5 Million Units in 4 Months: What the Huawei Mate9 Run Tells Engineers
The Mate9 shipped about 5 million units in 4 months after its late 2016 launch, yet it still gave up margin to hold volume. This explainer breaks down the pricing pressure, the cost drivers inside the phone, and what that lesson means for anyone building a hardware bill of materials today.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
Why 5 Million Units in 4 Months Still Was Not Enough
When the Mate9 reached roughly 5 million units in 4 months, the number looked like proof that Huawei had crossed into the top tier. It had. What the figure did not show was the price at which those units moved. A flagship that sells fast can still sell badly, because the average selling price falls as discounts, carrier bundles, and channel incentives pile up.
The year 2016 had three players fighting for the same premium shelf: Apple above, Samsung beside, and Huawei pushing up from below. Every discount Huawei offered to defend volume cut directly into gross margin. Volume grew; profit per unit did not.
For an engineer reading this years later, the useful question is not whether 5 million units in 4 months was a win. It is which parts of the phone carried the cost, and which of those costs a design team can actually control. That is where the story turns into something practical.
The short version: a fast ramp hides cost problems. Slow ramps expose them. Either way the cost is already in the part.
Where the Money Goes Inside a Flagship Chassis
A premium phone is mostly a mechanical problem wrapped around a battery and a board. The unibody shell, the mid-frame, the camera ring, the SIM tray, the antenna inserts, and the speaker grille are all machined or die-cast metal. On a flagship these parts are held to tight tolerances because they must mate with glass and with each other at a visible seam.
The shell alone is usually the single largest mechanical cost. A 7000-series aluminium unibody starts as a forging or a billet, then goes through roughing, semi-finishing, finishing, and anodizing. Anodizing is not a cosmetic afterthought; hardcoat and colour anodizing add process steps and a yield risk that shows up as scrap.
The mid-frame carries the stiffness. It also carries the tolerance stack. If the frame is out by 0.05 mm, the glass back may not sit flush, and the whole assembly line slows down. That is why phone frames are typically specified at ±0.02 mm to ±0.05 mm on critical features, with surface finish around Ra 0.8–1.6 μm before coating.
None of this is exotic for a machine shop. It is ordinary 3-axis and 4-axis work on aluminium, run at volume. The cost is not the machining itself; it is the fixture count, the inspection labor, and the scrap rate at ramp.
When a Milled Metal Body Stops Making Sense
Milled aluminium looks premium and feels stiff, but it is a poor choice when the target retail price is low. A machined unibody can cost several times a die-cast or stamped frame at the same volume. If the phone is meant to sell in a mid-tier band, the metal body is usually the first thing to go.
The break-even depends on volume and on wall thickness. Thin walls under about 0.8 mm get expensive fast, because tool deflection and chatter drive the scrap rate up. Deep pockets with a high aspect ratio need long, slender tools and slow feed rates. Cycle time climbs, and cycle time is money at any volume.
Die casting wins when geometry is complex and volume is high. Stamping wins when the part is flat and the tolerance is loose. Machining wins when the part is small, tight, and needed before the tooling is ready. That last case is the prototype stage, and it is where most programs actually start.
There is also a thermal reason. Metal conducts heat away from the battery and the SoC, which helps sustained performance. A plastic frame insulates. On a flagship that markets speed, that difference is part of the product, not just the bill of materials.
How Price Cuts Reach the Machine Shop
When the retail price of a flagship drops, the pressure travels down the supply chain in a predictable order. First the brand cuts channel margin. Then it asks contract manufacturers for a cost-down. Then the contract manufacturer asks component and machining suppliers for the same. Each step usually lands on the smallest supplier last.
A cost-down request on a machined part almost always targets cycle time, material, or inspection. Cycle time can be cut by loosening a tolerance or combining operations on a 5-axis machine. Material can be cut by switching from 7075 to 6061, or from billet to near-net forging. Inspection can be cut by moving from 100% to sampling, which is a risk decision, not a savings decision.
The trap is that loosening tolerance late in a program is expensive. Fixtures are already built, the anodizing process is already qualified, and the assembly line is already tuned. A change that saves a few seconds per part can cost weeks of re-qualification.
This is the real lesson of 5 million units in 4 months. Cost is locked in early, during design. Once the ramp starts, you are mostly choosing which risk to accept, not which cost to remove.
What a Design Team Should Do Before the Ramp
Get a manufacturability review before the tooling is cut. A DFM pass on the shell, frame, and small metal parts will usually find features that are hard to hold, thin walls that will chatter, or tolerances tighter than the function needs. Fixing those on screen is cheap. Fixing them after first article is not.
Pick the process for the volume, not for the drawing. Prototypes belong on 3-axis, 4-axis, or 5-axis mills, where no tooling is needed and changes are a program edit. Production belongs on the process with the lowest total cost at that volume, which is often die casting or stamping for phone-scale parts.
Decide the inspection plan early. On a visible cosmetic part, 100% inspection of critical dimensions is normal, and full reports should be available on request. On an internal bracket, sampling may be enough. The plan should match the consequence of a bad part, not the habit of the buyer.
Finally, treat surface finish as a functional spec, not a note. Anodizing thickness, colour consistency, and bead-blast texture all affect yield. If the finish is not defined with a number, the supplier will define it for you.
Machining vs Die Casting vs Stamping for Phone-Scale Metal Parts
Rough guide for small aluminium parts at 0.5–2 mm wall thickness
| Factor | CNC machining | Die casting | Stamping |
|---|---|---|---|
| Best volume band | 1 to 10,000+ parts | 50,000+ parts | 100,000+ parts |
| Tooling needed | None | Die, weeks of lead time | Progressive die |
| Typical tolerance | ±0.005 mm to ±0.05 mm | ±0.1 mm to ±0.3 mm | ±0.1 mm on flat features |
| Wall thickness | Down to 0.8 mm, cost climbs | Down to 1.0 mm, easy | 0.5 mm and below |
| Surface finish | Ra 0.8–1.6 μm as machined | As-cast, needs finishing | As-stamped, needs finishing |
| Design changes | Program edit only | Die rework | Die rework |
| Best use | Prototypes, tight features | Complex 3D shapes | Flat frames, shields |
| Weak point | Cycle time at high volume | Porosity and draft angles | Limited 3D geometry |
The Takeaway
If you need tight metal parts before tooling exists, machine them. If you need a million complex frames, cast them. Choosing the wrong one at the wrong volume is what eats the margin, not the retail price alone.
Questions Engineers Ask After Reading This
Does a high unit volume automatically lower the cost per part?
Only if the process matches the volume. CNC machining has almost no fixed tooling cost, so the first part and the ten-thousandth part cost roughly the same per unit once the program is stable. Die casting and stamping carry a large fixed cost that must be spread over the run, so their per-part price falls sharply with volume.
Below a few thousand parts, machining is usually cheaper overall. Above that, the amortized tooling cost of casting or stamping starts to win, provided the geometry allows it.
How much does a tighter tolerance actually cost?
It depends on whether the tolerance sits on a critical feature or a cosmetic one. A ±0.005 mm callout on a mating bore is a normal grinding or fine-boring operation. The same callout on a large free surface forces slower passes and more inspection, and it may push the scrap rate up.
The practical rule is to tighten only the features that stack into the assembly. Everything else should sit at ±0.05 mm or looser.
Why does anodizing affect yield on a phone shell?
Anodizing is a chemical conversion of the aluminium surface, so it responds to alloy, grain structure, and surface prep. Different alloys take colour differently, and hardcoat builds a thicker oxide layer that can crack at sharp corners.
That means the finish is a process risk, not just a cosmetic step. It has to be qualified together with the machining, ideally on the same alloy and the same geometry as production parts.
What is the fastest way to get a metal housing prototype?
Machine it. A 3-axis or 5-axis mill can cut an aluminium housing from billet with no tooling, and design changes are just a program revision. That lets a team test fit, finish, and feel before committing to a die.
Typical practice is to machine the prototype, freeze the geometry, and only then move to the production process.
Can a supplier hold tight tolerance and a good finish at the same time?
Yes, but they are separate operations. Machining sets the dimensional accuracy, and finishing sets the surface. Bead blasting or polishing after machining can move a dimension slightly, so the process order matters.
Parts should be inspected after finishing, not before, if the finish removes or adds material on a critical face.
How should a buyer compare quotes from different processes?
Compare total landed cost at the real annual volume, not unit price at one quantity. Include tooling amortization, inspection cost, scrap allowance, and the cost of a design change. A quote that ignores one of those is not comparable.
It also helps to ask each supplier which process they would choose for the part and why. The reasoning tells you more than the number.
Send Us the Part, We Will Tell You the Process
Upload a STEP file and we will return a quote with a free DFM analysis within 12 hours, plus a clear recommendation on whether the part belongs on a mill, a die, or a press.
12-hour quoteFree DFM analysisNo minimum order quantityNDA on request
More From the Shop Floor
We publish setup notes, tooling trials and inspection data from the factory floor.
