Automotive

Chevrolet Bolt: how two latent cell defects passed end-of-line testing

Case file #100·November 21, 2026·6 min read·analysis by Peter Stasko

Case file

  • What happened: Between 2020 and 2021, GM recalled all roughly 142,000 Chevrolet Bolt EVs built worldwide after a series of battery fires, most occurring at or near full charge.
  • Scale: Two recalls expanded repeatedly in 2021 until coverage reached the entire Bolt population; owners were instructed to cap charging at 90 percent and park outdoors, away from structures.
  • Root cause: Two simultaneous cell-manufacturing defects at LG Energy Solution — a torn anode tab and a folded separator — creating conditions for internal short circuits inside validated, end-of-line-tested cells.
  • The bill: About 1.9 billion US dollars in module replacement costs, the bulk of it reimbursed by LG Energy Solution.

Here is an uncomfortable observation from two decades of chasing supplier defects: the scariest failures are not the ones your test bench misses on day one. They are the ones the bench was never asked to find. Every Bolt cell that left LG's line had passed electrical end-of-line testing. Then it sat in a customer's garage, charged to 100 percent, and burned. The gap between what a cell does in ninety seconds of testing and what it does after a thousand full cycles with a latent mechanical flaw inside it — that is where this recall lived.

~142kBolt EVs recalled worldwide
~$1.9bncost of battery module replacement
2latent cell defects, both EOL-invisible

The situation

The Bolt launched in 2016 as GM's first mass-market EV, with LG Energy Solution supplying lithium-ion cells assembled into modules at a GM facility. On paper, a mature, validated supply chain: PPAP-approved parts, a qualified supplier, volumes ramped over years. The paperwork signalled nothing.

Serial cell production is a different animal. Winding and stacking run at speed; anode tabs are thin metal foil that tears; separators are polymer films microns thick that fold. A torn tab and a folded separator in the same cell make a slow-motion failure — a burr or exposed edge that only manifests as a short after repeated cycling at sustained high state of charge. End-of-line electrical screening — capacity, internal resistance, self-discharge — sees a healthy cell. At that moment, it is one.

How it unfolded

The public record shows a pattern every quality professional will recognise. Initial fires triggered a first recall in 2020 with a software remedy: a reduced charge ceiling intended to limit the conditions for thermal runaway. Fires continued. The scope was expanded, then expanded again, until in mid-2021 GM abandoned software containment and recalled the entire population for module replacement. Each expansion was an admission that the previous root-cause analysis had been incomplete — that the mechanism was broader, or the defect population larger, than the prior 8D had concluded.

The interim customer instructions — cap at 90 percent, park outdoors, don't charge overnight unattended — were textbook containment. GM had no corrective fix for months, so it attacked usage conditions instead. Crisis management done correctly, even when it looks humiliating in the press.

Root-cause anatomy

Technically, a two-defect interaction. A torn anode tab leaves a burr or rough edge; a folded separator thins or removes the electrical barrier between electrodes. Either alone might be survivable. Together they build a path for internal shorting that develops over time and worsens at high state of charge — which is exactly why the 90 percent cap was rational containment, not panic. The physics was well understood. The detection was not.

Organisationally, the deeper root cause is a supplier whose serial production drifted into failure modes that neither in-process controls nor outgoing inspection caught — and an OEM quality system with no independent window into it. Two years into serial production, PPAP is a historical document. If nobody audits the actual winding parameters, tooling wear, film handling, the approval basis quietly expires.

An end-of-line test can only find the failures you already imagined; the ones that kill you are designed in the gap between your PFMEA and reality.

Where the quality system failed

Failures stack. Here, at least four disciplines went down together:

  • PFMEA detection ratings. The torn tab and folded separator evidently carried detection scores assuming end-of-line electrical testing would catch them. It could not. A detection rating of 2 or 3 attached to a test blind to the mechanism is not a conservative judgement — it is fiction, and the RPN arithmetic built on it is fiction too.
  • Supplier process control. These were process defects born on LG's lines. Whatever in-process checks existed — vision on tabs, separator alignment — either didn't cover these modes or toleranced them as cosmetic.
  • 8D discipline. The serial recall expansions tell us the initial root-cause work closed on an incomplete mechanism or an underestimated defect population. An 8D that closes before the escape population is bounded isn't closed; it is deferred, with interest.
  • Field data as an audit gate. Early fire reports were the only instrument that could see this defect. The question every OEM should ask: how fast did field evidence trigger a reopened investigation versus a defended closure?

What would have caught it

Nothing cheap. Concretely: sustained destructive physical analysis on serial production — weekly teardown of cells pulled from the line, examined for tab integrity and separator layup, not just electrical performance. In-line vision with 100 percent geometric verification of tab cutting and separator positioning, tied to tooling-wear triggers. VDA 6.3-style process audits at the cell plant on a recurring cadence, with winding and stacking treated as special characteristics — the battery-industry equivalent of safety-critical machining. And PFMEA reviews that reverse the logic: start from what can physically go wrong in the process step, then ask honestly whether any downstream test sees it. Scoring detection optimistically to keep the RPN green is how you buy this recall.

One more thing: cycle-life testing on production-representative samples at full charge, not just benign profiles. The defect manifested at high state of charge for a reason. Test the way customers use the product.

My take

I have no insider knowledge of this case — everything above comes from the public record — but I have lived its smaller cousins. In greenfield automotive plants I have watched a supplier's serial production drift: a changed film supplier here, a worn cutter there, silently invalidating a PPAP everyone still treated as scripture. When I built a QA/QC department from scratch for over 900 people, the lesson I encoded into the system was simple: approval is a moment, surveillance is a habit. We cut external failure costs dramatically at SNOP — not through smarter end-of-line testing, but through QRQC on supplier process data, treating the supplier's own signals as our first alarm and field returns as the final audit I could not delegate. The Bolt case is that philosophy priced at 1.9 billion dollars. Supplier chargebacks recovered most of it, which is the right commercial outcome — but no CFO should mistake reimbursement for a system that works. The only real reimbursement is never needing one.

What this means on your floor

  • PPAP is a snapshot. With no recurring process audit at the supplier, your approval basis expires silently the day their process drifts.
  • Audit every PFMEA detection rating against the physical mechanism, not the test report. Can this test actually see this failure mode, in the state where it occurs?
  • Never let an 8D close until the escape population is bounded. Every recall expansion is an 8D that closed early.
  • Treat field data as your final quality audit — the one test bench the customer runs for free, at full severity.

Two microscopic flaws — one torn tab, one folded film — walked past every gate the quality system had built, because the gates were watching for something else. The Bolt recall is not a story about lithium-ion chemistry. It is a story about the limits of end-of-line thinking, and the 1.9 billion dollar reminder that detection belongs in the supplier's process, in honest risk documents, and in the field data you stop explaining away. The cells passed the test. The system did not.

This case file analyses publicly documented events and reports. I had no involvement in the engagements described; company statements and official findings are matters of public record. The lessons and opinions are my own.

Peter Stasko

Peter Stasko

Corporate operator across automotive and aerospace — Airbus, SNOP and Witte Automotive. Building production AI hands-on since 2016.

Seeing the same failure mode on your floor?

Every case in this library ended as a headline. A one-hour conversation is cheaper. No slides, no pitch — just an experienced pair of eyes on your situation.

Book a conversation