A 6000 word Claude backgrounder on grid-scale and data center timelines, September 2026. The takeaway for my article is that it takes 6-24 months to install battery storage to reduce demand on the grid. It is faster and cheaper than building new gas generation. Around the world, few new gas generators are planned. The US is an important exception. It generates CO2. This was pulled for my research and ** not ** factchecked fully. Caution!
The question hides four different clocks
“How long does it take to install batteries?” is one of those questions where the answer ranges from four days to nine years depending entirely on what you mean by “install.” The confusion is not semantic pedantry. It sits at the center of nearly every current fight over resource adequacy, data center interconnection, and capacity market design, because the various parties to those fights are routinely quoting different clocks at each other without saying so.
There are four:
Clock 1 — Physical installation. Setting containers on pads, torquing busbars, pulling cable. Measured in days to weeks. This is what battery vendors quote in press releases.
Clock 2 — Construction. Notice-to-proceed through mechanical completion: site prep, foundations, drainage, fencing, the medium-voltage collection system, the substation, the fire suppression, the SCADA. Measured in months.
Clock 3 — Development to commercial operation. Site control, permitting, offtake, financing, procurement, construction, energization, commissioning, market registration. Measured in years.
Clock 4 — Queue position. How long the project sits in an interconnection process before any of the above can begin in earnest. Also measured in years, and in most Western markets this is now the binding constraint.
The International Energy Agency put usable numbers on the spread in a May 2026 commentary. Median construction time for utility-scale batteries globally runs about 275 days — close to solar PV at roughly 220 days, and far below gas at over two years or nuclear at more than six. But the IEA’s more important finding is the gap between that figure and total time to market: in Europe, the United States and Japan, battery projects typically take around two to two-and-a-half years to become operational. Timelines are shorter in China and parts of the Middle East. The agency’s summary is blunt — non-construction phases often account for more than half of total project timelines.
That ratio is the single most useful thing to carry into any conversation about battery deployment speed. The batteries are not the slow part. Almost nothing about the schedule is determined by lithium.
Clock 1: the physical install is genuinely fast, and getting faster
The reason batteries can be built quickly at all is modularity. A modern grid-scale system is a fleet of factory-assembled enclosures, each containing racked lithium-iron-phosphate cells, a battery management system, thermal management, and increasingly the power conversion electronics. The site work is foundations, cabling, and interconnection hardware. There is no boiler, no turbine hall, no fuel handling, no cooling tower.
Hornsdale, South Australia (2017) remains the canonical demonstration, and it is worth restating precisely because it is so often garbled. Elon Musk publicly wagered in March 2017 that Tesla would build a 100 MW system within 100 days of contract signature or provide it free. Contracts were signed at the end of September 2017; the 100 MW / 129 MWh Hornsdale Power Reserve was plugged into the South Australian grid on 1 December 2017. That is roughly 60 days on the grid-connection clock and comfortably inside the 100-day guarantee. Some secondary sources say “100 days,” others “60 days,” and at least one project-management retrospective says Tesla finished about 40 days early. These are not contradictions so much as different start points, but the discrepancy is worth flagging rather than resolving silently — Hornsdale gets cited as precedent constantly and the number people quote is often untraceable to a specific milestone.
What Hornsdale actually proved was narrow: that if you have a site with an existing wind farm connection, a motivated state government, a supplier willing to eat schedule risk, and no meaningful queue, the physical build is not the constraint.
Tesla’s Megablock (2025–26) is the industrialized version of that lesson. Announced in September 2025 and now entering production at the company’s Brookshire, Texas Megafactory outside Houston, Megablock bundles four Megapack 3 units with a medium-voltage transformer and switchgear into a single pre-engineered assembly. The specific claim that matters for schedule: an individual Megapack installation can require up to 24 cable connections per unit, while a Megablock needs three busbar connections. Tesla says this cuts installation time by 23% and construction cost by up to 40%, and that it can commission 1 GWh in 20 business days at a site density of 248 MWh per acre. Tesla’s VP of energy and charging, Mike Snyder, has stated the 20-business-day figure publicly.
Treat the number as a vendor target rather than a demonstrated field average — it has not yet been independently verified across a portfolio of sites — but the underlying engineering logic is sound and the direction of travel is real. The industry is systematically moving labor from the field into the factory. Tesla reported over 77 GWh of Megapack units operating worldwide, with 8.8 GWh deployed in Q1 2026 and 13.5 GWh in Q2. The Texas plant targets 50 GWh/year of Megapack 3 and Megablock capacity, with first deliveries expected around the end of 2026.
Manufacturing throughput is a separate speed story. For the Saudi projects supplied by Sungrow and BYD, manufacturing of the BESS units was reported complete in 58 days before shipping from Taicang port in Jiangsu. Chinese suppliers can now build a multi-gigawatt-hour order faster than most Western jurisdictions can process a grid connection application.
Inside a data center, the physical install can be a week. A colocation operator retrofitting an active facility from valve-regulated lead-acid to lithium-ion described unloading the UPS, switching to bypass, demolishing the old strings, and installing, starting up and commissioning 24 lithium-ion cabinets in one week per system. Eaton notes that lithium UPS batteries ship at 30% state of charge and need roughly four hours of charge time, with cabinet systems shipped disassembled and assembly and commissioning included in the price. None of this is schedule-critical for a facility. The batteries in a UPS lineup are the fastest thing in the building.
Clock 2: construction benchmarks, market by market
Here the numbers start to diverge sharply by geography — and the divergence is almost entirely institutional, not technical.
United States
Spearmint Energy, ERCOT. Two 100 MW / 200 MWh standalone systems, Tierra Seca in Del Rio and Seven Flags in Laredo, built by Mortenson using Sungrow’s PowerTitan 2.0 platform. Spearmint closed more than $250 million of construction financing in Q2 2025 and began construction; both projects began operation in late December 2025. That is roughly six to eight months from notice to proceed to operation for 400 MWh combined — about as fast as US utility-scale construction gets, and it happened in the one US market with a genuinely functional interconnection process.
Burns & McDonnell, West Texas. Three 10 MW / 20 MWh standalone systems, referred to collectively as the Texas Trio, energized on what the EPC described as an aggressive five-month construction schedule. Small systems, but a useful floor.
Georgia Power. A useful counterpoint from a regulated utility. The company announced in May 2025 that construction was underway on 765 MW of BESS across four sites authorized through the Integrated Resource Plan process. Groundbreakings at Hammond and McGrau Ford took place in April 2025, with projected commercial operation dates spanning May 2026 (Moody, 49.5 MW) through October 2026 (McGrau Ford Phase I). Roughly 12–18 months from groundbreaking, with the Hammond project leveraging existing infrastructure at a retired coal plant.
Energy Vault, SOSA Energy Center. A 150 MW / 300 MWh project in Madison County, Texas, acquired from Shell subsidiary Savion, which began construction in Q4 2025 with expected commercial operation by Q2 2027 — about 18 months. The company also commenced off-site construction on three further projects in Q4 2025 with expected commercial operation between Q4 2027 and Q4 2028, which tells you something about how far ahead of field mobilization procurement now has to run.
Saudi Arabia and the Gulf
Bisha, Asir Province. Saudi Arabia’s 500 MW / 2,000 MWh system, supplied by BYD with 122 prefabricated units and built by PowerChina Hubei Engineering, connected to the grid in January 2025 — eleven months after the start of construction. For a 2 GWh facility, that is remarkable, and it is the single best benchmark available for what a large BESS build looks like when permitting and interconnection are handled administratively rather than adversarially.
The five-site Saudi Energy program. National Grid Saudi Arabia awarded Alfanar (EPC) and BYD Energy Storage (design, supply, installation supervision, testing, commissioning) a contract in January 2025 for five facilities totaling 2,500 MW / 12,500 MWh across Riyadh, Rabigh, Dawadmi, Al Jouf and Qassim. Grid-tied charge and discharge testing completed at the end of May 2026, with commissioning announced in early June. That is roughly seventeen months from contract award to commissioning for 12.5 GWh across five sites — the largest grid-scale storage deployment in the world at the time, requiring 2,364 system cabinets.
For comparison: 12.5 GWh is roughly twice the total operating grid-scale battery capacity in Germany.
Australia
Collie Battery Stage 2, Western Australia. Neoen delivered 341 MW / 1,363 MWh ahead of schedule for the start of a four-hour AEMO capacity contract on 1 October 2025 — less than fifteen months from the start of construction.
Waratah Super Battery, New South Wales. The cautionary tale, and the most instructive project on this list. Akaysha Energy began construction at the former Munmorah coal station site in May 2023 and finished the battery build in October 2024 — seventeen months for 850 MW / 1,680 MWh and roughly 2,600 battery units. Excellent. Then commissioning began. First-stage energization completed September 2024. Partial operation as a grid shock absorber began in August 2025 at about half capacity. In October 2025 a transformer failed. Replacement took approximately ten months, and only that fast because the original manufacturer, Wilson Transformer Company in Victoria, still held the engineering records and had long-lead materials secured. As of early September 2026 the rebuilt transformer has been energized and the project is approaching full capacity.
Construction: 17 months. Construction start to full operation: about 40 months. The delta is entirely commissioning, testing, market registration and one piece of failed high-voltage equipment.
China
Chinese timelines are on a different scale, and Chinese sources report them with a specificity that Western press releases usually avoid.
Luzimiao, Inner Mongolia (CECEP Wind Power / 节能风电). A 200 MW / 800 MWh standalone storage station in Ulanqab. Per the company’s account carried by Caijing, Sina Finance and Securities Times: project filing and construction start on 5 June 2025; full-station energization achieved 16 December; back-feed completed 18 December; three full charge-discharge cycles completed and full-capacity grid connection achieved 30 December 2025 — one day ahead of the scheduled milestone. Just over six months from project registration to full grid connection for 800 MWh.
This one has since been corroborated by an independent site visit. A reported piece in 企业观察报 (Enterprise Observer), republished in July 2026, confirms the June 2025 filing and 30 December 2025 full-capacity connection, describes 160 battery cabinets on a four-hour LFP configuration with round-trip efficiency around 90%, and adds commercial detail the company release omitted: the station sits on Inner Mongolia’s standalone storage list and receives a discharge compensation of ¥0.35/kWh for ten years, which accounts for roughly 90% of its revenue and is decoupled from spot prices. The site is staffed around the clock, and the operator is candid that early operation included days with negative margins from buying high and selling low.
One correction to the data center framing. The company release ties Luzimiao to the local “Grassland Cloud Valley” (草原云谷) computing industry. The independent reporting foregrounds a different load: Ulanqab is China’s largest prefecture-level ferroalloy production center, with annual consumption measured in the tens of billions of kWh, and submerged-arc furnaces that cannot easily be stopped once running. Both are real, but the storage-for-data-centers rationale is the promotional framing; the industrial load is what the on-the-ground reporting emphasizes. Treat Luzimiao as a source-grid-load-storage integration case with a compute component, not as a data center storage project.
Dabancheng, Xinjiang (same developer). 100 MW / 400 MWh. Construction started September 2025. Facing a complex site and a compressed schedule, the regional company established a daily-control, weekly-dispatch, monthly-review management cycle and put more than 160 workers on site per day at peak. All buildings and equipment installation in the storage area were complete in three months. Back-feed completed 4 February 2026, entering trial operation, with grid-compliance testing to follow.
Linfen Ru’an, Shanxi. Reported as entering operation in April 2024, five months ahead of schedule, as part of a program in which 24 new stations were commissioned in under three years.
The aggregate picture: China added roughly 65.87 GW / 190.02 GWh of new-type storage grid connections in 2025, up 69% year on year, following the removal of mandatory renewables-plus-storage co-location requirements and a shift the industry describes as moving from policy-driven to market-driven. The IEA puts China’s 2025 battery additions at just over 63 GW, of which about 55 GW was utility-scale.
Germany
Germany is where the gap between “how fast can you build a battery” and “how fast can you connect one” is widest, and where the reporting is most granular.
Current large projects, per a September 2026 survey in ingenieur.de (VDI Verlag):
| Project | Size | Construction start | Target commissioning |
|---|---|---|---|
| Klostermansfeld, Saxony-Anhalt (BW ESS) | 1,000 MW / up to 5,700 MWh | 1 July 2026 | — |
| Waltrop, NRW (Trianel/Luxcara/BKW) | 900 MW / ≥1,800 MWh | 24 August 2026 | End 2028 |
| Boxberg, Saxony (LEAG) | 400 MW / 1,600 MWh | 31 August 2026 | Late 2027–early 2028 |
| Philippsburg, Baden-Württemberg | 400 MW / 800 MWh | Under construction | End 2027 |
| Förderstedt, Saxony-Anhalt (Eco Stor) | 300 MW / 718 MWh | November 2025 | Full operation 2027 |
| Gundremmingen, Bavaria (RWE) | 400 MW / 700 MWh | 29 October 2025 | Early 2027 |
| Jänschwalde, Brandenburg (LEAG) | 1,000 MW / 4,000 MWh | In preparation | Late 2027–early 2028 |
The Waltrop schedule is the one to sit with: construction began in August 2026, major components are scheduled to arrive from July 2027, and commissioning is planned for end 2028. Roughly 28 months, with an eleven-month gap between site start and equipment delivery. That is not a battery installation schedule. That is a heavy electrical infrastructure schedule dictated by long-lead procurement.
RWE’s Gundremmingen project is the fastest on the list at roughly fifteen months, and the reason is stated plainly: it reuses the existing grid connection of the nuclear plant currently being decommissioned on the site. More than 200 containers holding around 850,000 LFP cells and over 100 inverters, on a €230 million budget, at a site where the high-voltage connection already exists.
Kupferzell. TransnetBW’s 250 MW / 250 MWh grid booster in Baden-Württemberg is the most complete German case study available, because the entire permitting and construction arc is documented publicly by the developer itself.
From TransnetBW’s own project diary and press releases: Fluence contracted as general contractor in 2022; permit documents submitted to the Stuttgart Regional Council; early-start permit for site preparation granted August 2023, with site setup from September 2023; full planning approval decision (Planfeststellungsbeschluss) on 20 March 2024, following a hearing in late November 2023 and a public participation process; official groundbreaking 10 June 2024. The 275-tonne 380/30 kV grid-booster transformer was delivered at the end of November 2024, routed through several village centers because of roadworks on the B19. Operating building concrete complete January 2025; the 30 subnest slabs — roughly 520 m² each, about 15,600 m² total, or slightly more than two football pitches — poured through early 2025; cable ducts installed March 2025; cabling scheduled June to September 2025. Estimated cost €188 million, using 428 CATL battery cubes. The building will be unstaffed in operation, controlled from TransnetBW’s main control center in Wendlingen.
Original target commissioning: 2025, and in one engineering contractor’s project description specifically 1 January 2025. Current target per trade reporting: during 2026.
A verification flag worth carrying: TransnetBW’s project diary explicitly promises regular updates through to commissioning. Its most recent entry, as of a direct check on 7 September 2026, is dated 15 May 2025 — sixteen months of silence on a page designed to track exactly this. That is not proof of anything; developers let web pages go stale. But no commissioning announcement has been published by the operator, and anyone stating that Kupferzell is in service should confirm it directly rather than inferring it from the 2026 target.
Commissioning here is also not purely technical. Local reporting notes that an independent system test follows completion of all construction work, and that final building-law acceptance by the county administration — primarily checking fire protection compliance — must precede the operating permit.
Clock 3: the commissioning tail nobody budgets for
Between “the batteries are installed” and “the asset earns money” sits a phase that consistently surprises developers and consistently gets omitted from the timelines cited in policy debates.
The sequence, roughly: back-feed the station (energize auxiliary systems from the grid), commission the protection and control systems, complete grid-compliance testing, perform capacity verification through full charge-discharge cycles, obtain the operating permit, and register with the market operator.
Chinese reporting names these steps explicitly — 反送电 (back-feed), 涉网检测 (grid-compliance testing), 三充三放 (three charge/discharge cycles) — and treats the interval between them as a project management KPI. Luzimiao ran from full-station energization to full-capacity grid connection in fourteen days. Australian and German projects routinely take months over the same ground.
Waratah is the extreme case: energized September 2024, partial service August 2025, full capacity September 2026. The Australian Energy Market Operator registration and System Integrity Protection Scheme service requirements added layers that a merchant arbitrage asset would not face, and the transformer failure was bad luck. But bad luck at the commissioning stage is not rare, and the replacement timeline — ten months, described as fast — tells you what a transformer failure costs when the manufacturer is not around the corner with the drawings.
For anyone modeling capacity accreditation or resource adequacy contributions, the commissioning tail is the part of the schedule most likely to slip and least likely to be disclosed in advance.
Clock 4: interconnection, which is usually the answer
Everything above is largely irrelevant in most Western markets, because the binding constraint is upstream.
United States
Lawrence Berkeley National Laboratory’s Queued Up series is the reference. The 2026 edition, published June 2026 with data through end-2025 and compiled with GridTracker across seven ISO/RTOs and 50 non-ISO utilities (~98% of installed US generating capacity), reports:
- Over 2,060 GW of total generation and storage capacity actively seeking connection.
- Roughly 8,200 projects, representing 1,312 GW of generation and approximately 749 GW of storage — storage down 16% year on year, as active gas capacity rose 86% to 253 GW.
- 549 GW with a draft or executed interconnection agreement but not yet operational, including 161 GW of storage.
- Median duration from interconnection request to commercial operation of over five years for projects built in 2025.
- Only 13% of capacity that submitted requests between 2000 and 2020 had reached commercial operation by the end of 2025; 75% was withdrawn and 10% remained active.
LBNL also cautions that FERC Order 2023 and related reforms are still being implemented and that it is too early to assess their effect — worth keeping in view when anyone cites queue reform as a solved problem.
Two widely repeated figures deserve a note. The “55 months average” for projects reaching COD in 2024, against 36 months in 2015 and 22 months in 2008, comes from third-party summaries of the 2025 edition rather than from LBNL’s own headline findings; the 2026 edition’s published metric is the median exceeding five years. Similarly, the frequently cited size tiers — under 20 MW moving through in 11–18 months, above 20 MW at 30 months or more, above 100 MW at a median exceeding four years — trace to trade-press reporting on an earlier edition and should be dated accordingly rather than presented as current.
Regional wait times commonly cited for data center load: a national average around four years, with Northern Virginia at approximately seven.
PJM
Relevant to the current docket activity. PJM has processed over 170 GW from July 2023 to December 2025 under its transition to a first-ready, first-served model, with 26 GW through the expedited Fast Track process and 30 GW scheduled for processing in 2026. The one-time Reliability Resource Initiative drew 94 applications totaling 26.6 GW against a 51-project cap, with studies completing at the end of 2026. Notably, nearly 5,700 MW of proposed storage was not selected, including 1,120 MW in Maryland, and most selected capacity is not expected online until 2030 or 2031.
Modo Energy’s analysis of the Fast Lane cohort found 71 battery projects totaling over 4 GW, with most expected to begin commercial operations between 2027 and 2029, and 97% of those disclosing duration specifying four hours or more.
PJM’s Reliability Backstop Procurement was filed with FERC on 31 July 2026 in Docket ER26-3380-000, addressing the 6,831 MW shortfall left when the 2028/2029 Base Residual Auction cleared short on 14 July 2026. It requires resources to demonstrate commercial operation no later than 1 June 2032 and subjects them to a COD feasibility evaluation covering interconnection status, site control, project financing, equipment purchases and construction schedules. Failure to pass the gating evaluation disqualifies a resource from offering.
That gating evaluation is, functionally, a regulator formalizing the four-clocks problem: PJM is no longer willing to take a developer’s word on schedule.
Status as of 7 September 2026: FERC has not yet ruled. Developer advisories anticipate an order around 29 September, with the central procurement window running 30 September to 21 October and results finalized 2 December, ahead of the 2029/2030 Base Residual Auction opening 9 December. Note that PJM’s own June 2026 CIFP proposal paper described a bid solicitation window of 10 September to 9 October and central procurement running to 20 November — the currently circulating dates differ, and the whole timeline is explicitly subject to FERC action. Anyone publishing against these dates should re-check the docket.
Germany
The German figures are almost comic in their scale mismatch. Per the Bundesnetzagentur, published via SMARD:
- 2025 connection requests for battery storage: 573.5 GW of power and roughly 1,583 GWh of capacity — up 43% and 139% respectively.
- Total applications received in 2025: 18,158, a 194% increase on 2024 (17,775 to distribution networks, 383 to transmission).
- 2025 connection commitments: 54.2 GW and 142.4 GWh, across 1,937 approvals — up 119% and 207% on 2024.
- At distribution level, 17,775 requests covering 406.8 GW and 1,185.9 GWh came in; 1,429 of that same year’s applications received a commitment, totaling 16.8 GW and 59.2 GWh.
- Transmission operators issued 71 commitments in 2025 covering 27.3 GW and 59.7 GWh.
The much-repeated “roughly 8% approval rate” is pv magazine Deutschland‘s arithmetic on the BNetzA data — 1,429 of 17,775 same-year distribution applications, or about 4% by requested capacity. It is a sound derivation, but it is a derivation, not a published regulator statistic, and it compares same-year applications against same-year decisions, which understates eventual approval since many applications are decided in later years.
The 573.5 GW figure overstates real pipeline, and the Bundesnetzagentur says so explicitly: developers file at multiple candidate connection points for one project, and where several are approved only one is normally built. The regulator also notes the dataset does not yet distinguish grid-coupled from generation-coupled storage, a gap it intends to close.
Installed base — a live discrepancy. Reporting citing BNetzA puts 1,292 battery storage systems at medium voltage and above in operation, totaling around 3.6 GW and 6 GWh. A separate Marktstammdatenregister evaluation, as of 15 October 2025, gives roughly 2.4 GW gross and 3.2 GWh usable for large battery storage. These are different definitions on different dates rather than a contradiction, but they are close enough in framing that they get quoted interchangeably. Say which one you mean.
Bavaria’s largest distribution network operator, Bayernwerk, and Mitnetz Strom have both indicated typical waits of at least five years for large storage connections. Since 1 April 2026 the four transmission system operators have applied a maturity-based process (Reifegradverfahren), collecting applications in fixed cycles and assessing project readiness rather than processing purely by filing date. The Bundesnetzagentur has also clarified that a physical connection does not automatically confer the requested capacity in both directions — a real issue for storage, which is a load when charging and a generator when discharging.
For data centers specifically: a September 2026 commentary in pv magazine Deutschland states that in Frankfurt, Germany’s largest data center hub, new large projects have essentially no prospect of a grid connection before 2030.
The equipment that isn’t the battery
Cells have stopped being the problem. LFP accounted for roughly 90% of newly installed stationary storage globally in 2025, up from well under 50% five years earlier, and battery costs fell more than 90% between 2010 and 2025. Cell lead times have drifted from around 12–16 weeks to 20–26 weeks, which is inconvenient rather than schedule-breaking.
The problem is everything downstream of the DC bus.
Large power transformers. Pre-pandemic lead times of roughly 7–14 months have stretched well beyond 24 months in most major markets, with some specialized units approaching 36–48 months, per PTR Inc analysis published in March 2026. European lead times for large power transformers are reported in the 48–60 month range. Asia-Pacific is the shortest at around 12 months. Wood Mackenzie’s Q2 2025 survey put power transformers at an average of 128 weeks and generator step-up units at 144 weeks. NEMA has told trade press that an order shipping in four to six weeks five years ago now ships in 80 to 210 weeks.
Medium-voltage switchgear. 2026 lead times reported by procurement specialists: 52–72 weeks for 5kV/15kV metal-enclosed, 60–80 weeks for 15kV/27kV metal-clad, 78–104 weeks for 38kV. Pad-mount transformers 40–65 weeks.
Integrated BESS with PCS and switchgear. Multi-megawatt orders now carry 20–35 week lead times, because the upstream transformer and semiconductor constraints propagate.
The structural causes are consistent across analyses: highly customized engineering, long manufacture-and-test cycles, dependence on a small number of specialist component suppliers (high-voltage bushings and on-load tap changers are repeatedly flagged), and a limited global base of manufacturers capable of the largest units. Domestic capacity expansions in the US do not automatically compress delivery windows; OEMs cite raw material constraints and technical labor shortages as limits on their own expansion.
Integrator capacity is now also binding. Fluence cut its FY2026 guidance on manufacturing delays despite record orders and a large backlog, following a Samsung SDI warning that demand would outstrip production. The practical procurement observation from that cycle: allocation and inventory have replaced price as the primary variable.
This is why the Megablock proposition matters beyond Tesla’s marketing. Bundling the transformer and switchgear into a factory-built assembly does not eliminate the lead time, but it moves it into a manufacturer’s production plan rather than a project’s critical path, and it standardizes the unit so that capacity can be built ahead of specific orders.
Permitting and social license
The other non-technical clock, and the one with the sharpest recent inflection.
The 16 January 2025 fire at Vistra’s Moss Landing Phase 1 facility in Monterey County, California — the largest battery storage fire in US history, burning for days, incinerating more than 50,000 batteries and prompting the evacuation of roughly 1,200 residents — triggered a permitting cascade that is still working through.
Within weeks: Orange County’s Board of Supervisors passed an urgency ordinance placing a moratorium on permitting large-scale BESS in unincorporated areas while new regulations were developed. Morro Bay adopted a 45-day moratorium. Assembly member Dawn Addis introduced AB 303, which would have removed the California Energy Commission’s AB 205 opt-in certification pathway for facilities of 200 MWh or more, imposed a 3,200-foot setback from homes and schools, and barred siting in environmentally sensitive areas. SB 283 (Laird) took a different tack, focused on fire-official oversight of permitting and maintenance. The CPUC proposed new maintenance and operations standards and stronger emergency response action plan requirements. By October 2025 Monterey County itself had taken the first formal step toward a moratorium of a year or more.
Solano County provides the pattern to watch: a moratorium imposed after an earlier fire, lifted in August 2025 but replaced with rules restricting BESS to industrial or manufacturing-zoned land.
Germany has had its own version. A battery storage fire in Bautzen in July 2026 required a fire service deployment lasting more than a day.
The practical planning implication, from BESS installation guidance: permitting and authority-having-jurisdiction engagement should begin three to six months before construction, and the utility interconnection application should be filed on day one of the project, before design is finalized. The cost of getting this wrong is not marginal — permit rejection after engineering is complete generates rework, and misjudged interconnection can add 18–24 months.
Data centers: a different problem with a different battery
Data centers use batteries for at least four distinct jobs, and conflating them produces bad timeline estimates.
1. UPS ride-through. Seconds to minutes of runtime bridging to generator start. Hyperscale operators typically design for 1–2 minutes; colocation for around 5; financial installations for 10–15. Physical replacement of a UPS battery bank is a one-week-per-system exercise; a full planned replacement program including assessment, specification, procurement and commissioning is typically scoped at around twelve weeks. VRLA strings get replaced every 3–5 years; lithium systems every 8–12.
2. Load smoothing. AI training workloads swing power draw violently, and batteries can absorb ramps from roughly 10% to 90% of capacity in milliseconds. Chinese industry commentary is direct on why this matters: with tens of thousands of GPUs in a training run, even a millisecond interruption destroys the work.
3. Virtual capacity expansion. This is the most interesting one from a regulatory standpoint. A battery lets a facility draw more peak load than its transformer capacity nominally allows, by discharging into the excess. Chinese reporting on the Shenzhen storage supply chain describes this explicitly as 虚拟增容 — virtual capacity increase — and frames it as the response to overseas grid upgrades that take three to five years or never get approved at all.
4. Bridge power. Covering peak load from day one of operation while the permanent grid connection catches up.
What this does to timelines
The bridge-power case has produced the most aggressive schedules on record.
xAI Colossus, Memphis. Construction began at a former Electrolux factory in April 2024; Phase 1 with 100,000 H100 GPUs went live 122 days later in July 2024, roughly 250 MW after the December 2024 expansion to 200,000 GPUs. The substation is the relevant data point: xAI and Memphis Light, Gas and Water designed and built a 150 MW substation in 97 days against a normal timeline of roughly 2.5 years. In the interim, temporary gas turbines and Tesla Megapacks carried the load — approximately 156 Megapacks supporting the first site, with 168 more delivered for Colossus 2. The turbines drew sustained legal and community opposition from the Southern Environmental Law Center on behalf of the NAACP and Young, Gifted & Green, and were partially demobilized once the substation energized.
Whatever one thinks of the environmental record, the engineering point stands: batteries were the only asset class that could be deployed on the timescale the compute buildout demanded.
Aligned Data Centers announced a 31 MW / 62 MWh battery alongside a Pacific Northwest facility explicitly sized to secure interconnection years earlier than traditional utility upgrades would allow — a two-hour system, planned operational in 2026.
The generalized claim. Analysis from Verse contrasts average grid-connection waits without storage (roughly 66 months) against 18–24 months for data centers paired with behind-the-meter storage and dispatch software that can qualify for accelerated interconnection mechanisms. Treat the specific arbitrage figure with caution — it comes from a vendor with a commercial interest in the comparison — but the mechanism is real and utilities are increasingly writing it into tariffs: storage plus a firm demand-response commitment converts a firm load request into a flexible one, which is a different queue.
Wood Mackenzie’s global head of energy storage, Allison Weis, has framed the broader trend: with average grid waits around four years, developers are pairing gas generators with storage as bridge power because bringing your own generation is faster than waiting. BNEF projects energy storage supporting 9.8 GWh of gas generation at data centers through 2030. At the extreme end, Pacifico Energy’s GW Ranch off-grid project in West Texas is planned with 1.8 GW of battery storage alongside 7.65 GW of gas.
China’s version: 算电协同
China has institutionalized the data-center-plus-storage pairing under the banner of 算电协同, roughly “compute-power coordination,” which was written into the 2026 Government Work Report. The concrete projects are worth knowing because they are the fastest-built integrated facilities anywhere:
- Ulanqab, Inner Mongolia (July 2025): China’s first data center green-power direct-connection and integrated source-grid-load-storage project — Zhongjin Data’s zero-carbon compute base, pairing 200 MW wind, 100 MW solar and 45 MW storage, with 848 GWh/year of self-generated green power and a 38.74% renewable substitution rate.
- Karamay, Xinjiang (December 2025): Guangdong Energy Group’s 300 MW solar compute-power project with 30 MW / 60 MWh storage and a 500-PFLOPS intelligent computing center — described as the first packaging of solar, storage, compute and network into a single integrated build.
- Qingyang, Gansu (April 2026): The first units of a gigawatt-class green power aggregation pilot for the East Data West Computing industrial park connected to the grid. Total planned capacity 2 GW; the first 1 GW phase went from construction start in August 2025 to grid connection in under six months.
- China Telecom Anhui intelligent computing center: A 25 MW / 200 MWh storage package, tendered and awarded to Kedaguochuang in March 2025.
The Ministry of Industry and Information Technology’s February 2025 action plan on high-quality development of the new-type storage manufacturing industry explicitly directed deployment of storage in data centers and intelligent computing centers as high-reliability, high-load scenarios.
Meanwhile the German industry association BVES has been pushing back on the framing of batteries and data centers as competitors for scarce connection capacity, arguing that integrated projects combining renewables, storage and AI compute should be treated as a single system in connection procedures rather than as separate queue entrants. That is essentially the same argument PJM is working through under its Bring Your Own New Capacity and Interim Resource Adequacy Service constructs.
Working numbers
For planning purposes, ranges that hold up against the evidence above:
| Scope | Realistic range | Notes |
|---|---|---|
| UPS battery string swap, one system | ~1 week | Plus 4 hours charge; batteries ship at 30% SoC |
| Full UPS replacement program | ~12 weeks | Assessment through commissioning |
| Commercial cabinet system | 2–4 weeks site work | Behind existing connection |
| Utility-scale containerized, construction only | 5–12 months | Median ~275 days globally (IEA) |
| Utility-scale, 100–300 MW, construction only | 9–18 months | Australian EPC benchmark |
| German 400 MW+ project, groundbreaking to COD | 15–28 months | Faster where an existing connection is reused |
| Chinese 100–200 MW, filing to grid connection | 6 months | Best observed; not a Western benchmark |
| Saudi 500 MW / 2 GWh, construction to grid | 11 months | Bisha |
| Total time to market, EU / US / Japan | 24–30 months | IEA; non-construction >half of it |
| US interconnection request to COD (median) | >5 years | LBNL, projects completed 2025 |
| Northern Virginia data center grid connection | ~7 years | Widely cited market figure |
| Data center + BTM storage, accelerated interconnection | 18–24 months | Vendor-sourced; mechanism is real |
| Large power transformer | 24–48 months | Europe 48–60 months for some units |
| MV switchgear | 52–104 weeks | By voltage class |
What to watch
Whether Megablock-class products actually compress field schedules. Tesla’s Brookshire plant reaching 50 GWh/year with pre-integrated transformers and switchgear would move a genuine bottleneck out of the project critical path. The 20-business-days-per-GWh claim needs a year of field data before anyone should plan around it.
Germany’s Reifegradverfahren and the AgNeS network tariff reform. The maturity-based connection process, effective April 2026, is the first serious attempt to triage a 573 GW request pile. The Bundesnetzagentur published its final AgNeS determination draft on 6 August 2026, consulting to 18 August, with capacity-based charges for storage replacing energy charges from 2029 and grandfathering for advanced projects. The current §118(6) EnWG exemption horizon of 4 August 2029 is driving a visible rush.
PJM’s Reliability Backstop Procurement COD gating. The requirement to document interconnection status, site control, financing, equipment purchases and construction schedules against a 1 June 2032 deadline is the first US capacity construct to formally underwrite schedule credibility rather than accepting it. How PJM scores storage against gas on COD feasibility will tell you a great deal about which technology the market is actually betting on for the 2028/29 shortfall.
Foreign entity of concern sourcing rules. With LFP cell supply concentrated in China and FEOC compliance now a financing precondition in US projects, the procurement calendar and the schedule are increasingly the same conversation.
Whether the post-Moss Landing permitting friction stabilizes or compounds. California’s moratoria have mostly converted into siting restrictions rather than bans, which is the survivable outcome. A second large fire in a populated area would not necessarily follow the same path.
Verification status
Second-pass review completed 7 September 2026.
Tier 1 — independently confirmed or primary-sourced
LBNL interconnection data. Read directly from Berkeley Lab’s Queued Up landing page and 2026 edition summary, not from secondary coverage. Dataset is CC BY 4.0, compiled with GridTracker.
German connection-request data. Verified against SMARD, the Bundesnetzagentur’s own transparency platform, which is the origin of the 573.5 GW / 54.2 GW / 18,158 figures.
IEA construction and time-to-market benchmarks. Signed IEA commentary (Wanner, Schoenfisch, Ringkjøb, 29 May 2026), CC BY 4.0, based on BNEF and Benchmark Mineral Intelligence data.
Kupferzell construction milestones. TransnetBW’s own project diary, fetched directly, plus operator press releases and the Stuttgart Regional Council’s planning decision.
Waratah Super Battery milestones. EnergyCo (NSW Government) project page plus Akaysha statements and contemporaneous trade coverage.
PJM RBP structure and dates. PJM’s Board decisional letter of 27 July 2026 and the ER26-3380-000 filing of 31 July 2026.
Tier 2 — single-sourced but credible; attribute
Bisha — 11 months from construction start to grid connection. MEED, January 2025. No second independent source found for the eleven-month figure specifically.
Spearmint’s ERCOT construction window. The six-to-eight-month figure is a derived estimate from two company announcements; Spearmint never states a construction duration.
Collie Stage 2 at under fifteen months. Utility Magazine, reporting Neoen’s claim.
Individual German project construction dates. The September 2026 ingenieur.de survey is well-sourced trade press citing BNetzA, SMARD, dena, VDE FNN and LEAG, but several individual project dates are single-sourced there.
Tier 3 — subject-authored; marked as such
Tesla’s Megablock claims. The 23% installation time reduction, 40% cost reduction, 248 MWh/acre and 1 GWh in 20 business days are company statements throughout, repeated across trade press without independent field verification. The deployment rate has not been audited by anyone outside Tesla.
Chinese project timelines — upgraded, with a correction. Luzimiao has moved from Tier 3 to Tier 2: an 企业观察报 site-visit piece independently confirms the milestones and adds unflattering operational detail the company release omitted. That reporting also corrected the framing — the Ulanqab load story is substantially ferroalloy, not data centers. Dabancheng remains Tier 3, resting entirely on the company account.
The structural caveat stands regardless of tier: “project filing to grid connection” in China is not the same regulatory object as “interconnection request to COD” in PJM. Chinese filing (备案) happens after siting and grid coordination are already substantially settled. The six-month figure is real but is not measuring the same interval as the LBNL five-year median. Do not put them side by side without saying so.
The China 2025 aggregate (65.87 GW / 190.02 GWh, +69%) comes from a company article citing unnamed industry statistics. The IEA independently estimates just over 63 GW for China in 2025 — broadly consistent, which raises confidence without confirming the precise figure.
The Verse comparison — 66 months without storage versus 21 months with a behind-the-meter battery — is vendor-authored by a company selling the dispatch software that enables the arbitrage. The mechanism is independently documented; the numbers are marketing.
Unresolved discrepancies
Hornsdale: 60 days, 100 days, or 40 days early? Contemporaneous wire reporting establishes grid connection on 1 December 2017 against a clock starting at contract signature at the end of September — roughly 60 days. A Project Management Institute retrospective says completion came about 40 days ahead of schedule. Aggregator sites state flatly that the build took 60 days, conflating the grid-connection clock with the construction clock. Safest formulation: Tesla delivered inside the 100-day guarantee.
PJM RBP procurement window. PJM’s own June 2026 CIFP proposal paper describes bid solicitation 10 September – 9 October and central procurement running to 20 November. Developer advisories circulating in August 2026 give 30 September – 21 October with results 2 December. The timeline is expressly subject to FERC action.
German installed base. BNetzA-sourced reporting gives 1,292 systems / 3.6 GW / 6 GWh at medium voltage and above; a Marktstammdatenregister evaluation as of 15 October 2025 gives roughly 2.4 GW gross and 3.2 GWh usable. Different definitions and dates, not a contradiction — but they get quoted interchangeably.
PJM shortfall size. One legal-industry summary characterizes the RBP as addressing an approximately 60 GW shortfall. PJM’s own filing and multiple developer advisories put it at 6,831 MW from the 2028/2029 BRA. The 6.8 GW figure is correct.
Transformer and switchgear lead times. PTR Inc (March 2026), Wood Mackenzie (Q2 2025), NEMA statements to Utility Dive, and procurement-firm data give overlapping but non-identical ranges. The spread here is deliberate.
Still open
Kupferzell’s operational status. TransnetBW’s project diary has not been updated since 15 May 2025 and no commissioning announcement has been published by the operator. Trade reporting gives a 2026 target.
FERC action on ER26-3380. No order as of 7 September 2026.
Bisha construction start date. The eleven-month claim would be stronger with a documented groundbreaking date.
Compiled September 2026. Sources include the IEA, Lawrence Berkeley National Laboratory, Bundesnetzagentur via SMARD, ingenieur.de and pv magazine Deutschland, Caijing/Sina Finance, 企业观察报, MEED, Energy-Storage.news, EnergyCo NSW, TransnetBW, RWE, PJM Inside Lines, Modo Energy, Southern Environmental Law Center, and DataCenterDynamics.