Samsung SDI Brings an Indoor Fire Test Claim to RE+

Samsung SDI gave its U8A3 prismatic uninterruptible power supply battery a North American debut at RE+ in Las Vegas on November 17, carrying a claim about fire performance that the company has not yet backed with a public certification record. According to The Elec, which reported the exhibit plans on November 16, Samsung SDI states that the U8A3 is the first product in the world to meet every criterion of UL Solutions’ indoor large-scale fire test, an evaluation that ignites a battery system at installation scale rather than at the cell or module level.

The product itself is narrower than the headline suggests. The U8A3 uses a lithium manganese oxide cathode rather than lithium iron phosphate, and it is built as a UPS rack aimed at the load swings that artificial intelligence data centers impose on backup power. Samsung SDI claims 33 percent greater space efficiency than its prior generation. The company is separately showing SBB 2.0, a 20-foot container made in the United States on LFP cells, following the entry of its Indiana prismatic LFP line into mass production. The Seoul Economic Daily reported on November 10 that the UPS battery took a third consecutive CES Innovation Award.

The verification gap. The fire test claim rests on the company’s own announcement. A buyer evaluating indoor siting works from a UL 9540 listing and a UL 9540A test report, both of which are documents an authority having jurisdiction can request and read. A press statement reporting that all evaluation criteria were met is a different artifact. Until the result appears in UL’s directory, the correct description of the U8A3 position is a vendor claim awaiting confirmation, not an indoor certification.

A second large-scale test at the same show. Ampace is presenting fire safety results for its PU100 system at RE+ across November 17 to 19, alongside data center, commercial and residential products. The underlying large-scale test was presented earlier by Dr. Drew Feng at Data Centre World Asia 2026. Ampace describes the test layout as representative of dense indoor installations and reports no propagation to adjacent modules or cabinets and no sprinkler activation, with a 24-hour post-test observation period that ran past the 12-hour re-ignition window. The company states the test met the criteria of the sixth-edition UL 9540A large-scale fire test with margin on several measures.

The contrast between the two disclosures is instructive. Ampace put observable outcomes on the record: what did not propagate, what did not activate, how long the system was watched afterward. Samsung SDI’s statement reports a conclusion. Both are vendor-supplied, and neither has published the section of a test report that a code official actually reads.

The configuration question. Large-scale fire testing produces data that an authority having jurisdiction weighs against a specific proposed layout, which means the value of a result depends on how closely the tested arrangement resembles the room in question. Ceiling height and side-by-side clearance are the two variables that move the answer most. A test run at a three-metre ceiling with zero clearance between units describes a different installation than a typical mechanical room with higher clearance and a different suppression design. A result generated on a UPS rack arrangement supports a UPS rack arrangement. It travels further only as far as the configuration section of the report permits, and that section is what neither announcement reproduces.

The deadline. The sixth edition of UL 9540A becomes mandatory in January 2027, which is roughly six weeks after the RE+ show floor closes. The interval explains the timing of both announcements. Vendors can run large-scale tests to the new edition now and publish whichever portion of the result they choose, and the first mover sets the vocabulary that later entrants have to argue against. Installation-level fire testing has become a marketing asset slightly ahead of becoming a compliance requirement, and that ordering rewards whoever publishes first and discloses least.

A third set of listings. Moment Energy, a Canadian repurposer, opened orders on November 17 for Nova, a 200 kW and 500 kWh stationary block assembled from retired electric vehicle packs at its British Columbia factory. The unit carries UL 9540 and UL 1973 listings with UL 9540A thermal runaway testing, and the factory holds UL 1974 for sorting and repurposing used EV batteries. The system is liquid cooled, measures 2.94 metres by 1.80 metres, is rated from minus 25 to 50 degrees Celsius, and is scheduled for first deliveries in the first quarter of 2027. Moment has not stated an indoor rating. Three vendors in one week therefore present three different things under the general banner of fire testing: a claimed indoor large-scale pass, a disclosed large-scale test summary, and a listing stack with module-level thermal runaway data.

Why indoor performance is a building question. A Latitude Media Open Circuit episode published November 17 described distributed data center developers approaching large downtown and suburban commercial landlords and offering payment for spare electrical capacity, often without the serving utility being aware of the arrangement. Jim Kapsis named fire codes, structural weight, cooling and fiber as the barriers to converting office buildings to compute use. Caroline Golin named storage as the answer to the power quality problem that spiky inference loads create. Fire code therefore sits at the top of the list of constraints in exactly the conversation where an onsite battery is proposed as the fix, which raises the commercial stakes on whether a given system can be sited indoors and on what evidence.

What is on the record as of November 17 is a company statement from Samsung SDI, a test summary from Ampace that names observable outcomes without specifying the room, and a mandate date in January 2027 that will convert the sixth edition from an optional credential into a requirement. The configuration pages, which determine how far any of these results travel, remain unpublished by all parties.


Sources