A battery is the only component in a hardware product where a manufacturing defect can kill someone. That single fact reorganises everything about sourcing it. The cell price matters, the cycle life matters, and neither determines whether a field failure becomes a recall, a lawsuit or a criminal investigation. What determines that is the chain of warranties, test evidence and insurance that sits behind the pack, and most foreign companies build that chain after they have chosen the supplier rather than before.
The commercial situation is familiar. The product needs a pack, the factory has a pack, the pack is cheaper than the alternatives, and the sample works. Eighteen months later a unit vents in a customer’s hand, the investigation finds that the BMS lacked a protection the specification never required, the assembler’s warranty expired at twelve months, the cell maker warranted only to the assembler, and the assembler has assets of a few million renminbi and a lease. The cell price was never the problem.
The most important conclusion is that battery sourcing is a contract design problem before it is a procurement problem. China produces more than three-quarters of the world’s lithium-ion battery supply, so the question for most hardware companies is rarely whether to source there. It is which part of the chain to contract with, and on what terms. Decide who bears which risk, for how long, with what evidence and with whose money, and the supplier selection becomes a matter of finding a party who can carry the allocation. Decide the supplier first and the allocation becomes whatever that supplier will accept, which for a small assembler is very little.
What makes a battery different from every other component
Four features separate a battery from a connector, an enclosure or a display module, and each has a legal consequence.
The first is classification. Lithium cells and batteries are dangerous goods for transport purposes, in Class 9, and the rules that govern their movement are technical instructions rather than commercial preferences. A shipment that is misdeclared is not merely non-compliant; it is a hazard to the aircraft or vessel carrying it, and the liability for the declaration sits with the shipper named on the transport document, which may be you rather than the factory.
The second is that failure is a safety event rather than a quality event. A defective enclosure is a return. A defective cell is a fire, and a fire in a consumer product produces personal injury claims, regulatory reporting obligations, a recall and, in serious cases, a criminal investigation into who knew what. The damages are not bounded by the order value, and the insurance that covers them has to exist before the event.
The third is that regulatory permission attaches to a configuration, not to a product family. A certification for a pack with one cell, one BMS and one enclosure does not cover the same pack with a substituted cell. Changing a supplier to protect margin can silently void the approval under which the product is sold, and the invalidity is discovered at customs or at the customer’s audit rather than at the change.
The fourth is that the supply chain is three layers deep and the warranties do not line up. The cell maker warrants to the assembler. The assembler warrants to you. You warrant to your customer. Each layer has a different period, a different remedy and a different cap, and the gaps between them are where the money is lost.
Chemistry, format and the obligations each creates
The three chemistries available from Chinese suppliers behave differently in the product and differently in the paperwork.
Lithium-ion cylindrical cells in formats such as 18650, 21700 and 4680 are the workhorses. They are named for their dimensions, they are produced in very high volume, and their quality is consistent because the manufacturing process is mature. They suit power tools, e-bikes, energy storage and consumer electronics, and they are the easiest to second-source because the format is standardised. Volume prices for 18650 cells in the 2,500 to 3,000 mAh range have recently sat at roughly $0.80 to $1.50 each.
The legal consequence of a standard format is that substitution is easy, which is both an advantage and a risk. It is easy for you to qualify a second cell maker, and it is also easy for the assembler to substitute a cell without telling you, because the pack will still fit. The contract needs a change control clause that names the approved cell and prohibits substitution without written approval and, where the certification is configuration-specific, without re-testing.
Lithium polymer pouch cells use a soft casing and can be made in custom shapes and thicknesses, which is why they dominate phones, wearables, drones and slim devices. They typically cost $0.30 to $0.80 per Wh and have shorter cycle life, in the region of 300 to 500 cycles. The legal consequence is that a pouch cell is usually custom, which means tooling, a longer qualification and a certification that is tied to one supplier’s dimensions. Moving to a second source means a new mechanical design and a new approval, not a like-for-like swap.
Lithium iron phosphate cells offer thermal stability, long cycle life in the range of 2,000 to 5,000 cycles, and a lower energy density than nickel manganese cobalt chemistries by roughly 30 to 40 per cent. They cost 20 to 40 per cent more per Wh than NMC. They are the right choice where safety and longevity outweigh weight, which covers stationary storage, many e-bikes and industrial equipment. The legal consequence is favourable: a chemistry with better thermal stability produces fewer abuse-test failures and a more defensible safety case, which shortens the certification path and reduces the warranty reserve.
A pack is not a cell. It is an assembly of cells, a battery management system, wiring, connectors, a casing and, usually, a thermal design. The BMS monitors cell voltages, balances the cells during charging, prevents over-discharge, and in a smart pack communicates state of charge and state of health to the host device. Most hardware companies buy packs rather than cells, and the choice between the two is the first structural decision in the programme.
Sourcing cells directly gives control over provenance and cost, and requires in-house capability in cell selection, matching, BMS design and assembly. It suits programmes above roughly a hundred thousand units a year with a dedicated battery engineering team. Sourcing complete packs is the right default for everyone else, at the cost of an assembler margin of perhaps 15 to 30 per cent and less visibility into the cell.
In practice the decision is staged rather than made once. Prototyping is served by off-the-shelf packs bought from a distributor or a marketplace, which are fast, cheap and limited in form factor, and which are entirely appropriate for proving that the product works. Pilot production is served by a tier-2 assembler building a custom pack, with non-recurring engineering costs in the region of a few thousand to ten thousand dollars and a development period of eight to sixteen weeks. Volume production is where direct conversations with cell makers become worthwhile, and where the assembler’s margin becomes a number worth negotiating rather than accepting.
| Chemistry | Best for | Transport and certification consequence | Warranty exposure |
|---|---|---|---|
| NMC cylindrical | Power tools, e-bikes, consumer devices | Standard formats, mature test evidence, easy second source | Substitution risk if change control is weak |
| LiPo pouch | Wearables, phones, slim devices | Custom dimensions, certification tied to one supplier | Mechanical and swelling failures, shorter cycle life |
| LiFePO4 | Storage, industrial, safety-critical | Better abuse-test margin, heavier and bulkier | Lower thermal risk, longer service life to warrant |
The supplier chain, and what each layer can actually warrant
China’s battery industry is stratified, and the stratification is not a marketing hierarchy. It reflects what each layer can evidence, and therefore what each layer can promise.
Tier-1 cell manufacturers such as CATL, BYD, EVE Energy, Lishen, CALB and Gotion produce at gigawatt-hour scale and supply global automotive and consumer brands. Their quality systems are world class and their test evidence is comprehensive. Their minimum order quantities are in the millions of cells, their lead times run six to twelve months, and they do not generally sell to a startup directly. This matters less than founders think, because their cells are available through the assemblers who buy from them.
Tier-2 pack assemblers, typically with a hundred to a thousand employees, are where most hardware companies should sit. They buy cells from tier-1 makers, design and assemble packs, hold the certifications that matter, and will work at minimum order quantities in the range of 500 to 5,000 units. They have engineering capability for custom designs and the commercial incentive to take a new customer seriously. The concentration is in Shenzhen and Dongguan, with clusters in Jiangsu and Zhejiang.
Tier-3 workshops compete on price, frequently use recycled or graded cells, and often present certificates that do not survive verification. The quality is inconsistent, the safety risk is real, and the entity is usually too small to carry a warranty. They are appropriate only for products sold into unregulated markets at very low cost, and even then the insurance position is difficult.
Screening for a tier-2 assembler is a matter of six checks, and each takes a day. Time in business, with five years as a reasonable floor for a supplier that will hold your certification. Factory size and capacity against your forecast, so that you are neither too small to matter nor too large for their line. Certifications that can be verified with the issuing body rather than merely displayed. International references that will take a call. Research and development capability, because a pack is a design and not an assembly. And transparency about where the cells come from, which is the question that separates a serious assembler from a trader.
The red flags are equally specific. A price 30 per cent or more below the market average is a statement about cell provenance. No third-party certificates, or certificates that cannot be traced to a register. Vague answers about cell brand and grade. A high minimum order for samples, which suggests the sample line is not real. Reluctance to share a factory address or allow an unannounced visit. And pressure to place a large order quickly, which is the behaviour of a business that expects the buyer to learn something later.
Due diligence follows the same order every time. Verify the certificates with the issuing bodies. Call the references. Commission a factory audit, which a third-party service will typically perform for a few hundred to a couple of thousand dollars depending on scope. Buy samples independently and test them rather than relying on the supplier’s data. And place a small pilot order before any volume commitment, because the pilot is where the process is revealed.
The layer that matters for warranty purposes is not the cell maker but the assembler, because the assembler is your counterparty. Two questions determine what the warranty is worth. Does the assembler have a balance sheet that can absorb a recall, and does the assembler have upstream recourse to the cell maker that it can pass to you. A tier-2 assembler with a strong cell relationship and a documented supply agreement can assign its upstream claims. A small workshop cannot, because it has no upstream agreement to assign.
Ask for the assembler’s cell supply agreement, or at least the cell purchase terms, before signing the pack agreement. If the assembler buys cells on the spot market, its upstream warranty is whatever the spot seller will give, which is usually nothing beyond a delivery inspection. That is the difference between a warranty that is real and a warranty that is a paragraph.
| Layer | What it sells | What it can warrant | What it cannot |
|---|---|---|---|
| Cell maker | Cells by the million | Cell performance to the assembler, on its own terms | Anything about your pack, your BMS or your enclosure |
| Pack assembler | Finished packs | The pack, for a period and a remedy it chooses | A recall larger than its balance sheet |
| Component suppliers | BMS, connectors, casing | Their component | The interaction between components |
| You | The product | The product to your customer, for longer than you are warranted | Recovery from a supplier that has no assets |
Qualification testing: the gates a pack must pass
Qualification is not a single test. It is a sequence of gates, each of which produces evidence that the next gate depends on, and each of which can stop the programme. Companies that treat it as a single certification event discover the sequencing problem late, when a design change made after certification invalidates the approval.
The gates below are the ones a pack programme normally passes through. The sample basis and the reference standards are set by the applicable standard and by your own specification, and should be confirmed with the testing laboratory rather than assumed.
Gate 0, specification and datasheet review. Before any sample is built, the specification states capacity, nominal and maximum voltage, continuous and peak discharge, charge profile, operating and storage temperature, cycle life, mechanical envelope, ingress protection, BMS functions and the safety requirements. The cell datasheet is checked against the specification rather than the reverse. The common failure at this gate is a specification written around a cell that cannot deliver the peak current at the specified temperature.
Gate 1, supplier qualification. The assembler is audited, its certificates are verified with the issuing bodies, its cell provenance is established, and its quality plan is reviewed. The audit should confirm cell matching tolerances, incoming inspection, traceability by lot, and the ageing regime. Reputable assemblers match cells within roughly 2 per cent on capacity and 5 milliohms on internal resistance; the numbers matter less than the fact that the assembler measures them and records them.
Gate 2, incoming cell verification. Cells are tested on receipt for capacity, internal resistance, open circuit voltage, self-discharge and dimensions, and the lot is recorded against the pack serial numbers it will go into. This gate is what makes a later claim possible. Without lot traceability, a field failure cannot be attributed to a cell batch, and the assembler will attribute it to the customer’s use.
Gate 3, design verification. The pack is tested for capacity, rate capability, thermal behaviour under load, BMS function including protection thresholds and balancing, charge and discharge cycling, and behaviour at the temperature extremes of the specification. This is where most design problems appear, and it is cheaper to find them here than at certification.
Gate 4, abuse and safety testing. The pack is subjected to conditions it should survive without fire: overcharge, external short circuit, crush, impact, nail penetration in some standards, thermal propagation in others, and drop. The applicable tests and the pass criteria come from the standard for the target market, and the results feed the safety case that the certification body will review.
Gate 5, transport testing. UN 38.3 is the transport test regime, and it is a precondition for lawful shipment rather than a market approval. It comprises eight tests: altitude simulation, thermal cycling, vibration, shock, external short circuit, impact or crush, overcharge and forced discharge. The report is issued to the applicant, covers a specific configuration, and must be available to the carrier.
Gate 6, market certification. Depending on the destination, the pack is certified under IEC 62133-2, UL 1642 and UL 2054, the applicable European directives, and the national schemes of Japan, Korea, India and others. Certification is configuration-specific and time-limited.
Gate 7, pilot and production acceptance. The first production lot is inspected against the golden sample, aged for a defined period at full charge, and released against an acceptance sampling plan. Production acceptance continues for the life of the programme, because the risk that matters is not the first lot but the fifteenth.
| Gate | What is tested | Reference | Sample basis | Who signs off | What typically fails |
|---|---|---|---|---|---|
| 0 | Specification against cell datasheet | Your specification | Documents | Your engineering lead | Peak current at temperature |
| 1 | Assembler capability and provenance | Audit checklist | One audit, annual repeat | Your quality lead | Cell matching records |
| 2 | Cell capacity, resistance, voltage, self-discharge | Cell datasheet, your specification | Per lot, sampled | Assembler quality, verified by you | Traceability to pack serial |
| 3 | Capacity, rate, thermal, BMS, cycling | Your specification | Engineering samples | Your engineering lead | Thermal margin at peak load |
| 4 | Overcharge, short circuit, crush, propagation | Applicable safety standard | Standard-defined | Third-party laboratory | BMS protection thresholds |
| 5 | Altitude, thermal, vibration, shock, short, crush, overcharge, forced discharge | UN 38.3 | Standard-defined | Testing laboratory | Vibration and shock on the enclosure |
| 6 | Market safety and EMC | IEC 62133-2, UL, CE and national schemes | Standard-defined | Certification body | Marking and documentation |
| 7 | Production conformity | Golden sample, AQL plan | Per lot | Your quality lead | Cosmetic and weld quality |
The gate sequence has one property that founders underestimate. Each gate is cheaper than the one after it and more expensive to skip. A specification error found at Gate 0 costs a meeting. The same error found at Gate 6 costs a redesign, a new certification and a delayed launch, and the delay is usually longer than the redesign because the laboratory queue is the constraint.
The calendar follows the complexity rather than the supplier’s optimism. A simple pack, such as a two-cell lithium polymer design with a basic protection board, typically runs from specification to certified sample in six to eight weeks. A complex pack, such as a multi-cell lithium iron phosphate design with a smart BMS, a custom enclosure and several market certifications, typically runs sixteen to twenty-four weeks, and the certification portion alone can absorb four to eight weeks of that. Non-recurring engineering charges sit in the low thousands to ten thousand dollars for a standard design and can exceed fifty thousand dollars where custom tooling and multiple certifications are involved. Volume production after design freeze usually takes three to six weeks, subject to cell availability.
Budget the certification time separately from the development time. Development is under the supplier’s control and certification is not, because the laboratory queue, the sample submission and any retest after a failure are outside it. A launch date that assumes certification runs in parallel with development is a launch date that will slip.
Transport classification and the documents that travel with the goods
Lithium batteries move as Class 9 dangerous goods, and the classification depends on how the battery is packed rather than on what it is.
Cells and batteries shipped on their own are UN 3480 for lithium ion and UN 3090 for lithium metal. Batteries packed with equipment are UN 3481 and UN 3091. Batteries contained in equipment, which is the case for a finished device with a built-in pack, are UN 3481 and UN 3091 as well, distinguished by the packing instruction rather than the UN number. The air transport rules then separate the three cases into packing instructions, with the lithium ion instructions running from the standalone case through packed with equipment to contained in equipment.
Two restrictions catch hardware companies repeatedly. Standalone lithium ion cells and batteries shipped by air are capped at a state of charge of 30 per cent, a rule that exists because a charged cell is a greater fire risk in a cargo hold. And the relief available for small batteries depends on the watt-hour rating, with cells at or below 20 Wh and batteries at or below 100 Wh qualifying for a lighter regime and anything above moving into fully regulated freight. Both rules are technical instructions that change, and both should be confirmed with the forwarder for the specific shipment rather than taken from an article.
| Shipment | UN number | Air packing instruction | Key restriction | Document owner |
|---|---|---|---|---|
| Cells or packs alone | 3480 | Lithium ion cells and batteries | State of charge cap by air; watt-hour threshold for relief | Shipper |
| Packs packed with equipment | 3481 | Lithium ion, packed with equipment | Outer packaging and quantity limits | Shipper |
| Device with pack inside | 3481 | Lithium ion, contained in equipment | Protection against short circuit and damage | Shipper, usually the factory |
| Lithium metal cells | 3090 | Lithium metal cells and batteries | Different watt-hour and lithium content limits | Shipper |
The documentation set is not negotiable and it is not the factory’s private business. It comprises the UN 38.3 test summary, which the air rules require to be made available, the safety data sheet, the dangerous goods declaration signed by the shipper, the packing certificate where the mode requires one, and commercial documents that describe the goods consistently. Inconsistency between the invoice description and the dangerous goods declaration is the single most common cause of a shipment being held, and the hold usually happens at the transit point rather than at origin, which means the goods sit in a warehouse while the launch date passes.
The party who signs the declaration carries the liability for it. Where the factory arranges transport, the factory is the shipper and the exposure is theirs, which is the reason to let them arrange it. Where you arrange transport from the factory gate, you may become the shipper, and you should know that before the first shipment rather than after. Where the factory arranges it, the contract should require it to provide the declaration and the test summary to you, because your customer and your insurer will ask for them.
Two practical points reduce the friction. Use a forwarder that handles batteries routinely and ask for the acceptance criteria of the airline before booking, because not every carrier accepts every class of battery shipment. And treat a factory that offers to describe the goods as “electronic components” to obtain a cheaper rate as a factory that has just transferred a serious regulatory risk onto you.
Supplier liability under Chinese law
Chinese product liability starts from a position that is stricter than many foreign buyers assume and weaker in practice than the text suggests.
A producer is liable for injury or property damage caused by a defective product, and the standard is not negligence; the producer is liable once the product is shown to be defective and the damage to be caused by it. A seller is liable where it supplied the product and cannot identify the producer or the supplier above it. A defect in this context means an unreasonable danger to person or property, and a product that does not comply with the applicable safety standard is treated as defective.
The practical difficulty is not the standard but the defendant. A tier-3 assembler with a small balance sheet can be liable in full and unable to pay. Enforcement requires assets, and a judgment against a company that has been wound down and reconstituted under a similar name is worth very little. The consequence for contracting is that the identity and financial substance of the counterparty matter more than the wording of the liability clause, and the wording matters only to the extent that it can be enforced against something.
There are three ways to close that gap, and a serious programme uses more than one. The first is to contract with an entity that has substance, which may mean requiring the assembler’s parent or operating company rather than a trading subsidiary to sign. The second is to take security, in the form of a parent guarantee, a retention held against the final payment, or a deposit that is only released on the expiry of the warranty. The third is insurance, in the form of a product liability policy that names you as an additional insured and that responds to a recall.
The regulatory layer sits on top of the civil one. Defective consumer products are subject to a recall regime administered by the market regulation authorities, under which a producer that identifies a safety defect must investigate, report and recall, and the authority can compel a recall where the producer does not act. For a foreign brand, the recall obligation usually reaches you through your importer and through the destination market’s own regulator, and the cost lands on you before it lands on the assembler. That is why the recall cost allocation belongs in the contract rather than in the insurance file.
Criminal exposure exists where products that do not meet safety standards are produced or sold, and the threshold and the penalties are set by law and by judicial interpretation. The relevant point for a foreign buyer is documentary. An organisation that can show a specification, a test regime, a supplier audit and a response to a known defect is in a different position from one that cannot, and the records that demonstrate diligence are the same records that support the civil defence.
| Risk | Who bears it by default | What to negotiate |
|---|---|---|
| Cell defect causing a pack failure | Assembler to you, if it can be traced | Lot traceability and upstream claim assignment |
| Design defect in the pack | You, as the designer of record | Design responsibility clause and a design review record |
| Injury from a field failure | Producer and seller jointly, in theory | Named insured, guarantee, retention |
| Recall cost | You, in practice, in the destination market | Cost allocation formula and a field failure trigger |
| Certification invalidity after a substitution | You, as the certificate holder | Change control and re-testing at the supplier’s cost |
Allocating warranty and recall cost across the chain
The warranty chain is where battery programmes quietly fail, because the periods do not line up and nobody draws the diagram.
A cell maker typically warrants cells for a period measured from shipment of the cells, on a pro-rata basis, to its direct customer. A pack assembler typically warrants the pack for a period measured from shipment of the pack, with a remedy that is repair or replacement rather than a credit for the downstream cost. Your customer, whether a distributor or an end user, expects a period measured from retail sale, and in some markets a statutory minimum applies regardless of what your contract says. The difference between the cell warranty start date and the customer warranty start date can easily be a year, and that year is your exposure.
Four mechanisms close the gap, and they should be negotiated together rather than one at a time.
Pass-through and assignment. The pack agreement should require the assembler to assign to you its rights against the cell maker in respect of cells incorporated into your packs, and to co-operate in a claim. An assembler that refuses is telling you that its upstream position is weaker than it presented.
Lot traceability. Every pack should carry a serial number that maps to a cell lot, and the assembler should retain that mapping for longer than the warranty period. Without it, a failure cannot be attributed, and an unattributed failure is treated as the customer’s fault.
Defect credit rather than repair. The remedy should be a credit for each defective unit, plus the cost of returning or scrapping it, rather than a repair service. Repair is appropriate for a unit returned by a single customer; it is useless when ten thousand units are in the field.
A field failure trigger. The agreement should define a failure rate above which the supplier funds a lot-level containment or recall, and should state how the rate is calculated and who pays for the investigation. The threshold is a commercial negotiation; the existence of a threshold is not.
Three further protections belong in the same negotiation. A spares obligation, so that replacement packs remain available for the service life of the product. A recall cost formula that allocates the cost between the parties by cause rather than by who is holding the goods. And insurance, with the assembler naming you as an additional insured and providing a certificate annually, because a policy that is allowed to lapse is not a protection.
On the statutory side, where the contract sets no inspection period, a buyer must notify defects within a reasonable time and, in any event, within two years of receipt, unless a warranty period has been agreed. A contractual warranty period displaces that default, which means the number you write in the contract is the number that governs your claim against the supplier. Write it to be at least as long as your exposure to your own customer, and check that the remedy survives the end of the supply relationship.
| Layer | Typical period | Typical remedy | The gap | The fix |
|---|---|---|---|---|
| Cell maker to assembler | Measured from cell shipment | Pro-rata credit | Starts before the pack exists | Assignment of claims |
| Assembler to you | Measured from pack shipment | Repair or replacement | Shorter than your customer’s expectation | Extended term, defect credit |
| You to customer | Measured from retail sale | Replacement, sometimes consequential | Longer than your upstream cover | Warranty reserve, insurance |
| End user to you | Statutory minimum in some markets | Repair, replacement or refund | Not controllable by contract | Price the reserve into the product |
Certification, market access and export control
Certification is the gate between a pack that works and a pack that can be sold, and it is configuration-specific in a way that catches companies that treat it as a one-off.
The core instruments are IEC 62133 for cell and pack safety, UN 38.3 for transport, and the destination schemes. In Europe, the applicable directives require safety and electromagnetic compatibility evidence, and the battery regulation adds obligations that phase in over several years, covering carbon footprint declaration, recycled content, a battery passport and supply chain due diligence. In the United States, UL 1642 covers cells and UL 2054 covers packs, the Federal Communications Commission rules cover electromagnetic compatibility, and the incentive and trade measures for batteries are tied to sourcing and assembly conditions that change more often than the safety standards. In Asia, PSE in Japan, KC in Korea and BIS in India each impose their own scheme. For sale inside China, the national standards for portable and other lithium batteries apply.
Verification is a separate task from collection. Suppliers present certificates that are expired, that cover a different model, or that were issued by a body without accreditation. The check is to take the certificate number to the issuing body’s register, confirm the model designation matches the pack you are buying, confirm the applicant, and ask for the full test report rather than the certificate. Accreditation of the laboratory can be checked against the international mutual recognition arrangements and the Chinese national accreditation scheme.
Three failure modes recur. A cell certificate is presented as a pack certificate, which it is not, because a pack has its own enclosure, BMS and thermal design. A certificate for one configuration is presented for another, which is the normal consequence of a supplier substitution. And a certificate is allowed to expire during the product’s life, which is discovered when a customer’s compliance team asks for a current copy.
Export control runs in the other direction. China has added certain battery materials, and more recently selected battery technologies and equipment, to its export control lists, and the scope has changed more than once. A technology licence, a process transfer or the export of production equipment to a plant outside China may require a licence, and the classification is not always obvious from the description. Check the current catalogue before agreeing to any transfer of technology, and treat the answer as time-limited.
| Market | Instrument | What it covers | Who holds it | What invalidates it |
|---|---|---|---|---|
| Global transport | UN 38.3 | Transport safety of the configuration | The applicant, usually the manufacturer | A change of cell, BMS or enclosure |
| International | IEC 62133 | Cell and pack safety | The applicant | Configuration change, expiry |
| European Union | CE marking under the applicable directives, plus the battery regulation | Safety, EMC, and phased battery-specific duties | The economic operator placing it on the market | Configuration change, missing due diligence |
| United States | UL 1642, UL 2054, FCC | Cell and pack safety, electromagnetic compatibility | The applicant | Configuration change, lapsed surveillance |
| China | National standards for lithium batteries | Domestic sale | The manufacturer | Configuration change |
Where foreign companies get this wrong
Buying to a cell datasheet rather than a pack specification. The datasheet describes a cell at a defined temperature and rate. The pack has an enclosure, a BMS, a thermal path and a duty cycle. A programme specified from the datasheet will pass the bench test and fail the field, and the failure will appear at Gate 3 or later when the redesign is expensive.
Accepting a test report that covers a different configuration. The report is genuine and the pack is not the one it describes. The consequence is a shipment that is non-compliant, a customs hold, and a certification that cannot be relied on if a customer audits the file.
Misdeclaring a shipment. The freight saving is a few hundred dollars and the exposure is a criminal offence in the destination country, an insurance void and a carrier claim. The consequence is disproportionate to the saving in a way that no finance team would approve if it were presented as a risk decision.
Relying on a warranty from an entity with no assets. The clause is enforceable and the defendant is empty. The consequence is that the buyer carries the recall alone, which for a small company is an existential event rather than a bad quarter.
Treating certification as a one-off. The certificate is obtained for the launch configuration and the supplier changes a cell eighteen months later to protect margin. The consequence is that the product is sold without valid approval, and the discovery happens through an audit rather than through a failure.
Single-sourcing a cell during an allocation cycle. Tier-1 cell capacity is allocated to the largest customers first, and a small buyer can lose supply with little notice. The consequence is a line stoppage at the worst moment, and the cost of a second source qualified in a hurry is higher than the cost of qualifying it in advance.
Cost, and where the money actually goes
Cells account for 60 to 70 per cent of pack cost, the BMS for 10 to 15 per cent, the circuit board and wiring for 5 to 10 per cent, the casing and structure for 5 to 10 per cent, and assembly and testing for the remainder. Those proportions describe the purchase price, not the cost of owning the pack. The total includes certification per configuration, the freight premium for dangerous goods, insurance, a warranty reserve that should be several per cent of revenue for a new chemistry, the carrying cost of two to three months of safety stock, the cost of qualifying a second source, and the end-of-life obligation in markets that impose one. A programme that budgets the purchase price and not the rest will find that the pack is the largest unplanned cost in the product.
Practical checklist
- Write the pack specification before contacting any supplier, and derive it from the duty cycle rather than from a cell datasheet.
- Require lot traceability from cell to pack serial, and confirm that the assembler retains the mapping beyond the warranty period.
- Verify every certificate with the issuing body, check the applicant and the model, and obtain the full test report.
- Confirm that the UN 38.3 report covers the exact configuration you will ship, including the enclosure and the BMS.
- Ask for the assembler’s cell supply terms, and require assignment of its upstream warranty claims.
- Negotiate a defect credit rather than a repair remedy, and a field failure trigger that funds containment.
- Require product liability insurance naming you as an additional insured, with an annual certificate.
- Take a parent guarantee or a retention where the assembler’s balance sheet cannot carry a recall.
- Agree the transport classification and the document set with the forwarder before the first shipment, and never let anyone describe the goods as something they are not.
- Check the export control position before any technology transfer, and check the destination market’s battery-specific duties before the launch date is fixed.
The battery is the component where the engineering and the legal cannot be separated, and the programmes that survive a field failure are the ones that built the warranty chain before they built the pack. A supplier that can evidence its process, trace its cells and carry its share of the risk is worth a higher unit price than one that cannot, because the difference is not a cost. It is the difference between a recall that is survivable and one that is not.
For the contract terms that carry this allocation, see commercial contracts and IP protection. The guides that continue this sequence are China product certification, supply chain risk management in China and warranty and defect liability in China manufacturing.
Frequently Asked Questions↑ Back to article
Is it safe to source lithium batteries from China?
Yes, provided you work with certified, audited suppliers. China produces the majority of the world's lithium batteries, including those used by top global brands. The key is thorough due diligence: verify certifications (UL, CE, IEC 62133, UN 38.3), inspect the factory, test samples independently, and specify clear quality requirements. Risks come from cutting corners with unvetted low-cost suppliers, not from China as a source per se.
What certifications do batteries from China need?
Required certifications depend on your target market. The most common are: CE (EU), UL 1642/2054 (US), FCC (US electromagnetic compatibility), IEC 62133 (international safety standard), UN 38.3 (transport safety), PSE (Japan), KC (Korea), BIS (India), and GB standards (China domestic market). Your supplier should provide test reports from accredited third-party labs — always verify certification authenticity directly with the issuing body.
How do I verify a Chinese battery supplier is legitimate?
Check their business license, certifications (verify with issuing bodies), factory location (visit in person or use a third-party auditor), years in operation (prefer 5+ years), client references (especially international ones), and R&D capabilities. Red flags include: prices 30%+ below market average, refusal to share factory photos or address, no third-party certifications, vague answers about cell sources, and high MOQs for samples. A factory audit by a reputable third-party service is the gold standard.
What is the typical lead time for custom battery packs in China?
Custom battery pack development typically takes 8–16 weeks from specification to first prototypes, plus another 4–8 weeks for certification and testing. Volume production lead time is usually 3–6 weeks after design is finalized. Off-the-shelf standard packs can ship in 1–3 weeks. Timelines vary based on complexity: a simple 2-cell LiPo pack with standard BMS is on the shorter end, while a multi-cell LiFePO4 pack with smart BMS, custom enclosure, and multiple certifications will be longer.
How much do lithium batteries cost from Chinese manufacturers?
As of late 2025, cylindrical 18650 cells (2500–3000mAh) cost $0.80–$1.50 each in volume; 21700 cells cost $1.20–$2.00. LiPo pouch cells range from $0.30–$0.80 per Wh. Custom battery packs range from $5–$50+ depending on capacity, chemistry, BMS features, and certification. LiFePO4 packs are typically 20–40% more expensive per Wh than NMC. Cells account for 60–70% of total pack cost. Prices fluctuate with raw material prices — lithium carbonate swings can shift cell prices by 20–50% in a year.
What battery chemistries are available from Chinese manufacturers?
Chinese manufacturers offer three main lithium battery chemistries, each with different trade-offs. Li-ion cylindrical cells in formats like 18650 and 21700 offer consistent quality and wide availability for power tools, e-bikes, and consumer electronics. LiPo pouch cells can be made in custom shapes and thicknesses, making them ideal for smartphones, wearables, and slim devices. LiFePO4 cells offer exceptional safety and long cycle life of 2000–5000+ cycles but at 30–40% lower energy density, preferred for EVs and stationary storage.
What is the difference between tier-1, tier-2, and tier-3 battery suppliers in China?
China's battery industry is highly stratified. Tier-1 cell manufacturers like CATL, BYD, and EVE Energy produce at gigawatt-hour scale and supply global automakers but have enormous MOQs in the millions of cells with 6–12 month lead times, putting them out of reach for most startups directly. Tier-2 pack assemblers with 100–1000 employees are the sweet spot for most hardware companies, sourcing cells from tier-1 makers and assembling custom packs with MOQs as low as 500–5,000 units. Tier-3 small workshops compete almost exclusively on price, often using recycled cells with inconsistent quality and fake certifications, and should be avoided for regulated markets.