Elon turned on the Brookshire, Texas plant on August 6. It will add 50 GWh a year to U.S. battery capacity. Half of that would be enough to drastically reduce, at a fraction of the cost, the roughly $16 billion a year PJM pays for standby electric capacity it doesn’t expect to use. It probably makes unnecessary the auction PJM wants to start in a few weeks, which is likely to add billions more. Storage could probably do the same on the other U.S. grids.
Texas had a heat wave this summer that produced record demand. Storage and solar added enough effective capacity for the grid to handle the load without difficulty. Renewables produced 100% of California’s electricity 18% of the time, and more storage could probably double that. Australia has been adding both batteries and data centers, and rates there are going down this year.
Dave
1. The authoritative number: Tesla’s own capacity table
The manufacturing table in Tesla’s Q2 2026 shareholder deck (Form 8-K exhibit 99.1) is the only capacity figure Tesla publishes in a filed document, and it is materially more conservative than the figures circulating in trade press:
| Region | Product | Installed annual capacity | Status |
|---|---|---|---|
| California (Lathrop) | Megapack | 40 GWh | Production |
| Shanghai (Lingang) | Megapack | 20 GWh | Production |
| Texas (Brookshire) | Megapack | 50 GWh design | Production Aug 6, 2026 |
| Nevada (Sparks) | Powerwall | >6 GWh | Production |
Tesla attaches its standing caveat that installed capacity is not the current production rate, and that limitations may emerge as output approaches capacity.
This is the single most important line in the whole file. Tesla’s own filing puts Shanghai at 20 GWh, not the 40 GWh repeated in essentially every press account — including Tesla’s own 2025 marketing, which said global Megapack capacity was “expanding to 80 GWh” with Shanghai. Section 3 explains why 20 GWh is the more plausible figure.
2. Brookshire, Texas — the new capacity
Announced at the September 9, 2025 “Las Megas” event during RE+ in Las Vegas, and confirmed operational on August 6, 2026 via Tesla’s Megapack account:
- Designed capacity: 50 GWh/year, dedicated to Megapack 3 — roughly 10,000 units at 5 MWh each
- Groundbreaking to operation in 16 months; construction began April 2025
- Location: Brookshire, Waller County, in the Empire West Industrial Park, ~35–40 miles west of Houston
- Investment reported at ~$200M (≈$44M facility, ≈$150M equipment)
- Two buildings totaling >1.65M sq ft: ~1M sq ft manufacturing, ~600k sq ft warehousing
- Staffing tied to the Waller County abatement (10-year, up to 60% property tax): 375 by end-2026, 750 by end-2027, 1,500 by 2028
Tesla’s stated timeline slipped earlier, not later: the Q2 2026 shareholder letter said production was “on track to begin this year,” and it began in early August.
3. Shanghai — the ramp nobody in the English-language press has priced correctly
Chinese sources are far more detailed here, and they are what makes the 20 GWh filing figure legible.
Official milestones (Tier 1, Shanghai municipal government and Tesla China statements):
- Signed into Lingang New Area April 9, 2023
- Groundbreaking May 23, 2024
- Completion acceptance December 27, 2024 — seven months
- Production ceremony February 11, 2025, first Megapack off the line
- Site ~200,000 m²; welding, paint and final assembly in-house; welding-shop automation reported above 90%
- Planned output stated consistently as 10,000 units/year, “近40GWh” (close to 40 GWh)
- Mike Snyder (VP, energy and charging) at the ceremony: the plant would begin its capacity ramp that quarter, and Megapack supply was constrained by capacity
- First export shipment March 21, 2025, to Australia; subsequent output to serve domestic and Asia-Pacific markets
The actual output numbers, and why one of them is wrong:
Two figures circulate for Shanghai’s 2025 production, and they are irreconcilable:
- 21世纪经济报道 (Sept 3, 2025): the plant’s 1,000th Megapack came off the line July 29, 2025 — roughly 5.5 months after start of production. At 3.9 MWh per Megapack 2 XL, that is about 3.9 GWh cumulative.
- TechWeb / Sina Finance (Feb 2, 2026): Shanghai produced more than 2,000 Megapacks in 2025 — roughly 8 GWh.
- 36氪 (Jan 2026): Shanghai produced about 26 GWh in 2025, which it put at 55.7% of Tesla’s global deployments.
The 36氪 figure cannot be right. 26 GWh requires roughly 6,600 units. The plant had built 1,000 by end-July; reaching 6,600 would require 5,600 units in five months — more than a tenfold increase in run rate, from ~180/month to ~1,100/month, on a line whose nameplate implies ~830/month at full tilt. The TechWeb unit count and Tesla’s own 20 GWh installed-capacity line are mutually consistent with a slow ramp; the 26 GWh figure is not consistent with anything. The 55.7% share it cites is simply 26 ÷ 46.7 (Tesla’s 2025 global deployments), which makes it look sourced when it is arithmetic on an unsupported numerator.
Working conclusion: Shanghai in its first year ran at roughly 20% of its announced 40 GWh design capacity, and Tesla’s filed 20 GWh figure likely reflects what is actually tooled and commissioned rather than the announced target. This is the central factual finding here and it is worth an independent check before publication.
4. The 2027–2030 question: there is no Tesla guidance
Neither English nor Chinese sources contain a Tesla statement of the form “X GWh/year in 2028.” Every GWh figure above is annual nameplate, undated beyond “once fully ramped.”
What can be summed:
| Site | Nameplate | Source quality |
|---|---|---|
| Lathrop | 40 GWh | Tesla filing |
| Shanghai | 20 GWh installed / 40 GWh design | Filing vs. announcement — unresolved |
| Brookshire | 50 GWh | Tesla statement, undated ramp |
| Sparks (legacy Megapack line) | ~3 GWh | Trade press |
Summed at design values this produces the 130–133 GWh/year figure carried by pv magazine and ess-news. Summed at filed values it produces roughly 110 GWh. Neither is a Tesla forecast for any particular year.
The only company-sourced documents that actually span 2027–2030:
- The Waller County hiring schedule. Full 1,500-person staffing in 2028 implies full rate no earlier than 2028. This is a contractual commitment tied to a tax abatement rather than a production forecast, but it is the firmest date on the record.
- The LG Energy Solution supply contract. $4.3B, Lansing, Michigan; LFP prismatic cells; term running August 2027 to July 2030, with options to extend up to seven years and to increase volumes. The Lansing plant is sized around 50 GWh/year — reported as enough to cover Megapack 3 output at Brookshire. The option structure rather than a fixed escalating volume is itself informative: Tesla did not commit to a number for those years either.
- Musk’s “100 GWh per year of stationary storage” remark — undated, predates the Texas plant.
Anything projecting Tesla at ~150 GWh in 2027 or ~1 TWh by 2030 is third-party CAGR extrapolation. The 240 TWh storage / 30 TW renewables figures from Master Plan 3 (2023) are a global build-out thesis, not a Tesla manufacturing target, and should not be allowed to migrate into a capacity table.
5. Capacity is not the binding constraint — deployment timing is
| Period | Deployed | Note |
|---|---|---|
| 2022 | 6.5 GWh | +64% |
| 2023 | 14.7 GWh | +125% |
| 2024 | 31.4 GWh | +113% |
| Q1–Q3 2025 | 32.5 GWh | Q3 alone 12.5 GWh |
| 2025 full year | 46.7 GWh | Q4 14.2 GWh |
| Q1 2026 | 8.8 GWh | −15% YoY |
| Q2 2026 | 13.5 GWh | +41% YoY, +53% QoQ, second-largest quarter ever |
Q2 2026 annualizes to roughly 54 GWh against ~110 GWh of filed nameplate — about 50% utilization. ess-news made the same observation in late 2025, putting the factories at roughly half designed output.
Tesla’s own framing on the Q2 2026 call (CFO Vaibhav Taneja): the energy business is inherently lumpy, deployments are tied to customer timelines and largely outside Tesla’s control, the order backlog is robust, and the company is building against both existing demand and expected data-center and electrification demand.
Asked directly by Dan Levy (Barclays) whether energy storage would be supply-constrained for the foreseeable future, Musk did not answer the supply half of the question. He pivoted to demand, arguing that batteries are the cheapest route to expanding usable US energy output — roughly 1.2–1.3 TW of installed generation against ~0.5 TW average usage, implying output could potentially double with storage alone — and that SpaceX bought Megapacks for data centers mainly to smooth power during training runs, where draw can fall 70% in 100 milliseconds.
Margin context: Q2 2026 energy gross margin fell to 20.4% from 39.5%, driven by a $240M warranty true-up tied to vendor cell issues on legacy deployments, the non-repeat of >$200M in Q1 tariff benefits, and falling industrial storage ASPs. Tesla guided long-run energy gross margin to the mid-to-low 20s. For a utility buyer, the warranty charge is the line item worth attention: it is Tesla absorbing a supplier’s cell defect across an installed fleet.
6. Product claims relevant to utility procurement
Megapack 3: ~5 MWh per unit, up from 3.9 MWh (+28%) in the same footprint, via a new 2.8-litre LFP cell co-developed with Tesla’s battery team. Thermal management borrows from the Model Y heat-pump design; connection points reduced 78%. Chinese coverage adds a detail the English reports omit: cable connections drop from 24 on the Megapack 2 XL to three busbar connections, and servicing no longer requires roof penetration.
Megablock: four Megapack 3 units plus transformer and switchgear as a pre-engineered medium-voltage block. 20 MWh AC, 25-year design life, >10,000 cycles, 91% round-trip efficiency at medium voltage inclusive of auxiliary loads, 248 MWh AC per acre, operating range −40°C to 60°C, 23% faster to install, up to 40% lower construction cost. Snyder’s headline target: 1 GWh commissioned in 20 business days.
That last figure is a target stated at a product launch, not a demonstrated result. It is the claim most likely to be quoted back at Tesla by a procurement officer, and it should be labelled as a target wherever it appears.
Autobidder was presented alongside as managing 600+ projects totaling 4 GWh, with a further 10 GWh coming online.
7. Cell supply — the real forward constraint
Chinese reporting (原 晚点 LatePost, via IT之家 and 21世纪经济报道) is considerably more specific than anything in the US press:
- Tesla’s model is to buy cells and assemble systems itself; Megapack contributes roughly 80% of energy segment revenue.
- Before 2024, CATL was the sole cell supplier to Tesla’s US storage factory. Musk said on a 2024 call that cell oversupply meant Tesla would add suppliers to reduce purchase prices.
- The supplier set then expanded to Panasonic, LG Energy Solution, CATL, BYD and Sunwoda (欣旺达), with EVE Energy (亿纬锂能) as the sixth, its Malaysia plant slated to supply Tesla US from 2026. Sources maintained CATL remained Tesla’s largest storage cell supplier after the expansion.
- Disputed: FinDreams (弗迪电池, BYD) was reported in mid-2024 as contracted to be a second cell source for Shanghai from Q1 2025. Tesla China publicly denied it as false (21世纪经济报道, February 2025). No subsequent confirmation either way has surfaced.
US-side vertical integration:
- Nevada LFP cell plant, using CATL-derived equipment, in early ramp as of spring 2026. Capacity given as 7 GWh at the Las Megas event; other reporting says a ~10 GWh target. Output earmarked for Megapack, not vehicles.
- LGES Lansing as above — the contract exists in the form it does because of tariffs and IRA/ITC domestic-content thresholds, not because Tesla wanted a second chemistry.
- Tesla said in Q2 2026 that battery pack capacity remains the main limiting factor on ramping production, alongside electronic components. Note this was stated in the context of vehicles; Tesla has not said pack capacity is the binding constraint on Megapack specifically.
The Malaysia routing and the Michigan plant are the same play: non-China-origin cells for US-bound Megapacks. That is the actual 2027–2030 constraint, and it is contractual and jurisdictional rather than physical.
8. Price position — why Shanghai output goes to Australia and Europe, not China
Chinese coverage is blunt about the gap. A 2-hour Megapack was listed at roughly $1.03M (list) and $940K (volume) — about $343/kWh and $313/kWh, or 2.23 and 2.04 RMB/Wh — at a time when Chinese domestic system tenders had fallen below 0.5 RMB/Wh, with 4-hour systems reaching about 0.426 RMB/Wh.
Two analyst caveats worth carrying:
- 鑫椤锂电’s 龙志强 argued the comparison is not apples-to-apples once export shipping, taxes and Tesla’s longer warranty obligations are priced in.
- 史家琰 doubted China would be Shanghai’s primary market at all, suggesting China-built Megapacks would target Europe, Australia and Southeast Asia. That has held: the Q2 2026 shareholder letter credits Shanghai’s ramp for record EMEA deployments, and Tesla’s August 2026 China statements describe Shanghai Megapack output as going to European and Oceania markets.
- A 人民日报 energy supplement (Feb 2025) quoted 杜笑天 predicting Tesla’s storage production cost would converge with Chinese manufacturers’ by around 2026. Worth revisiting against the Q2 2026 margin compression, which points the other way.
For scale context on the competitive floor: BNEF’s 2025 battery price survey put Chinese stationary storage pack prices at $70/kWh, down 45% year over year.
Absent from the record: no Chinese reporting on a Shanghai phase 2 (二期), no localization of Megapack 3 or Megablock to Lingang, and no Tesla statement to Chinese media about capacity beyond the 80 GWh Lathrop+Shanghai framing and the 50 GWh Houston plant.
9. Demand-side markers (context, not capacity)
- Australia: Western Downs with Neoen — phases 1 and 2 grid-connected at ~540 MW / 1,080 MWh, phase 3 adding ~305 MW / 1,220 MWh for a ~845 MW / 2.3 GWh total. 36氪 claims Tesla held 35% of Australian grid-scale storage share in 2025.
- Japan: Sapporo Helios ~50 MW / 104 MWh in commercial operation; an Orix agreement reported for Japan’s largest storage facility, product presumed shipped from Shanghai.
- Belgium: Sambreville, 76 MW / 304 MWh, in service expected 2027.
- United States: 720 Megapacks (~2.8 GWh) at xAI’s Colossus 2 in Memphis, reported September 2026 as the largest grid battery in the US.
- Order book: Esyasoft, up to $3B for 15+ GWh across the UK, Western Europe, GCC and India. NatPower in Europe — two conflicting characterizations, see discrepancy register.
Verification status
Tier 1 — publishable without caveat. Section 1 (capacity table), the Q2 2026 financial and deployment figures in Section 5, the Q2 call quotations and Musk’s non-answer on supply constraint, and the Shanghai official milestone chronology in Section 3. These rest on the Form 8-K exhibit, the earnings-call transcript, and Shanghai municipal government and Tesla China on-record statements. The deployment series in Section 5 is internally cross-checked: Q1 2026 at 8.8 GWh × 1.53 = 13.5 GWh, matching Tesla’s stated sequential increase; Q1–Q3 2025 at 32.5 GWh plus Q4 at 14.2 GWh = 46.7 GWh, matching the stated full-year figure. Both chains close.
Tier 2 — single-sourced but credible, publishable with attribution. The Brookshire construction and staffing detail; the Shanghai 1,000th-unit date (21世纪经济报道); the LatePost-derived cell-supplier roster and the CATL sole-supplier history; the Megapack pricing figures; the BNEF pack price. Also the LGES contract term dates, which come from trade coverage rather than a filing — if the August 2027–July 2030 window is load-bearing for your piece, pull the LGES disclosure directly, as this is the single most useful 2027–2030 datum in the file.
Tier 3 — subject-authored, mark as such. All Megapack 3 and Megablock performance specifications in Section 6 originate from Tesla’s own launch presentation and have not been independently tested. The “1 GWh in 20 business days,” “23% faster,” “up to 40% lower construction cost,” “91% round-trip efficiency” and “248 MWh per acre” figures are vendor claims. The Baidu Baike entry used only as a foil in the discrepancy register.
Tier 4 — flagged, not laundered. The 36氪 26 GWh output figure (Section 3, rejected on arithmetic). The 36氪 35% Australian market-share claim, which has no visible methodology. The “>$9B in six weeks / 43 GWh” order aggregation, which traces to a social-media tally rather than to company disclosures. The “global capacity exceeding 100 GWh” assertion. The Basenor and optimusk.blog material used anywhere in this file is content-farm output and was used only where it restated something traceable to a Tesla statement or a filing — it should not be cited.
Weakest section. Section 9 (demand-side markers) is the thinnest and should not be published as-is; each project figure needs to be pulled from the developer’s or offtaker’s own announcement. Section 8’s price comparison is directionally sound but the specific Megapack list prices date to early 2025 and Tesla has since guided ASPs down, so they understate the erosion.
The load-bearing unresolved item is discrepancy 1 — Shanghai at 20 versus 40 GWh. Every aggregate capacity figure in circulation, including the widely repeated 130–133 GWh, depends on which one is right. A single question to Tesla IR would settle it.
Compiled September 14, 2026. Sources: Tesla Form 8-K Q2 2026 exhibit 99.1; Tesla Q2 2026 earnings call transcript (July 22, 2026); Tesla Megapack official statements (September 2025, August 2026); 澎湃新闻; 21世纪经济报道; 上海市政府 / 临港新片区; 解放日报; 人民日报能源周刊; TechWeb/新浪财经; 36氪; OFweek储能网; 界面新闻; pv magazine; ess-news; Electrek; Utility Dive.