What Actually Drives the Price of a Custom Robot Battery Pack

Date Author Lot Trace

The unit price of a custom lithium-ion pack for an AMR, AGV, or industrial robot arm is determined by a handful of decisions made long before anyone quotes a number: which cells, what the BMS is required to do, what the enclosure has to survive, how many units the program will truly build, and which certifications have to be on file. A quote that does not separate those into line items for cells, BMS, mechanical hardware, interconnect, assembly labor, test, and certification is not a price — it is a negotiating position. The buyer's real job is to convert it back into a cost model using a common breakdown template, and to do that before the RFQ goes out, not after the PDFs come back.

Three quotes, one cold cup of coffee

It's 9:40 on a Thursday and I have three PDFs open on the second monitor. The first is nine pages with a clean table. The second is two pages, and the pricing section is a single row: "48 V robot pack, qty 250, unit price." The third has forty-one rows, and about half the labels are abbreviations from the supplier's ERP — I genuinely cannot tell whether "PCBA-ASM-01" means the BMS board or the entire pack assembly.

My coffee went cold around 8:15. I open a blank spreadsheet before I read any of them. Not after. Before. That ordering is the whole discipline, and it's the thing nobody tells you when you start sourcing these programs: the template has to exist before the first quote arrives, because whatever you read first will quietly become your mental model for what a battery pack costs, and the second supplier will get graded against it whether or not that was fair.

So I build columns. Eleven of them, in this order: cells, BMS, enclosure and mechanical, interconnect and harness, assembly labor, test and burn-in, certification and documentation, tooling and NRE, logistics and packaging, warranty and spares, and margin.

Then I read each quote a second time and drop every number into a cell. Where there's no number, I type a question mark. Not a zero. A zero is a claim. A question mark is an open item, and I want to see how many of them there are, because the count itself tells me something about the supplier.

Cells are the anchor, and they're priced by three things at once

Cells are the largest single line in almost every pack I've costed, and they move on chemistry, capacity, and continuous discharge rating — but the one buyers miss is the fourth variable, which is grade and traceability. A pack built on cells with a documented manufacturing lot and a matching test report costs more than one built on whatever was on the shelf. That difference is not a scam. It's the difference between being able to answer a customer's field-failure question in eighteen months and not being able to.

The discharge rating is where sizing discipline shows up in the quote. If you spec the pack against the robot's peak draw instead of its duty-cycle average, you buy more parallel strings, and you buy them in every downstream line item too — more cells, more interconnects, a bigger enclosure, more tab welds, longer assembly. One requirement, four line items. That's the pattern you're looking for.

The BMS is a step function, not a slope

Adding CAN to a board that already has a microcontroller is a firmware question and doesn't move the number much. Adding active balancing, a second contactor with weld detection, dual current sensing, an independent over-current path, or a wireless authorization input for a fleet's remote e-stop is a different board, a different layout, and sometimes a different processor family. Those are steps.

So when a supplier shows me a BMS line that moved by a large amount between two revisions, I don't ask why it's expensive. I ask which of the four things I asked for in that revision required a new board. If they can't name it, the number is soft.

The communication requirement deserves its own sentence here. If your fleet controller talks RS485 and the supplier's board only speaks CAN, the adapter isn't free and it isn't in the BMS line — it shows up in the harness. Which is why the harness gets its own column.

Where the mechanical and assembly lines actually come from

Ingress protection is the biggest single lever on the enclosure line, and it's not really the material — it's the sealing labor and the test. Moving from a splash-rated pack to something that survives washdown means a different gasket geometry, sealed glands, a pressure-equalizing vent, and a pressure-decay leak test at end of assembly. All four are line items, and three of them are labor.

Shock and vibration are quieter. An AMR that rolls over dock plates and a six-axis arm that whips at the end of a fast move do not load a pack the same way, and the difference shows up as wall thickness, mounting bosses, and how many threaded inserts somebody has to install by hand. Every insert is touch labor. Every threaded fastener is torque time. Assembly labor is the most understated line in these quotes because it's the easiest to compress on paper and the hardest to compress on the floor.

Anything potted adds cure time that runs on the clock. That's real cost, and it's legitimate, but you should see it, because a supplier who quotes potting and then doesn't pot the pilot units is telling you where they cut.

Test is a menu, and you pick from it

End-of-line insulation resistance and hipot should be table stakes. Above that, the menu includes pack-level capacity verification, thermal soak, charge/discharge cycling, and incoming cell screening. Each is a line item and each buys you a different kind of quiet.

Pack-level capacity verification costs more and catches mismatched cells before they become a warranty conversation. Whether you want it depends on how expensive a field failure is for you — and that's a business question, not an electrical one. I've bought it on packs that ship in two-hundred-unit fleet orders and skipped it on prototypes, and I'd make both calls again.

Certification is not overhead, and this is where quotes get slippery

UN 38.3 is required for transport regardless of application, and it applies to the pack as a design type — a pack assembled from cells that carry their own UN 38.3 reports still needs its own test summary. Ask for the test summary by name. It is a document that is supposed to exist and be available, and a supplier who says "yes, UN 38.3, included" without producing the summary has told you that you're buying a promise.

The UL question depends on how the pack is classified. UL 2054, UL 2271, and UL 2580 describe different applications and carry different test programs, and IEC 62133-2 is a separate track that some customers accept and others don't. I'm not going to tell you which one your robot needs, because that depends on your customer, your insurer, and your AHJ — but I will tell you that the classification decision is a cost decision, and it changes the certification line item by a lot.

What matters for comparing quotes: whether certification appears as its own line with a scope statement, or is buried in a general "engineering and documentation" bucket. Buried certification is the single most common way a cheap quote stops being cheap. Either the work isn't being done and you'll find out at the port, or it is being done and you have no visibility into whether the scope covers what you need.

Shipping counts here too. Dangerous goods packaging, class 9 labeling, and the state-of-charge limit that applies to air transport of cells and batteries handled separately from equipment all show up as logistics lines. If they're missing, they're not free.

NRE, tooling, and the quantity nobody wants to name

Tooling and NRE only reduce unit price when they're divided by a quantity that will actually be built. Two quotes can land at the same unit price with completely different structures — one with no NRE and a higher unit cost, one with NRE and a lower unit cost — and they cross over at different volumes. If your program is 250 units and the supplier amortized tooling across 5,000, the unit price you're holding is fiction, and the correction arrives as a mid-program surprise nobody can defend in a budget review.

So the question I ask is not "what's the unit price." It's "what quantity is this unit price amortized over, and what happens to my unit price if we build half of it?" The answer needs to be in writing before the PO, not discovered on the second release.

What a missing line item actually means

There are only three explanations for a blank cell.

It's absorbed into overhead. That means you can't audit it and you can't compare it, and when the supplier's costs move, you'll have no idea which line moved.

It's deferred to a change order. The NRE shows up after you've committed, when the tooling has to be cut and there's no competing bid in play. This is the expensive version.

Or it genuinely isn't being done. Which is fine for some lines and catastrophic for others, and the only way to know which one you're looking at is to ask directly and watch what happens to the tone of the email.

The comparison only works if the template is the same

The reason I build eleven columns before reading anything is that a quote is a narrative. It's arranged to be persuasive. The nine-page quote puts its strongest numbers first and quietly folds labor into "assembly." The two-page quote relies on you not wanting to open the attachment. The forty-one-row quote buries the thing you'll care about in row thirty-three.

A shared template flattens all of that. Same columns, every supplier, question marks where the numbers should be, and a note on each line about which requirement in my spec produced it. When I hand that to a supplier and ask them to fill it in, I learn more from how they respond than from anything in the original PDF. The ones who push back earnestly and explain why a line can't be separated are telling me they understand their own cost structure. The ones who agree instantly and return a filled template with suspiciously round numbers are telling me something else.

The last thing I do before I close the laptop is write one sentence at the top of the sheet: which design decisions in this quote would be cheap to change today and ruinous to change after tooling. Supersized enclosure, potted assembly, a custom BMS layout, a certification scope that doesn't cover the second product variant. Those are the rows where the price isn't really the price — it's the cost of the options I'm giving up.

That sentence is usually the only thing anybody remembers from the meeting.

Lot Trace

About the author

Lot Trace

Lot Trace is a byline shared by sourcing, quality, and compliance professionals who have spent their working lives on the buyer's side of custom battery pack programs. They have walked pack assembly lines in the Midwest and in Guangdong, reviewed cell lot records and UN 38.3 test summaries, and sat through PPAP reviews for robot programs built on both sides of the Pacific.

More about us