I spent a weekend two years ago rebuilding the bill of materials for one mid-tier Android phone. One device, one SKU, £399 at retail. I worked from teardown photography, component price sheets, and two conversations with people who would talk if I promised not to name them. I wanted a single number: what hardware industry margins actually look like on a phone that is neither an iPhone nor a sub-$100 handset — the segment where most venture-backed device companies actually live.
The components came to roughly $247. Display and touch stack was the biggest line. The SoC was second. Memory and storage were third and by far the most volatile. Add about $18 for assembly and test at the contract manufacturer, then another $22 or so for freight, duties, and a warranty reserve I set at three percent because that was the figure two different people told me was normal. Call it $287 landed.
Now the other side. £399 including UK VAT is about £332 ex-VAT, roughly $415 at that month's exchange rate. The channel — carriers, Amazon, the electronics chains — takes its cut before the maker sees anything. At fifteen points, the company collects around $353.
So: $66 a unit. Nineteen percent gross margin on a well-reviewed phone at a price people were willing to pay.
Then I did the part that changed how I read every hardware press release since. I divided the fixed costs in.
The margin is a volume problem wearing a component costume
This is the thing that takes a while to internalize: hardware industry margins in the mid-tier are set less by what you pay for parts than by how many times you get to divide your fixed costs.
A single phone generation at a small OEM costs somewhere between $30M and $50M before a unit ships — the engineering team for eighteen months, tooling, regulatory certification in every market you sell in, carrier lab testing, and the software group that has to keep existing devices patched while building the new one. My estimate for the device I was modeling was $38M, and I should be honest that tooling amortization is the one line I never got a straight answer on.
Ship 400,000 units and that's $95 a unit. You are underwater on a nineteen-point gross margin. Ship four million and it's under $10, and you have a business.
None of that is fixable by negotiating harder with suppliers. A company shipping half a million phones a year buys its Qualcomm silicon and its Samsung memory at something close to list, in a market where the two largest buyers have contracts nobody else gets to see. It cannot raise price, because Xiaomi and Google's A-series set a ceiling in every review that gets read. It cannot cut the bill of materials without cutting the specs the reviews are about.
Which leaves one lever: the size of the company.
Do layoffs at a phone maker mean the company is dying?
Usually not on their own. Hardware industry margins in this segment leave almost nothing to cut but people, so headcount reduction is the first competent response to a volume shortfall, not the last desperate one. The layoff tells you the company missed its unit forecast. It does not tell you whether the company has a plan. What tells you that is which people left.
I now sort every hardware restructuring into one of two piles, and the sorting takes about twenty minutes of reading.
Pile one: the company cut the future. Core hardware engineering, industrial design, the platform software team, the people who hold the carrier relationships. Those functions are the fixed cost that buys the next generation. A company that cuts them has decided — whatever the statement says — that there may not be a next generation at the current scale.
Pile two: the company cut the present. Regional sales entities, marketing, retail operations, categories that were never going to reach volume. That is a company narrowing to the markets where its unit economics work. It reads badly in a headline and it frequently works.
The history here is not subtle. LG left handsets in 2021 after roughly six years of divisional losses, having spent most of that stretch trimming the second pile and never recovering volume. HTC sold a large part of its phone engineering group to Google in 2018 for $1.1B — that was the first pile, and the handset business never came back from it. BlackBerry stopped designing its own hardware in 2016 and licensed the name instead. Meanwhile, most quarterly estimates over the past decade have put Apple alone above two-thirds of the entire industry's handset operating profit. Everyone else is dividing a remainder.
What to read instead of the press release
| The announcement says | What I check | What it usually means |
|---|---|---|
| "Restructuring to focus on core markets" | Which countries lost their sales entity, and whether engineering headcount moved | Pile two. Often survivable. |
| "We are not exiting the smartphone market" | Regulator filings (FCC and equivalents) for unreleased model numbers in the last two quarters | Two quiet quarters is the real exit, whatever the denial says |
| "Streamlining to invest in AI" | Whether the AI story lowers BOM, raises ASP, or creates recurring revenue | If none of the three, it is a fundraising narrative, not a business model |
| "Reducing roles across the organization" | Whether OS and security update commitments got quietly shortened | Support windows are expensive and ongoing; shortening one signals the software team is gone |
| Nothing at all about supply | Supplier and ODM commentary on order commitments | Makers commit cash to components two to three quarters out. One that stopped ordering already decided. |
Three checks that take under an hour:
- Search the FCC ID database, and the equivalent regulator wherever the company certifies, for filings from the last two quarters. Unreleased model numbers mean a roadmap. Silence means a decision.
- Pull the stated OS and security update window for the current flagship and compare it against what was promised on the previous one. A shortened window is a software-headcount signal no statement will admit to.
- Read the last two quarters of commentary from their contract manufacturer, where it is public. Order books surface there before they surface anywhere else.
The highest-signal item is the supply one. Everything else can be spun. Component commitments are cash going out against inventory that has to sell, and a company that has stopped placing them made its decision months before the communications team wrote anything.
I want to be fair to the companies, because the analyst reflex is to treat every restructuring as a confession. The margin structure in this segment punishes size in the middle. You are too small to get the pricing, too committed to specs to go downmarket, too unbranded to charge upmarket. Cutting to fit that is not panic. It is arithmetic somebody finally did out loud.
The number I got wrong
I reopened that spreadsheet about eighteen months after I built it, mostly out of curiosity, and updated three lines: memory, storage, and the silicon generation the phone would have to move to. Memory and NAND were in the upper half of their cycle. The new SoC cost more than the old one, as it does nearly every generation. The retail price had not moved, because the competitive set had not allowed it to.
Gross margin per unit went from about $66 to about $31.
Nobody at that company had done anything wrong between my two versions of the file. They shipped a good phone, on time, to decent reviews, and watched half their margin disappear into a commodity cycle and a silicon roadmap they do not control. That is the part I had to sit with. In this segment you can execute well and still lose ground, and when that happens the only honest move available to management is to become a smaller company that survives at the volume it actually has.
So when the layoff notice and the we-are-not-leaving statement land on the same afternoon, stop trying to work out which one is the lie. They are usually both true.
If a hardware company cuts the people who build the next product, believe the layoffs; if it cuts almost everyone else, believe the denial.
