7%
Total (7 ISOs)~33,270 MW~6.5%

Source: FERC 2025 Assessment of Demand Response (2024 data). *ISO-NE counts only "active" resources at peak; its Forward Capacity Market demand resources are ~3,300–3,900 MW. Note this ~33 GW of ISO demand response overlaps the DOE "~33 GW VPP" figure, largely the same assets. Don't add them.

Residential VPP standouts

Puerto Rico

~500 MW
LUMA "battery sharing", 81,000+ home batteries. The cleanest standalone residential VPP (not inside a mainland ISO).

Vermont

~70 MW
Green Mountain Power, the state's largest dispatchable resource, ~4,500 customers.

Texas, ERCOT ADER

~241 MW
Aggregated-DER pilot, qualified across 9 aggregators (May 2026); a 500 MW cap leaves room to grow.

California has huge storage, but most is utility-scale, not residential VPP (its DSGS fleet did dispatch ~539 MW in one 2025 event). Most state residential programs roll up into their ISO's total above, don't double-count.

By company: big numbers that don't add up

Every figure below uses a different definition: owned vs managed vs potential, power (MW) vs energy (MWh), US vs global. They overlap each other and the ISO totals. Useful for scale, useless to sum.

CompanyWhat it isCapacityScope / caveat
UplightDERMS software>8.5 GW managedN. America. Software-managed inside utilities' own programs, not owned.
VoltusC&I demand response>7 GWN. America. A subset of ISO DR totals. Private (its SPAC was cancelled).
CPower (NRG)C&I demand response~6 GWUS. Enrolled C&I DR, already counted inside ISO markets.
TeslaPowerwall fleet6.7 GW globalGlobal, not US, 1M+ Powerwalls, only ~1 in 4 VPP-enrolled. No official US figure.
Renew HomeSmart thermostats~3 GWUS. Demand-side load, not batteries. "50 GW" is a 2030 goal.
EnergyHubDERMS software~2–3 GW managedUS. Managed across dozens of utility programs; figure is moving.
SunrunResidential solar+storage~425 MW / ~4 GWhUS. Largest residential VPP. 425 MW dispatchable ≠ 4 GWh installed energy.
LeapDER market-access API~370 MW nominatedUS. On partners' hardware, overlaps other aggregators. A classic double-count.
Generac / ecobeeThermostat DR108 MW deliveredUS+Canada. 108 MW actually delivered vs ~2.8 GW modeled potential.
Green Mountain PowerUtility resi VPP~75 MWVermont. Inside the ISO-NE total.
Sonnen (Shell)Residential batteries~250 MWhThat headline is Germany, in MWh not MW. US is a regional patchwork.
The four traps

Global vs US (Tesla's 6.7 GW is worldwide). Energy vs power (Sonnen's MWh, Sunrun's GWh ≠ dispatchable MW). Managed vs owned (Uplight, Leap, EnergyHub "manage" capacity that lives on others' hardware). Delivered vs potential (Generac's 108 MW real vs 2.8 GW modeled). Mix these up and you double- or triple-count the same battery.

Policy & energy markets

The rulebook that gates the upside. Five things to internalize: FERC Order 2222 is real but mostly not live yet; almost every residential VPP today is a retail program; the 25D tax credit died end of 2025; federal momentum reversed in 2025; and the deepest barriers are boring and structural.

What it requires (issued Sept 2020): every ISO/RTO must let aggregations of DERs, as small as 100 kW, compete directly in wholesale energy, capacity, and ancillary markets alongside big power plants, with rules to prevent double-counting. Revolutionary in principle; slow in practice.

ISO / RTOStatusFull implementation
CAISO (California)Live, furthest along, built on its prior DER modelIn effect
ISO-NE (New England)FERC-approved; phasing in~2026
NYISO (New York)Compliant; final pieces in progress~end 2026
PJM (Mid-Atlantic)Approved; long runway~2028–2029
MISO (Midwest)Two-phase compliance2027 → 2029
SPP (Central)The laggard, still finalizing~2030
ERCOT (Texas)Not FERC-jurisdictional, runs its own ADER pilotsee below
Founder takeaway

Outside CAISO (and outside ERCOT's own programs), direct wholesale residential VPP revenue is mostly a 2026–2029 future event. Today's bankable money is utility-program money. Model around program revenue now; treat wholesale stacking as upside as Order 2222 lands.

ERCOT ADER (Texas): because Texas is outside FERC, it runs its own Aggregated DER pilot, often a faster path. The cap was raised to 500 MW (Mar 2026), with ~241 MW qualified across 9 aggregators (May 2026) and participants including Tesla, Sunrun (via NRG/Reliant), Sonnen, and Octopus/Enphase. This is a big reason Texas is so active.

ProductWhat you're sellingAnalogy
EnergyThe actual kWh dispatched, paid at the local marginal price (LMP)Getting paid for groceries you actually sell
CapacityA promise to be available at peak, $/MW-day, whether or not you're calledA retainer for being on call
Ancillary servicesFast reliability services (regulation, reserves), mostly an availability payment + performance bonusA fueled fire truck idling, plus extra when it rolls

The same battery fleet can often stack all three, subject to anti-double-counting rules (you can't sell the identical MW twice at the same instant). Unlocking that stack is exactly Order 2222's job.

Door A, Wholesale (FERC/ISO)

Aggregator bids into the ISO's markets. What Order 2222 unlocks. Used today mostly by larger C&I resources & big aggregators.

Door B, Retail (state/utility)

Customer enrolls in a utility/state program (demand response, BYOD/BYOB); the utility handles any wholesale interaction. ConnectedSolutions, GMP, Xcel.

Why nearly every residential VPP you can join in 2026 is Door B: Order 2222 isn't fully live in most regions; wholesale rules are brutal for small resources (a home is ~5 kW vs. the 100 kW floor → you must aggregate 20+ homes just to qualify); the "10 MW telemetry cliff" makes device-level wholesale metering uneconomic; and a fixed $/kW-summer utility payment is simply easier to underwrite than volatile wholesale revenue. Residential is only ~10% of total US VPP capacity today.

ProgramStateWhat it paysStatus (2026)
ConnectedSolutionsMA$275/kW-summer (~$1,200–$1,500/yr)Active
ConnectedSolutionsRI$225/kW (legacy $400/kW)Active
Energy Storage SolutionsCT$300–$550/kW-yr over 10 yrsNew model Apr 2026
BYOD / Powerwall leaseVT (GMP)Up to $10,500; or $55/mo leaseActive
DSGSCA$/kW capacity + 2026 bonusActive but budget-curtailed
ELRPCA$2/kWh (non-residential)Residential sunset end-2025
SGIPCAIncome-qualified storage rebateGeneral tiers closed
Renewable Battery ConnectCO (Xcel)$350/kW up to $5,000Reopened May 2026
ADER pilotTX (ERCOT)Wholesale revenue (~$400/yr/PW via Tesla)Active, 500 MW cap
VPP pilotVA (Dominion)TBD, up to 450 MWFiled Dec 2025, pending
BEDRP ("PowerOn")PR$1.25/kWh to aggregatorsActive, ~70k batteries
Reconciling with your KC article

Your consumer article's "$50–$300/yr" is the conservative national / OEM-locked range. The Northeast BYOD programs (ConnectedSolutions, GMP) are the high end at $1,000–$1,500/yr. Both are true, the spread is the whole point, and the homeowner calculator lets you see it.

DOE "Pathways to Commercial Liftoff: VPPs" (2023, updated Jan 2025) is the anchor document: 80–160 GW by 2030, 10–20% of peak, ~$10B/yr savings. RMI's VP3 (Virtual Power Plant Partnership, since 2023) is the industry coalition, members now include Ford, GM, Generac, NRG, Renew Home, Uplight, ecobee, Leap.

The tax change reshaping everything

Section 25D (homeowner-owned, cash/loan) expired Dec 31, 2025. Section 48E (third-party-owned lease/PPA) survives under current law: a project that begins construction by ~July 4, 2026 keeps a runway out to 2030, otherwise it must be in service by end-2027. With 25D gone, a buyer gets no federal credit in 2026, but a leasing company still claims 48E and passes savings through. This pushes the market toward third-party ownership (TPO), projected ~60–65% of installs exiting 2026. That's good for VPPs, owner-aggregators (Sunrun) can enroll their fleets at scale, and VPP payments help "replace" the lost tax credit.

Don't over-claim TPO as a rescue

Freedom Forever (the #2 residential installer) filed Chapter 11 on April 15, 2026 despite a dealer+TPO hybrid; Wood Mackenzie projects the residential market shrinks ~19–21% in 2026.

Federal momentum reversed in 2025: DOE cancelled billions in grid awards and loans (the Sunnova "Project Hestia" VPP loan guarantee was terminated; Sunnova filed bankruptcy June 2025); a 3–2 Republican FERC (chair Laura Swett) keeps Order 2222 in effect but rollback is a live risk; and a Section 301 tariff on Chinese grid batteries rose to 25% (Jan 1, 2026), raising BESS costs. The market keeps growing even as policy support recedes.

  • Metering & telemetry cost, the #1 residential obstacle. Above CAISO's 10 MW threshold, real-time per-device metering can be required, impractical for thermostats.
  • Double-counting & baselines, you can't be paid for the same MW in two programs; and paying for "avoided" load rests on a gameable counterfactual.
  • Interconnection queues, DERs still must interconnect; backlogs and hosting-capacity limits cap participation.
  • Utility resistance & cost-shift arguments, incumbents argue paying DERs shifts costs to non-participants; FERC had to force PJM/MISO/SPP to shorten review timelines.
  • Market-rule complexity, wholesale rulebooks were written for big generators; adapting them to a fleet of homes is genuinely hard.
  • Wholesale-vs-retail jurisdiction, the deepest fault line: FERC governs wholesale, states govern the retail customer, and a BTM aggregation is simultaneously both. Who controls dispatch, data, and overrides remains unresolved.

Calculators: make the economics intuitive

Two sandboxes. The first is the homeowner's view (what a battery earns in a program). The second is the aggregator's view (the business you'd actually build). Both use real 2025–26 benchmark defaults, labeled illustrative, not quotes.

① Homeowner VPP earnings

Pick a real program and a battery setup. Watch how wildly the answer swings, the same battery earns ~$100/yr in an OEM-locked program and ~$1,500/yr in a Northeast BYOD program.

Estimated annual VPP income
Per battery
VPP-only payback on hardware

The honest takeaway (consistent with your KC article)

Batteries earn their keep on backup and self-consumption under TOU / NEM 3.0. VPP income is a bonus on top, never the reason to buy. A $12k battery earning $200/yr is a 60-year payback on that stream alone.

② Aggregator unit economics

The business view. Defaults are anchored to the cleanest public numbers in the space, Voltus's 2021 investor deck (~40% gross margin, ~$20k gross margin per MW-year, target ~10:1 LTV/CAC). Move the sliders to feel what makes the model work or break.

Total annual gross margin
Gross margin / MW-yr
Lifetime value / MW
LTV : CAC ratio

Self-quiz: lock it in

Flip the flashcards to drill the vocabulary, then run the multiple-choice check. Recall beats re-reading.

Flashcards

tap to reveal
1 / 1

Multiple choice

Score: 0 / 0 answered

Sources & glossary

The glossary doubles as your flashcard deck. Below it, the primary sources, grouped by tier so you know what to trust. Capacity (GW) and regulation facts are high-confidence; USD market-size reports are not.

Glossary

Primary sources

Confidence tiers

High, cite freely: DOE Liftoff, Wood Mackenzie, Brattle, RMI (capacity & economics); FERC.gov & the ISO/RTO trackers (regulation); company press & SEC filings (player scale). Low, directional only: USD "market size" numbers from research-report mills (they disagree ~3× on the base year). Lead with GW, never USD TAM.

Fast-moving figures

A few numbers here change often and are worth re-checking before you lean on them: the exact CA DSGS 2026 budget, GMP's current fleet MW, PJM's Order 2222 date (a recently-delayed, still-moving target), and ERCOT ADER's enrolled MW (climbing monthly). The capacity totals and program payment rates are the most stable.