TL;DR · 30-second read
The Short Version
Sungrow, a Chinese maker of giant batteries for power grids, set one of its shipping-container-sized battery units on fire on purpose, with every fire-fighting system switched off.
Inside, temperatures hit about 1,260 degrees Celsius, hot enough to melt copper. The fire burned for seven and a half hours, but it did not spread to the identical units packed right beside it. An independent safety-testing company ran the test.
Why it matters: the computer warehouses behind artificial intelligence need enormous amounts of electricity, and batteries help supply it. Proof that a fire stays in one box could let batteries sit closer together, on less land.
Sungrow said on October 21, 2026, that its PowerTitan 3.0 battery energy storage system (BESS) completed a large-scale fire test under the requirements of UL 9540A: 2026 and CSA/ANSI C800-2025. The test was conducted by UL Solutions, witnessed by Fire Risk Alliance and observed by Jensen Hughes. It used 28.56 MWh of liquid-cooled storage and a 7.4 MW medium-voltage substation laid out like a real utility-scale installation. A thermal runaway event deliberately started in one container did not spread to four adjacent containers or to the substation over 7.5 hours, even though peak internal temperature in the burning container reached 1,259.8°C.
Sungrow describes PowerTitan 3.0 as the industry’s first stacked-cell BESS with an AC-block design to complete such a test, and says all three PowerTitan generations passed their large-scale fire tests on the first attempt.
Executive Summary
The headline result is containment: with the batteries fully charged, the containers placed back-to-back, active fire suppression and thermal management disabled, and the roof vents of the burning unit opened to feed it oxygen, the fire stayed inside one 20-foot container. Cells in neighbouring containers stayed below the temperature at which they would themselves enter thermal runaway, and every container remained structurally intact.
That matters beyond Sungrow’s product line because fire propagation is what drives how far apart battery containers must sit, and separation distance is what drives the land a storage project consumes. For developers trying to add storage at or near power-constrained AI data center campuses, where land, substation capacity and setbacks are all contested, test evidence that containers can be packed tightly without a single fire becoming a site-wide event is a practical input to siting, not just a safety talking point.
The test is one configuration of one product, and fire officials and installation codes, not manufacturers, ultimately decide how close equipment can sit. But it shows the kind of system-level evidence that higher-density battery platforms will increasingly need to bring to the permitting table.
A Test Designed to Remove the Safety Net
Thermal runaway is the self-sustaining chain reaction in which an overheating lithium-ion cell releases heat and flammable gas, triggering its neighbours. Most battery systems rely on layers of active protection, such as cooling systems, gas detection and fire suppression, to interrupt that chain. Sungrow’s test switched those layers off. The system was at 100% state of charge, meaning maximum stored energy and therefore maximum fuel. The initiating container sat back-to-back with its neighbour, the substation was included in the layout, and the rooftop pressure-relief panels were opened fully to maximise airflow.
Under those conditions, the only thing standing between one burning container and its neighbours was passive design: the enclosure, internal barriers, cell arrangement and spacing. That is the right question to test, because active systems can fail or be overwhelmed in a real incident. UL 9540A itself is a test method rather than a simple badge: it generates data on how a battery fire behaves and whether it spreads, which fire officials and engineers then use to judge installations. A result showing no propagation under deliberately adverse conditions is the kind of data those reviewers look for.
Why Container Spacing Becomes an AI Campus Variable
The mechanism linking this test to AI infrastructure runs through land and layout. Separation distances between battery containers, and between batteries and electrical equipment such as substations, exist to stop one fire from cascading. Installation standards such as NFPA 855 and the local authorities who enforce them weigh fire test data when deciding what spacing is acceptable. The more conservative the assumption about propagation, the more ground a storage project consumes for the same capacity.
AI data center campuses are where that trade-off bites. Large AI facilities are often limited by how much power the grid can deliver and how fast, and batteries are one tool operators and utilities use to manage peaks, firm up supply and provide backup. Space inside a campus fence line or next to its substation is scarce, and every metre of setback competes with buildings, cooling plant and electrical gear. Sungrow’s test showed 28.56 MWh of storage arranged with containers back-to-back and a 7.4 MW substation in the same block, with neighbouring cells staying below their thermal runaway threshold while the initiating container peaked at 1,259.8°C. That is direct evidence on exactly the spacing question siting engineers face.
Who is affected: data center developers evaluating on-site or adjacent storage, utilities placing batteries near large new loads, insurers pricing those assets, and fire departments that must plan for incidents next to critical facilities. None of them can treat one manufacturer’s test as a blanket licence for tighter layouts, but the bar it illustrates, system-level evidence including the substation, is the bar dense, co-located storage will be judged against.
Denser Boxes Raise the Stakes
PowerTitan 3.0 combines two design choices that make a fire test more consequential. Stacked cells pack more energy into a given volume, and the AC-block architecture places the battery and its DC/AC power converter, the equipment that turns stored direct current into grid-ready alternating current, inside a single 20-foot container. More energy and more electronics per box mean more potential fuel in each enclosure and fewer, more consequential units per site.
Sungrow’s claim to be the first stacked-cell, AC-block BESS to complete such a test is drawn around a narrow product category, as many industry firsts are. The more meaningful point is that energy density and fire containment pull in opposite directions, and density gains are only bankable for dense sites if containment keeps pace. Sungrow’s statement that PowerTitan 1.0, 2.0 and 3.0 all passed on the first attempt suggests a deliberate design discipline, though the company has not published the earlier results alongside this one for comparison.
What One Burn Can and Cannot Establish
Several features add weight to the result. The test was run by UL Solutions, with two fire-engineering firms witnessing or observing, and it lasted 7.5 hours under continuous monitoring. Containers remained structurally intact, and Sungrow says the initiating container could still be lifted as a single unit afterwards, which bears on cleanup and site recovery. Smoke and fire effluent were captured and treated rather than released.
Other claims are harder to weigh. Sungrow calls its simulation platform the world’s largest of its kind and says simulated fire behaviour closely matched the physical test, but it has not published the model outputs or the comparison metrics. A single test also cannot capture every real-world variable: different spacing, weather, terrain, adjacent buildings, or the gas accumulation that can occur when vents do not open. And the purification system used to treat smoke in the test would not exist at a real site, so the effluent question for neighbours of a dense campus remains a planning issue in its own right.
Background
Sungrow, based in Hefei, China, describes itself as a globally leading supplier of solar (PV) inverters and energy storage systems. Its PowerTitan line serves utility-scale storage projects, the large battery installations that grid operators and developers use to store power and release it when needed. PowerTitan 3.0 is the third generation, moving to stacked cells for higher energy density and integrating power conversion into each container.
As battery storage has scaled up, fire safety has become one of the main factors shaping permitting, insurance and site design. Large-scale fire testing, in which a full system is burned under controlled conditions to see whether fire spreads between units, has become a key way manufacturers demonstrate that a single failure will not become a site-wide event. That question carries extra weight wherever storage is sited near critical facilities such as substations and data centers. Source: Three Generations, All First-Time Passes: PowerTitan 3.0 Validates System-Level Safety Under Extreme Large-Scale Fire Test, Sungrow’s announcement of PowerTitan 3.0’s large-scale fire test results under UL 9540A: 2026 and CSA/ANSI C800-2025.Sources