Case study · Electronics manufacturing · May 2026 · 8 min read

The Note 7:what the hardest recall in electronics taught manufacturing

Samsung recalled the Galaxy Note 7 twice in 2016, once for its original battery and once for the replacement from a second supplier. Every cell involved had passed final test. The lessons Samsung drew from it changed battery manufacturing for the whole industry.

Hands repairing a consumer electronics device with circuit boards exposed

Case study · Electronics manufacturing

In August 2016, Samsung launched the Galaxy Note 7 to strong reviews. Within weeks, a small number of units were overheating and catching fire. On 2 September, Samsung recalled roughly 2.5 million phones and traced the fault to batteries from one supplier. Replacement phones shipped with cells from a second supplier. When some of those also failed, Samsung made the hard call and discontinued the product entirely in October. Aviation regulators banned the phone from aircraft by name. The direct cost was estimated at more than five billion dollars.

What makes the Note 7 a manufacturing story rather than a design story is what the investigation found. In January 2017, after testing 200,000 phones and 30,000 batteries alongside three independent labs, Samsung published the root causes in full, an unusually transparent step for a failure of this scale. There were two different defects, from two different factories, with one thing in common: both were process variances that were invisible at final test.

02Two suppliers, two process escapes

The first cell failed because the pouch casing left too little room at the corners. The negative electrode could deflect in the upper corner of the cell, and in rare cases the electrodes eventually met. This was a dimensional variance, the kind of thing that drifts gradually in tooling and assembly fixtures, batch by batch.

The replacement cell, ramped quickly to meet the demand the recall created, failed differently: welding burrs on the positive tab were occasionally tall enough to pierce the separator, and some cells were missing insulation tape. Weld quality and a missing component, two classic line-level escapes, made more likely by a rapid increase in volume.

Every one of those cells passed end-of-line testing. The defect rate was on the order of a few failures per million, far below what sampling-based quality control can catch, and exactly the kind of tail that only appears when millions of units meet the real world. The verdicts arrived weeks later, in the field.

Macro view of an electronic circuit board — The failures were physical and tiny: a deflected electrode corner, a weld burr taller than its neighbours.

03The lesson the industry took

Samsung's response is the most instructive part, and it deserves credit. The 8-point battery safety check it introduced afterwards is, in essence, upstream process surveillance: X-ray inspection of every cell's internal geometry, disassembly audits, charge and discharge cycling, leak detection by volatile-compound sensing, and accelerated stress tests. Measurement moved from the end of the line into the line, and much of the industry followed.

That is the general lesson, and it extends far beyond batteries. When a defect is born in a process, in a fixture that drifted or a weld parameter that crept, final test is the most expensive and least informative place to look for it. The signal existed earlier: in the dimensional data, in the weld telemetry, in the vision feed. What was missing was something watching those signals continuously and connecting them to outcomes.

This is precisely the problem class Vertex-edge is built for, and it is why our EV battery problem brief reads the way it does: predict the end-of-line outcome from the upstream signature, flag the drifting tool before it ships risk at scale, and keep a signed record of which batch saw which process state. The Note 7 programme did not lack testing. It lacked a continuous memory of its own manufacturing, and the entire industry learned from what that cost.

Keep reading

Physical AI

Prometheus: Bezos's $18B bet on the physical world — and what it means for Europe

Jeff Bezos has quietly raised around $18 billion for Prometheus — an AI built to design and manufacture physical things, from jet engines to drug compounds. Almost no one in Europe is talking about it. Here's why his bet, Siemens' answer with NVIDIA, and one very blunt line from Siemens' own CEO add up to a fight for the factory floor.

Read

AI development

GTC Taipei: the factory is the product now

At GTC Taipei, Jensen Huang stood in front of Taiwan's manufacturing leaders and described the AI factory as the next industrial asset class. For anyone who works on a plant floor, the most interesting message was in the supporting slides.

Read

Recognise one of these failure patterns from your own floor?