Between 1 January 2025 and 4 October 2026, eMOBINO identified 39 recall campaigns for fire risk in the high-voltage battery of electric and plug-in hybrid vehicles. Together they cover more than 1.3 million vehicles. We counted each defect once, however many countries filed it, and we read the regulators' records and the makers' own defect reports before the press.
The list is a floor, not a census. More campaigns exist than we could identify, and for most of them only the vehicle count of one or two countries is known. Even so, the pattern in the record is clear enough to act on, and it is more useful to a battery pack engineer than any single headline about one model.
Most of the recalls start inside the cell
We sorted every campaign by the defect wording of the record. In 21 of the 39 campaigns the record names a fault inside the cell or module. In five more the battery overheats and the record gives no cause. Four campaigns name the battery management software as the defect, six are pack assembly faults and three are sealing, structure or cooling faults.
Where the reports explain the cell fault, they describe a production step that went wrong:
- Separator damage, where the cathode puts local stress on the separator layer
- Misaligned or shifted electrodes that lead to an internal short circuit after many cycles
- A torn cathode edge that folds over inside the cell
- Lithium plating after an equipment setting error left the anode and cathode capacity out of balance
None of the records blames the cell chemistry or the cell design as such. These are production faults that escaped the tests in place. A cell with such a fault has a normal capacity and a normal resistance when it is new, within what a production test can resolve. It passes formation, grading and the pack maker's incoming inspection, and it becomes an event after months or years of use.
Most of these faults can be found in the plant with the right inspection. After the recalls, one cell supplier added camera monitoring of electrode stacking and another tightened the limits of its automated inspection. Both steps could have been in place from the start of production.
The same cell fault reaches several vehicle makers
The makers' reports to the United States regulator name the cell supplier. Read side by side, they show that a cell fault is rarely the problem of one vehicle maker. One separator fault appears in the reports of three vehicle groups in almost the same words. One electrode alignment fault appears at four vehicle makers.
One maker's report says that its investigation began when it learned of battery failures in another maker's vehicles with the same battery. It decided on a recall ten days later. A vehicle maker that buys a cell also buys the cell maker's process history, together with every other customer of that line. Whether it learns of a fault in days or in months depends on the supply contract, not on the physics.
Plug-in hybrids hold more than half of the recalled vehicles
Plug-in hybrids are 28% of the campaigns and 57% of the recalled vehicles. This is a share of recalled vehicles, not a failure rate. We have no count of plug-in hybrids and electric cars on the road to divide by, and two very large campaigns carry most of the share.
The record cannot prove that plug-in hybrid packs fail more often. Their duty cycle does offer an explanation. A plug-in hybrid pack is small, so the owner charges it to full almost every day. The car then stands at full charge for hours, often overnight in a garage. A long-range electric car is rarely charged to full and rarely stays there.
In our view the plug-in hybrid is one of the hardest duty cycles for a battery pack in a passenger car, and it has often been validated as if it were the easiest.
A full battery is the condition that turns a defect into a fire
In 18 of the 39 campaigns the owner is told to limit or stop charging until the repair. The limits in the records are 70%, 75%, 80% and 90%. Seven campaigns tell owners not to charge at all.
Several reports are direct about the reason. A module short-circuits when the vehicle is fully charged and parked. Cells short-circuit when charged to full or nearly full capacity. The fires themselves are often not charging fires. One report describes five thermal events: two during fast charging, two with the vehicle parked and not charging, and one while driving. Another describes two vehicles that caught fire switched off, parked outside and not connected to a charger.
A charge limit is the only lever a maker has across a whole fleet within days. It needs no parts and no workshop. It works only if the vehicle enforces it and not the owner.
Software is the first remedy, and it does not always hold
For 87% of the recalled vehicles, software is the remedy or its first step. The update looks for a cell that loses charge or voltage faster than its neighbours, and the battery is replaced only where the software flags it. This is a reasonable first answer. There are not enough packs or workshop hours to replace several hundred thousand batteries at once.
In four campaigns the maker's own report states that an earlier software remedy did not hold. Fires or cell venting occurred in vehicles that already had the update, and the maker filed a new recall. In one case the final remedy is a new battery for every vehicle.
Monitoring is a race between detection and failure. It works for a fault that grows over weeks. It does not work for a fault that goes from normal to short circuit within one charge, or when the battery management system is asleep while the vehicle is parked.
Traceability decides the size and the speed of a recall
Each report to the United States regulator gives the maker's estimate of the share of recalled vehicles that have the defect. In the small campaigns it is 100%: the maker and the supplier can name the cells, and the recall covers 14 vehicles, or 51. In most of the large campaigns it is 1%, and in two cases 5%. The largest campaign recalls about 320,000 vehicles for a fault estimated in about 16,000.
The same split shows in time. From the first fire known to the maker to the recall decision took between ten days and 33 months in the eleven campaigns where we could measure it. The fast campaigns are those where a teardown found a physical defect and the supplier could tie it to machines and dates. The slow ones are those where the root cause stayed open.
What a pack programme should change
We read the record as four lines of defence, and each is decided years before the first field event:
- Prevent. Audit the cell process, and buy the data together with the cell: a serial number on every cell, the inspection result per cell, and the link from the machine to the vehicle.
- Detect. Design the battery management system to find a failing cell while the vehicle is parked, and prove it on cells with known faults.
- Contain. Keep margin at the top of the charge window, make the charge ceiling a parameter that can be changed in the field, and design the pack so that one cell does not take the vehicle.
- Repair. Keep the module replaceable in a workshop, and hold a field plan that reaches every market on the first day.
eMOBINO's advice to anyone building a battery pack follows from these records. Do not test only what the regulation asks: write your own test cases, especially for abuse and misuse, and repeat them with cells from different batches. Ask the cell supplier for the inspection data of every cell, not only a certificate for the lot. Make the battery management system check the cells while the vehicle is parked, not only while it drives or charges. Keep some margin below 100% charge. None of these measures is expensive against a single recall campaign.
If these points are not decided in the design phase, a recall decides them later.
