What Actually Causes a Lithium Battery to Catch Fire (Hint: It's Rarely the Cell Alone)

"It Was a Bad Cell." Sure. Now What?

Open any news story about a battery storage fire and you'll find the same comment within the first ten replies: "Bad cell."

That explanation is true and useless at the same time. Actually, it's worse than useless — it's the reason the same failures keep happening.

I'm a quality and brand compliance manager at an energy storage integrator. I review cells, packs, and system documentation before anything goes to a customer — roughly 200 vendor submissions a year. In Q1 2023, we rejected about 15% of a 200Ah 48V LFP cell batch. Not because the cells were defective. Because they were inconsistent. The vendor's response: "That's within industry standard."

We rejected them anyway, and the vendor redid the batch at their own cost. That decision changed how I look at the entire question of battery fires.

So let me reframe the question that everyone keeps asking wrong: it isn't why do lithium cells fail. It's why does one failed cell turn into a building on fire. Those have completely different answers.

The Real Problem Isn't Cell Quality. It's Consistency.

I've been doing this long enough to be honest: most thermal runaway incidents I've traced weren't caused by a single bad cell. They were caused by cells that were slightly different from each other.

Here's the mechanism, stripped of jargon. Two cells come off the same line. Same nominal capacity. Different internal resistance by maybe 3%. You weld them into the same pack. They charge and discharge at slightly different rates. The higher-resistance cell runs a little hotter. Hotter means it ages faster. Faster aging means higher resistance. Higher resistance means more heat. It's a spiral, and by the time anyone notices, that cell is the weakest link in a chain holding real energy.

Now here's the part that surprised me when I started auditing factories: this is a process control problem, not an inspection problem. You cannot catch it by testing finished cells. You catch it by controlling the winding tension, the electrode coating thickness, the moisture content in the dry room. By the time it shows up in a finished cell, the defect was baked in three weeks earlier.

From the outside, it looks like a battery fire is a quality escape. The reality is that it's usually a manufacturing discipline failure that nobody saw, because the cells passed every test they were given.

The Second Thing Nobody Designs For: Propagation

This one took me years to fully internalize, and I think it's the single most important point in this entire article.

A cell failing releases a limited amount of energy — a few hundred watt-hours in a 200Ah 48V module. That's a cell-level event. Bad, contained, survivable.

The neighboring cell igniting, and the one next to that, and the one next to that — that's a system-level event. That's a fire that needs the fire department.

The industry has standards specifically for this. In the US, UL 9540A evaluates thermal runaway fire propagation — the whole point is to demonstrate that a failure is contained rather than cascading. NFPA 855 covers how stationary storage gets installed and spaced. In Europe and Asia, IEC 62619 covers industrial cell and pack safety. These exist for a reason, and they are not the same thing as UL 1973, which certifies cell-level safety. I've seen spec sheets that conflate them, sometimes deliberately.

Here's my frustration with the marketing layer: plenty of vendors write "UL 9540A tested" on a datasheet without stating the conditions. Was the cell spacing 10mm or 25mm? Was there intumescent barrier material between modules? Did the test pass, or did it fail at the module level and pass at the unit level? All of those details matter, and most of them never make it into the sales deck.

I'd argue that the physical design — cell spacing, inter-cell thermal barriers, venting paths, particle filtration — determines the fire outcome far more than the brand on the cell does. A well-designed enclosure can contain a mediocre cell. A badly-designed enclosure will cheerfully propagate a world-class cell into the same catastrophe.

What This Actually Costs

Nobody wants to talk about the cost side, because it's unflattering. But it's where the case for getting this right actually lives.

Rework. In 2024, we had to retrofit 132 modules on a 2.4 MWh commercial project because we hadn't written inter-cell spacing requirements into the original 2022 contract. The vendor's spacing was "within their standard." It didn't meet the propagation requirements our customer's insurer wanted. Cost to us: just over $46,000 in rework, plus 11 weeks of schedule slip. The customer's project-finance diligence review is what caught it, not our own audit.

Insurance. This is the fastest-moving consequence, and I don't think most buyers have caught up. Commercial and utility-scale storage now gets underwritten with three standard questions: what cell manufacturer, has the system been UL 9540A tested, and at what spacing. Can't answer? Premiums go up 20-40% or coverage gets declined outright. That's before you've even put a system in the ground.

Permitting. Authorities having jurisdiction have gotten much sharper about this since 2023. A project with a clean UL 9540A report and proper setbacks moves through approval in weeks. One without a report sits in review. Sometimes for months.

And the reputational cost — a fire makes local news, but a recurring problem makes industry news. That one doesn't come back.

Okay. So What Do You Actually Specify?

I'm going to keep this part short, because if you've read this far, you already understand the problem. The solution is much less interesting than the diagnosis.

Write consistency windows into the contract, not just nominal specs. Not "200Ah" — but "200Ah ±2%, internal resistance within a 5% band across the batch, DCIR measured at 50% SOC and 25°C." If a vendor won't agree to a window, that tells you something.

Ask for the UL 9540A report itself. Not the summary. Not the marketing one-pager. The actual report with test conditions, cell spacing, and pass/fail at each tier — cell, module, unit, installation.

Don't compare apples to grapefruits on specs. A 200Ah 48V cell is nominally around 9.6 kWh. A Tesla Powerwall 3 is spec'd at 13.5 kWh usable capacity with 11.5 kW continuous output — but that's a complete AC-coupled system, not a raw cell. The number of times I've seen these two figures put side by side in a comparison slide is genuinely astonishing. They measure different things.

Look at who controls the process, not just the product. This is why the factory matters more than the label. Public reporting has confirmed EVE Energy ships cells into Tesla's supply chain — and what that really buys you isn't the brand name, it's evidence that a third party with extremely low tolerance for process drift has already stress-tested the manufacturing system. Similarly, EVE Energy's Indonesia battery plant, with a publicly stated timeline landing in 2025–2026, matters for procurement teams because regional manufacturing shortens supply chains and sidesteps some of the tariff and export-control friction that's been reshaping cell sourcing since 2023. More qualified sources, more leverage, fewer excuses for long lead times.

None of this is exotic. It's specification discipline, and process discipline.

Which brings me back to where I started. "Bad cell" is a comfortable answer because it lets everyone off the hook. The uncomfortable answer is that lithium batteries catch fire when manufacturing consistency drifts, when system design ignores propagation, and when buyers don't specify either one. You can't inspect your way out of a problem you never wrote into the contract.