Why classification matters more than product selection
A residential battery specification decision made today is a decision about which of three overlapping frameworks a product sits inside: its cell chemistry, its certification stack, and its battery-management-system (BMS) generation. Each framework changed substantially over the 2010-2025 period, and each still has open questions running into 2026-2030. Evaluating a product against yesterday's version of any one framework understates today's risk, or overstates today's differentiation. This report sets out all three frameworks as they stand, with the evidence for how each one got there.
Chemistry classification
Three cathode chemistries account for essentially the entire residential lithium-ion category across the period. Classification matters because chemistry constrains everything downstream of it: cycle life, thermal behaviour, and the safety certification burden a product needs to carry.
| Chemistry | Cell energy density | Cycle life (80% DoD) | Thermal behaviour | 2025 residential position |
|---|---|---|---|---|
| NCA (nickel cobalt aluminium) | ~260 Wh/kg | 1,000-2,000 cycles | Higher-voltage cathode; more sensitive to overcharge/overheat abuse | Legacy: original Tesla Powerwall chemistry, since superseded in Tesla's own product line |
| NMC (nickel manganese cobalt) | 300+ Wh/kg | 1,000-2,000 cycles | Shares NCA's thermal management demands; can release oxygen under thermal runaway | Declining: dominant 2016-2019, BNEF projects ~1% of stationary storage share by 2030 |
| LFP (lithium iron phosphate) | 95-205 Wh/kg (CATL, 2024) | 2,500-9,000+ cycles; next-gen toward 15,000 | Higher thermal runaway onset temperature; no oxygen release on failure | Dominant: BNEF projects LFP will hold approximately 95% of the stationary storage market from 2026 |
The classification lesson: energy density, the metric NCA and NMC compete on, is the wrong optimisation target for a device that sits stationary in a garage for ten years and does not need to minimise weight or volume the way a vehicle battery does. LFP's cost trajectory reinforces the same conclusion at the procurement level: BNEF's 2025 survey put average LFP pack prices at US$81/kWh against NMC's US$128/kWh, a 37% discount that has widened from a 32% cell-level discount in 2023.
The safety certification framework
Every residential battery product in a major market today should be evaluated against a four-layer certification stack, each layer addressing thermal runaway from a different vantage point. The stack took until 2020 to fully assemble, five years after the category's first mass-market product shipped, which is itself a specification-relevant fact: a product's approval date, not just its current certification status, indicates which generation of the framework it was originally designed against.
| Standard | Layer | What it certifies | In force from |
|---|---|---|---|
| IEC 62619 | Cell / module | Overcharge, overdischarge, short-circuit, crush, thermal abuse and vibration testing at the cell level | 2017 (2nd ed. 2022) |
| UL 9540A | Fire behaviour | Large-scale fire propagation test method: whether a single-cell failure propagates to adjacent cells, modules and the structure | Developed 2014-15 |
| UL 9540 | Certified product | System-level product certification; de facto US market-entry requirement | 2016 (2nd ed. 2020, 3rd ed. 2023) |
| NFPA 855 | Installation | Spacing, siting, fire-service access and ventilation rules for stationary ESS installations | 2020 (2nd ed. 2023) |
A specification note worth carrying forward: UL 9540's Third Edition (2023) formally distinguishes AC-coupled from DC-coupled ESS and adds functional safety requirements: a product certified only against the Second Edition (2020) has not been evaluated against that distinction. For any large-scale or multi-site deployment, requesting the certification edition number, not just confirmation of "UL 9540 certified," is a materially different level of diligence.
Case study: fleet risk when a manufacturing defect outlives the recall
The LG Energy Solution recall is the category's most consequential safety event, and it is instructive less for what happened than for how long it took to resolve. NMC cells manufactured between 21 January 2016 and 30 June 2019 developed a lithium-plating defect causing internal short-circuit risk. The recall proceeded in waves across Australia (Aug 2020, expanded Dec 2020 and Jun 2021, Software Fix Recall Aug 2022), the United States (CPSC notices Dec 2020, Aug 2021, Oct 2022), the United Kingdom (Mar 2021) and New Zealand (Jun 2021, Sep 2022). By September 2023, LG had reported 73 thermal-runaway incidents globally: 28 in Germany, 13 in Australia, 10 in the United States.
Two facts matter for fleet-risk assessment specifically. First, the ACCC accepted a court-enforceable undertaking from LG in May 2024, five years after the affected manufacturing window closed, and then accepted a variation to that undertaking in November 2025, expediting replacement and refund access; as of December 2025 LG's own tracker recorded 931 unrectified units under the Replacement Recall and 1,358 under the Diagnostic Software Recall. Second, a February 2025 explosion in Schönberg, Germany involved a 2019-manufactured unit outside the original recall's defined scope, after which LG preventively capped charge levels at 75% for affected systems. Legacy fleet risk from a chemistry-and-manufacturing-window defect does not resolve on the recall's original timeline; it persists as an operational liability for years afterward, with remediation activity still active in the most recent reporting.
The mitigating context: separate German research (RWTH Aachen) put the annual fire risk for home battery storage systems at approximately 0.0049%, around 37.5 incidents in 2023 against roughly 770,000 installed units. The recall reflects a manufacturing-window quality failure, not a base-rate risk indictment of the chemistry or the category.
BMS generation as a technology-maturity signal
Battery-management-system sophistication is a fourth classification axis, increasingly independent of chemistry. Four generations are identifiable in the engineering literature and map cleanly onto the residential category's own history.
| Generation | Era | Capability |
|---|---|---|
| Gen 1 | ~2014-2016 | Passive cell protection: voltage and temperature cut-offs, limited software intelligence |
| Gen 2 | ~2016-2019 | Digitalised state-of-charge estimation, active balancing, mobile app connectivity |
| Gen 3 | ~2019-2022 | Cloud connectivity, over-the-air update capability, state-of-health reporting, early VPP integration |
| Gen 4 | ~2022-2026 | AI/ML-driven multi-modal sensing, predictive thermal-runaway modelling (8-13 minutes ahead of onset), VPP dispatch bundled as a software layer |
The LG recall's over-the-air mitigation, capping state of charge to 90% and later 75% across thousands of fielded units in multiple countries, was the first large-scale proof that a Gen 3 BMS could function as a remotely deployable safety intervention rather than a static protection circuit. It was also proof of that approach's limit: the ACCC reported a subsequent fire involving a unit that had already received the software update, confirming an OTA cap mitigates but does not eliminate risk when the underlying fault sits in cell manufacturing rather than in software logic. For specification purposes, BMS generation should now be treated as materially as chemistry: by 2025, roughly 40% of new German residential buyers were selecting products specifically for VPP compatibility, a software capability rather than a cell-chemistry one.
What the classification implies for 2026-2030 planning
Three planning-relevant conclusions follow from the record. First, chemistry risk for new deployments is now concentrated at the margins: BNEF projects LFP holding roughly 95% of stationary storage from 2026, so a specification defaulting to LFP is defaulting to the market consensus, not taking a position. Second, certification-edition currency is a real diligence variable, not a formality: the four-layer stack is mature but still iterating (UL 9540 Third Edition's AC/DC ESS distinction, 2023), and legacy NMC-chemistry fleet risk is explicitly expected to continue generating incident and remediation activity through at least 2026-2027. Third, BMS generation is becoming the primary basis for technology differentiation between otherwise similar LFP products, and evaluating it requires asking about software and grid-interactivity capability specifically, not inferring it from chemistry or capacity specifications alone.
Two emerging chemistries sit outside this framework and are worth tracking rather than specifying against today. Sodium-ion cells (CATL's first generation at 160 Wh/kg, targeting 200 Wh/kg in generation two, with reported small-scale cell pricing near US$77/kWh) could undercut current LFP cell pricing if scaled to volume production, but lower energy density limits displacement of LFP in space-constrained residential applications before the early 2030s on current evidence. Solid-state technology remains an EV-first development track: Toyota's confirmed 2028 EV target and QuantumScape's "before 2030" target are both vehicle commitments, and no major manufacturer has publicly committed to a stationary-residential solid-state timeline. The historical EV-to-stationary technology transfer pattern (NCA and NMC both arrived in residential storage after establishing themselves in vehicles) suggests any residential solid-state crossover would lag EV commercialisation by several years.
Full year-by-year record, including the complete country-level 2025 market data, sits alongside review.solar's companion history record and Signal on this research. The market this chemistry stack now serves is sized in the 2025 behind-the-meter storage data, and the grid-scale fleet's revenue reality in the TBx Capture Index.