The 20-foot container became the industry's unit of account

A standard 20-foot ISO container now holds roughly 5 MWh of usable LFP energy in a liquid-cooled configuration using 314 Ah cells, and 6.25 MWh with 587 Ah cells entering volume production. That is a four- to fivefold density gain over the 40-foot HVAC-cooled cabinets of 2019–2020, which typically topped out around 1.5–2 MWh. The footprint never changed; the cells, thermal architecture and packing efficiency did.
Volume tells the same story. BloombergNEF put global storage additions in 2024 at roughly 69 GW/169 GWh, close to double the prior year, and the consultancy's outlook has cumulative deployments passing 400 GW by 2030. Almost all of that utility-scale capacity arrives on a truck, pre-assembled, pre-tested and pre-wired. The IEA's battery cost tracking shows why: pack-level prices fell to around USD 115/kWh in 2024, a decline of roughly 20% in a single year, and DC blocks from Chinese integrators have traded in the USD 70–100/kWh band on an ex-works basis.
Factory integration is the economic engine behind those numbers. Assembling racks, coolant loops, fire suppression, gas detection and controls in a controlled environment cuts field labour hours by 60–70% compared with stick-built enclosures, and it moves commissioning risk from a muddy site in February to a test bay with a calibrated load bank. For developers, the container is no longer a component — it is the project's atomic unit, and the entire schedule, financing model and O&M contract is built around it.

What is actually inside a 5 MWh container

The dominant architecture is a DC block: LFP cells in 1P52S or 1P104S rack configurations, rack-level DC/DC optimizers or a simple busbar, a liquid cooling loop running glycol at 20–35 °C, a redundant chiller or air-cooled heat exchanger, aerosol and water-based fire suppression, aspirating smoke and combustible gas detection, and an HVAC system sized for the local design ambient. Cells sit at roughly 3.2 V nominal, and a 5 MWh unit typically strings 12–14 racks in parallel.
The alternative is an AC block, which integrates the power conversion system, medium-voltage transformer and switchgear into or alongside the enclosure. AC blocks were the norm five years ago; they are now the minority in utility-scale procurement because separating PCS from the battery lets the owner re-bid inverters, mix vendors and decouple the DC and AC procurement cycles. DC blocks also pack 20–40% more energy into the same footprint.
Auxiliary load is the specification buyers under-read. Chillers, pumps, controls and heating draw 2–5% of nameplate energy in hot climates, and a container rated at 5 MWh at 25 °C may deliver only 4.4 MWh at 45 °C ambient without derating. Ask for the derating curve across 0–50 °C, the coolant setpoint strategy and the condensation-control logic. The numbers you actually get paid on — round-trip efficiency, availability and degradation — all depend on those details. For a deeper look at how LFP chemistry behaves under daily cycling, battery storage technology covers the cell-level trade-offs.

Block prices, project costs and the numbers that matter

Container price is not project cost. A 100 MW/200 MWh installation in the United States will typically see the DC blocks account for 40–50% of total capital expenditure, with the balance absorbed by transformers, medium-voltage collection, civil works, interconnection upgrades, EPC margin and owner's costs. The gap between a quoted USD 90/kWh block and a USD 280/kWh installed cost is entirely explainable — and entirely where budgets go wrong.
| Format | Typical usable energy | Cooling | Typical duty |
|---|---|---|---|
| 10-ft container | 0.5–1.0 MWh | Air | C&I backup, small microgrids |
| 20-ft AC block | 2.5–3.9 MWh | Liquid or air | Distribution-level, behind-the-meter |
| 20-ft DC block | 5.0–6.25 MWh | Liquid | Utility-scale 2–4 hour duty |
| 40-ft DC block | 6.0–8.0 MWh | Liquid | Land-constrained sites |
| Multi-unit block (20 MWh class) | 20 MWh | Liquid | GW-scale clusters, minimal cabling |
Reference points worth keeping. A two-hour turnkey system in the US market has been landing at roughly USD 250–350/kWh in 2024–2025 depending on interconnection scope, while Chinese domestic tenders for 2-hour DC blocks have cleared below USD 100/kWh. Round-trip efficiency for a well-integrated LFP system sits at 85–92% AC-to-AC, and step response to a dispatch signal should be under 200 ms. Degradation guarantees typically promise 70% of beginning-of-life capacity at year 10 or 15, capped by a throughput limit expressed in MWh. Those four figures — price, RTE, response time, warranty floor — do more to determine lifetime revenue than any other part of the datasheet.

Sizing: DC/AC ratio, duration and augmentation

DC/AC ratio is the single most consequential design decision and the least standardized. For a two-hour system, integrators commonly land between 1.1 and 1.3; for four-hour systems, 1.5 to 2.0 is normal. Push the ratio higher and you buy fewer inverter dollars per megawatt-hour, but you clip the top of the charge and discharge curve and lose revenue in the highest-priced intervals. In markets with steep evening peaks — ERCOT, the UK's Balancing Mechanism, Australia's NEM — a lower ratio often pays for itself within three years.
Duration is a moving target because the market it serves keeps changing. A system designed for 2025 frequency response revenue may need to pivot to 2032 capacity adequacy, and the difference is four hours versus two. Designing racks with spare physical space and pre-installed coolant headers adds 3–5% to capital cost and preserves the option to add DC blocks later, which is far cheaper than a new interconnection request.
Augmentation modelling deserves its own line item. LFP systems lose 1.5–3% of usable capacity per year depending on cycle depth and temperature, and a warranty written against nameplate rather than tested capacity quietly transfers that risk to the owner. Build the augmentation schedule into the financial model from day one, and specify whether the vendor supplies replacement modules at a fixed price or an indexed one. For a full view of how these systems are packaged and delivered, the lithium battery product range shows how container formats map to different grid applications.

Fire testing and permitting decide the schedule

UL 9540A is the gate. The test protocol evaluates thermal runaway propagation at cell, module, unit and installation level, and the installation-level result is what an authority having jurisdiction will actually read. A container that passes unit-level testing without propagation can qualify for reduced separation distances under NFPA 855; one that does not may be pushed to 3 metres or more from property lines, which on a constrained site can kill the project outright.
NFPA 855's 2023 edition caps stored energy per fire area at 600 kWh for units that have not undergone large-scale fire testing, requires hazard mitigation analysis for larger installations, and mandates gas detection, emergency shutdown and water supply provisions. The International Fire Code's Chapter 12 mirrors much of this. In Europe, IEC 62933-5-2 and the EN 50604 family govern, and Germany's VDE-AR-N 4110/4120 grid codes add their own ride-through and reactive power requirements.
The practical lesson from projects that stalled in 2023–2024 is that fire review, not equipment supply, is the critical path. Engage the AHJ before the interconnection application, not after. Commission an independent third-party review of the UL 9540A report for the exact model and firmware revision being procured, because certificates do not always travel with product updates. And budget for a water supply and drainage design — many jurisdictions now require it regardless of whether the suppression agent is water-based.

Where containerized systems earn their keep

Solar co-location is the largest single application globally. AC-coupling a 100 MW/200 MWh block to an existing PV plant lets the owner clip the midday generation peak that would otherwise be curtailed, then shift it into the evening. In markets with high curtailment — Chile, California's CAISO, parts of Iberia — the incremental revenue from captured energy can exceed the storage capital cost within six to eight years. DC-coupling captures 2–4% more round-trip efficiency but complicates retrofits and constrains the PV inverter to a shared DC bus.
Standalone grid services remain the highest-value use case per megawatt-hour. The UK's Containment and Balancing Reserve markets, ERCOT's ECRS, and Australia's FCAS markets all pay for sub-second response, and a containerized system's 200 ms step response is precisely what those products require. Revenue stacking — frequency response in the morning, energy arbitrage in the evening, capacity adequacy in the background — is now standard practice and drives the control software requirements.
Behind-the-meter deployments at industrial sites follow a different logic. A 5 MWh container at a cold-storage facility or a foundry is sized against a demand charge, not an energy market, and pays back in 4–6 years in markets with demand charges above USD 15/kW-month. Microgrids and island systems use the same hardware with a higher DC/AC ratio and black-start capability. A portfolio of delivered installations across these categories is documented in the project references if you want to see how container counts scale with site constraints.

Procurement: the questions that separate a good contract from a bad one

Insist on AC-to-AC round-trip efficiency measured at the point of interconnection, at a stated ambient temperature, with auxiliary loads included. Vendor datasheets frequently quote DC-to-DC or cell-level figures that flatter the number by 4–6 percentage points. Ask for the measurement standard — IEC 62933-2-1 is the reference — and for the test report from a unit of the same configuration.
Availability guarantees should be 97–98% with a clear definition of what counts as unavailable, and a liquidated damages mechanism tied to it. Capacity guarantees need both a beginning-of-life test protocol and a degradation curve, plus a defined remediation path: module replacement, software recalibration or financial compensation. Spare parts obligations should cover 15 years, name the specific components held in regional stock, and specify lead times for cells, chillers and PCS modules.
Cybersecurity is now a procurement line item rather than an afterthought. IEC 62443 compliance, signed firmware, role-based access and a documented patch cadence are baseline requirements for any asset connecting to a utility SCADA system, and several US utilities have begun rejecting equipment without them. Finally, ask who owns the energy management system and whether it can dispatch multiple vendor brands — lock-in at the EMS layer is more expensive to escape than lock-in at the container layer. Teams comparing integration partners can review the platform overview or talk to an engineer about specific site constraints.

Commissioning, warranties and the twenty-year question

Commissioning is where a containerized project either proves its numbers or discovers them. The capacity test should be performed at the point of interconnection, at the site's design ambient temperature, with a calibrated metering package. Round-trip efficiency, ramp rate, step response, SOC accuracy and reactive power capability all need to be measured against the contract, not the datasheet. A well-run commissioning programme takes four to six weeks for a 100 MW site and should include a full black-start sequence if the asset is contracted to provide it.
The twenty-year question is really a fifteen-year question, because that is when most LFP containers require their first major augmentation or repowering. Cell degradation, coolant maintenance, chiller replacement at year 8–12, and software obsolescence all land in the same window. Owners who specified modular racks, standardized coolant fittings and open EMS protocols will spend a fraction of what owners of proprietary architectures spend. Those who did not will find that a 5 MWh container with a discontinued chiller is worth considerably less than its nameplate suggests.
Financing has adapted. Standalone storage in the US now qualifies for the 30% investment tax credit under the Inflation Reduction Act, and domestic content and energy community adders can push the effective incentive above 40%. In Europe, capacity market reforms in the UK and Italy have made merchant revenue stacks bankable enough for non-recourse debt. The technology risk premium has compressed accordingly — containers are now financed like infrastructure, not like a science project.
The container itself has become commoditized. What separates projects is everything around it: the fire review completed early, the DC/AC ratio chosen for the market the asset will serve in 2035, the augmentation schedule written into the model, and the efficiency measured at the meter rather than in a lab. Those decisions are made in the twelve months before a purchase order, and they are worth more than any single line on the specification sheet.