Tesla and Sunrun discharged a fleet of household batteries into California's grid for three evening hours during a heat wave. Here is what that number means against the extra heat-wave demand, against China, and in dollars.
On the evening of 9 Sep 2026, Sunrun and Tesla dispatched batteries enrolled in two state programs: the Energy Commission's Demand Side Grid Support (DSGS) and the utilities' Emergency Load Reduction Program (ELRP). DSGS calls an event when the day-ahead wholesale price tops $200/MWh.
Average output per Powerwall ≈ 4.7 kW (517 MW ÷ 110,000). Energy delivered is not published; with a 20% backup reserve kept in each battery, it was most likely 1.1–1.5 GWh over the three hours estimate.
CAISO reported a 9 Sep peak of 50,042 MW, the highest of 2026 (record: 52,061 MW in 2022). "Incremental demand" depends on the baseline, so here are four reasonable ones. Because CAISO's metered load already has the batteries' output netted out, the grey bar is the increment the grid would have faced without the VPP (reported peak + 580 MW − baseline). The orange bar is the 580 MW dispatch on the same scale.
Caveats: this assumes the dispatch overlapped the peak hour (timing not published). Measured against the reported 50,042 MW instead, the shares are 9–22%. CAISO expected to keep reserves that evening, so this was price relief and insurance, not a blackout averted.
California already runs the world's most battery-heavy evening peak. The residential fleet is a small slice of it.
Utility-scale batteries in CAISO hit a discharge record of 12,990 MW on 9 Jul 2026, covering 36% of demand at 8:20 pm. Installed CAISO battery capacity: 17,225 MW (2 Sep 2026). For scale, a typical gas peaker plant is 50–250 MW, so 580 MW equals a mid-size gas plant.
China's single events are larger, but they are a different animal: city-level aggregation of commercial and industrial (C&I) load, or grid-company storage. Normalizing by peak load is the fairest comparison.
Share of the relevant grid's peak load. Shenzhen figure is available adjustable capacity, not one verified event.
| Place | Event / resource | Size | Resource type | Paid per kWh |
|---|---|---|---|---|
| California | Tesla + Sunrun VPP, 9 Sep 2026 | 0.58 GW | Home batteries | ELRP $1–2; DSGS paid per kW |
| California | Grid batteries discharge record, 9 Jul 2026 | 12.99 GW | Utility-scale storage | market |
| Shanghai | City VPP response, 24 Jul 2026 | 1.65 GW | Mostly C&I load, some storage | up to ¥6 (~$0.84, Aug 2025 event) |
| Shenzhen | VPP Management Center, adjustable | ~1.3–1.4 GW | C&I, EV charging, storage | — |
| Jiangsu | Grid storage discharge, 16 Jul 2026 | 7.45 GW | Grid-scale storage | ¥3–4 DR (~$0.42–0.56) |
| Shandong | 144 storage stations discharging together, 11 Jul 2025 | 8.04 GW | Grid-scale storage | ¥0.28 spot (Jinan VPP) |
| Sichuan | New-type load management system, Aug 2024 | 7 GW | Load curtailment, 3M+ user responses | — |
| Australia | ~600,000 battery homes nationally; at Victoria's record peak (Jan 2026) they drew 1.4 kW less than solar-only homes | n/a | Home batteries (passive, not dispatched) | varies |
| Puerto Rico | LUMA/Sunrun emergency battery DR, largest 2025 event | 0.048 GW | Home batteries | — |
| UK | Demand Flexibility Service, winter 2024/25 (44 events) | 3.9 GWh total | Households shifting use | Octopus 52p |
Yuan converted at ~¥7.1/$ estimate. Puerto Rico's fleet is ~500 MW nameplate but its largest single dispatch was 48 MW. No Chinese residential-battery event of comparable size was found.
The goal is the same (shave the evening peak), but the toolkit is top-down and C&I-led.
State Grid and China Southern Grid run the "new-type load management system" 新型负荷管理系统, which can reach factories, malls, air-conditioning and EV chargers directly. National load-management / demand-response capacity exceeds 90 GW (~6% of peak).
Shanghai, Shenzhen and Guangzhou run VPP centers that bid aggregators' C&I load into response events. Nationally ~530 VPPs with >18 GW of adjustable capacity (summer 2026); policy targets are ≥20 GW by 2027 and ≥50 GW by 2030 《关于加快推进虚拟电厂发展的指导意见》.
153 GW / 396 GWh of new-type storage installed by end-June 2026 (+61% y/y). Provinces dispatch hundreds of stations at once: 7–8 GW in Jiangsu and Shandong.
If voluntary response is not enough, "orderly power use" 有序用电 curtails industrial users by schedule. California's equivalent is rotating outages, which hit 400,000+ customers in Aug 2020.
"Worth" depends on what you would have had to do instead. Six lenses, from narrowest to broadest.
| Lens | Calculation | Value | Status |
|---|---|---|---|
| Energy that night replace ~1.3 GWh at peak prices | ~1.3 GWh × $200–500/MWh | $0.3–0.7M | estimate |
| Blackout insurance only if it had prevented load shedding | ~1.3 GWh × ~$10,000/MWh value of lost load | ~$13M | estimate not the case on 9 Sep |
| What California actually pays DSGS Option 3 implied price | 580 MW × ~$81/kW-season (Option 3 budget $19.5M ≈ 241 MW) | ~$47M / season | estimate most realistic |
| Full capacity rental buy 580 MW of system resource adequacy | 580 MW × $28.65–42.54/kW-mo × 12 (CPUC system RA benchmarks) | $200–300M / yr | estimate upper bound; a 3-hour, event-limited resource earns less |
| Build a gas peaker instead | 580 MW × ~$1,700/kW (E3, frame CT online ~2030, nominal) | ~$1.0B capex | estimate |
| Program-level savings | Brattle (commissioned by Tesla/Sunrun): DSGS-type programs | up to $206M net by 2028 | sourced sponsor study |
Bottom line: the one night itself was worth well under $1M of energy. The real value is standing capacity: about $45–50M a season at the price the state actually pays, $200–300M/yr if valued as full system capacity, versus ~$1B to build a gas plant that sits idle most of the year.
Money flows from the state (DSGS) or utilities (ELRP) to the aggregator, which passes part of it to homeowners.
the entire 2026 DSGS Option 3 budget ($19.5M for ~241 MW, $42.7M for ~530 MW with extra funding), ≈$81/kW-season, shared across all VPPs. Utilities pay ELRP separately per kWh CEC
Sunrun owns many of the batteries (55% of the Powerwalls) and pays customers a flat fee, so it keeps most capacity revenue. Tesla passes ~$64.55/kW-season to DSGS owners, most of the ~$81 paid inference
Sunrun: flat $50–150 per battery. Tesla DSGS: per kW, up to $280 per Powerwall. Tesla ELRP: $2/kWh, typically up to $16 per event
| Payment | Rule | This event / season | Status |
|---|---|---|---|
| Tesla owner (ELRP) | $2.00 per incremental kWh; season typically $100–450 per Powerwall | ~$15–20 this night (8–10 kWh) | estimate rate from Tesla |
| Tesla owner (DSGS) | Per kW shown in events: $14.97/kW in September; $64.55/kW across May–Oct | ~$70 for September at 4.7 kW; ≤$280 season | estimate rates from Tesla |
| Sunrun customer (CalReady DSGS) | Flat $50–150 per battery per season (May–Oct, max 60 event-hours) | $0 extra for this night; ~$4.5–13.6M season for ~90k batteries | estimate |
| Aggregators (DSGS Option 3, all VPPs) | 2026 budget $19.5M (~241 MW), up to $42.7M (~530 MW) | $20–43M season | CEC |
For comparison, Chinese response payments are ¥3–6/kWh (~$0.42–0.84), roughly one-fifth to two-fifths of California's $2/kWh, but are paid to large C&I users rather than households. DSGS got only a ~$27M transfer for 2026 and has no funding for 2027; a proposal to move it into the utility-run ELRP was dropped from the final budget.
Small vs. the whole grid (1.2% of peak, 4.5% of grid-battery output), but meaningful vs. the heat-wave increment (8–18%). It proves 140,000 homes can act like one mid-size power plant on command.
China's events are ~2–14× larger in MW and 3–5× larger relative to peak load, but they come from factories and grid storage under grid-company control. California leads on the household-battery model.
One night of energy: under $1M. Capacity it stands in for: ~$45–50M a season at the state's price, ~$1B to build a gas peaker. Homeowners get ~$50–280 per battery per season; Sunrun keeps most of its capacity revenue, Tesla passes most on.