Every comparison of a China CNC machining supplier against a local shop makes the same move: five reasons offshore is cheaper, a seven-row table, a paragraph conceding that local sometimes wins, and a quote button. The number at the centre of them, that China is 30 to 50 percent cheaper, is not wrong. It is just not the number that decides, because it compares one unit price with another and stops before the freight invoice arrives.
Landed cost per part is not a percentage. It is a stack of lines with different scaling laws: a fixture that costs the same at 25 pieces as at 2,500, freight that scales with kilos rather than value, duty that scales with value rather than kilos, and a rework allowance that scales with the distance to the fix.
私たちが発言する権利のあること: we run CNC加工 and finishing in Huzhou, China, so we benefit when you buy offshore. That is exactly why the cases where offshore loses appear here with numbers on them.
Every worked example states its assumptions. The arithmetic transfers; our rates do not.

What does landed cost actually look like at 25, 250 and 2,500 pieces?
Eight cost lines decide a sourcing question and they do not scale together, which is the whole problem with reading a unit price.
The part. A 6061-T6 aluminium bracket, 118 x 76 x 22 mm, 0.35 kg finished from a 0.95 kg blank, three setups on a 3-axis mill, four features held to plus or minus 0.05 mm, one pocket, two through-bores, one tapped pattern, clear Type II anodise. Not a hard part. A harder part would change the levels, not the shape of the argument.
Stated assumptions, all of them ours. China billed machine rate USD 28 per hour on 3-axis at an export-quality shop with a CMM; local shop in the US Midwest or Western Europe at USD 80 per hour all-in. Material at 0.95 kg buy weight, USD 3.60 per kg in China and USD 7.20 per kg locally including cut charge. Margin of 20 percent on the China build and 30 percent locally. Anodise priced with a lot minimum, which is why it looks brutal at 25 pieces. Duty modelled at 25 percent of goods value as a planning figure, not a statement about your tariff line. Air freight at 25 and 250 pieces, LCL sea at 2,500.
| 原価要素 | China 25 | China 250 | China 2,500 | Local 25 | Local 250 | Local 2,500 |
|---|---|---|---|---|---|---|
| Cycle time per part, minutes | 14 | 11 | 8.5 | 16 | 13 | 11 |
| Machining at billed rate | 6.53 | 5.13 | 3.97 | 21.33 | 17.33 | 14.67 |
| 材料 | 3.45 | 3.45 | 3.45 | 6.90 | 6.90 | 6.90 |
| Deburr and clear anodise | 3.40 | 2.10 | 1.65 | 9.50 | 5.20 | 3.90 |
| Cost to make | 13.38 | 10.68 | 9.07 | 37.73 | 29.43 | 25.47 |
| Margin applied | 20% | 20% | 20% | 30% | 30% | 30% |
| Quoted unit price | 16.10 | 12.80 | 10.90 | 49.00 | 38.30 | 33.10 |
The price gap is 67 percent at all three quantities. It comes from the rate, not the volume, so a bigger order will never close it. The volume curve is shallower than buyers expect: 25 to 2,500 pieces takes the China unit price from 16.10 to 10.90, a third off. Sourcing is worth twice what quantity is worth, and the two compound.
Now the lines the quote does not show.
| Line item | 25 pcs | 250 pcs | 2,500 pcs |
|---|---|---|---|
| Quoted unit price, ex-works | 16.10 | 12.80 | 10.90 |
| Programming and fixture, amortised | 16.80 | 1.68 | 0.17 |
| First-article inspection and documents, amortised | 7.20 | 0.72 | 0.07 |
| In-process inspection and sampling | 0.30 | 0.30 | 0.28 |
| Rework and scrap allowance, 6% of goods value | 0.97 | 0.77 | 0.65 |
| Freight to your dock | 6.92 | 3.74 | 0.37 |
| Duty at 25% of goods value | 4.03 | 3.20 | 2.73 |
| Inbound handling and receiving inspection | 2.60 | 0.48 | 0.19 |
| Landed cost per part, China route | 54.92 | 23.69 | 15.36 |
| Local landed cost per part, same basis | 122.32 | 47.62 | 35.51 |
| Offshore advantage | 55% | 50% | 57% |
Sit with the 25-piece column, because it contradicts the advice you will read everywhere else. Fixed friction there is 26.60 per part: 16.80 of fixture, 7.20 of first-article inspection, 2.60 of receiving, more than the part's own unit price. Add 6.92 of freight, 43 percent of the unit price, and the offshore route still lands at 54.92 against 122.32 locally. That is what a three-to-one rate difference does to a light part. The case for making 25 pieces locally is not money; it is that you can drive over on Tuesday and change the design.
At 2,500 pieces the emphasis moves. The unit price is 71 percent of landed cost, and duty at 2.73 is the second-largest line, bigger than fixture, inspection and freight combined. At volume, negotiate the machine rate and the tariff classification; freight at 37 cents is noise.

Where does the offshore advantage actually disappear?
Two places, and duty is neither of them.
Duty is the cost everyone worries about and the one that matters least here, because a tariff is a percentage applied to a price you already cut by two thirds. Run four scenarios on the 250-piece bracket against the same local landed cost of 47.62.
| Duty rate modelled | Duty per part | Landed per part | Still cheaper than local by |
|---|---|---|---|
| 0% | 0.00 | 20.49 | 57% |
| 10% | 1.28 | 21.77 | 54% |
| 25% | 3.20 | 23.69 | 50% |
| 40% | 5.12 | 25.61 | 46% |
Even at 40 percent the part lands 46 percent below the local quote. That is not an argument for ignoring trade measures, only for putting the real number in the model instead of treating a headline as a veto.
Freight is different, because it scales with mass and there is no percentage to hide behind. Here is where offshore genuinely loses its edge.
Worked example, stated assumptions. A 4140 steel base plate, 12 kg finished from a 13 kg blank, 45 minutes of 3-axis work, one painted finish, 20 pieces. China price built from material at 0.85 per kg, machining at 28 per hour, paint at 4.00, 20 percent margin. Local price from material at 1.90 per kg, machining at 80 per hour, paint at 12.00, 30 percent margin. Air at 270 kg chargeable weight and 5.35 per kg. Sea LCL at 0.7 cubic metres plus documentation, with a minimum charge.
| Line item | Local | China by air | China by sea |
|---|---|---|---|
| Quoted unit price | 134.40 | 43.30 | 43.30 |
| Programming and fixture, over 20 pcs | 9.50 | 3.00 | 3.00 |
| Freight per part | 3.40 | 72.25 | 17.75 |
| Duty at 25% of goods value | 0.00 | 10.83 | 10.83 |
| First-article inspection, over 20 pcs | 1.60 | 9.00 | 9.00 |
| Rework allowance, 3% local and 6% offshore | 4.03 | 2.60 | 2.60 |
| Receiving and inbound handling | 0.60 | 0.60 | 0.60 |
| Landed cost per part | 153.53 | 141.58 | 87.08 |
| Advantage over local | none | 8% | 43% |
A 68 percent unit price advantage became an 8 percent landed advantage. That is inside the noise floor of one rework loop, one week of schedule slip, or one argument about whether a bore is in tolerance. On this part, by air, the local shop is defensible on money as well as on speed, and the buyer who quotes only unit price will never see it.
Sea freight restores the margin at 43 percent and costs about 27 extra calendar days. Whether those days are cheap is the next question, and it has an arithmetic answer.
How much does one rework loop cost you?
A rework loop is not a defect rate. It is a sequence of events from the first measurement to the replacement on your dock, and every step costs money that is not in the unit price.
Worked example, stated assumptions. The 250-piece bracket, goods value 3,200, delivered day 24. At incoming inspection 12 percent of the lot is out on one bore, 0.03 mm over nominal against 0.10 mm of assembly clearance. The shop says the part still assembles. Your drawing says otherwise. That disagreement is the loop.
| Cost line, buyer side | コスト |
|---|---|
| Sorting and re-inspection at your dock, 6 hours | 330 |
| Engineering and purchasing time on the disposition argument, 12 hours | 780 |
| Express re-ship of 30 replacement parts, 14 kg | 210 |
| Re-run first-article inspection on the replacement lot | 180 |
| Buyer-borne total | 1,500 |
| As a share of the order's goods value | 47% |
One loop consumed 47 percent of the value of the entire order. Compare it with the negotiation most buyers spend their time on: pushing the unit price from 12.80 to 12.20 is worth 150 on this order. You need ten clean orders to pay for one loop.
The supplier's side is what makes it drag. Their gross margin on 250 pieces at 12.80 is around 530; remaking 30 parts costs roughly 360, two thirds of the margin on the job. Their rational response is not to remake quickly but to argue that the parts conform, because the argument is cheaper. Nothing about that is malicious; every day of the argument is a day of your schedule.
The cheapest defence is not a lower price or a tighter tolerance. It is a tolerance review before the quote: list the eight dimensions that carry function, price the rest as reference, and put in writing what happens to a part that measures outside. A supplier who has agreed the critical features in advance has given up the argument, and giving up the argument is what makes a remake fast.

What does the calendar look like, day by day?
Lead time is not the sum of the process steps. It is the longest chain, and on machined aluminium the longest chain is almost never the machining. The table is in calendar days, because your customer does not care about weekends.
| ステージ | Local | China by air | China by sea |
|---|---|---|---|
| PO to deposit cleared | 0 | 2 | 2 |
| Material on the floor | 2 | 4 | 4 |
| Programming and fixture | 1 | 2 | 2 |
| First chip to last chip, 250 pcs | 4 | 4 | 4 |
| Anodise, lot queue | 7 | 7 | 7 |
| First-article inspection, documents, packing | 1 | 2 | 2 |
| Transit and customs clearance | 2 | 6 | 32 |
| Inbound and receiving | 1 | 1 | 2 |
| Total calendar days | 18 | 28 | 55 |
Anodising at five business days is seven calendar days, and that single row is longer than every machining operation in the job combined.
The offshore penalty by air is ten calendar days, not the three to five weeks that "shipping from China" implies. Seven of those ten days are finishing and customs, not machining. So below about 14 calendar days local is the only route, and between 18 and 28 days you are trading ten days for roughly half the landed cost.
One smaller effect belongs in your plan: the clarification tax. A question that would be a ten-minute walk to a local machinist becomes a full day offshore, because the answer arrives after your working day ends. Six questions asked one at a time cost six days; the same six in one message cost one.

Which capability differences are real and which are brochure copy?
Almost every published comparison scores this 9 out of 10 for the side paying for the article. This one scores both, from 1 to 5, and flatters neither.
| 寸法 | Local | 中国 | What it decides |
|---|---|---|---|
| 3-axis milling to 0.05 mm, 100 to 2,500 pcs | 4 | 5 | Nothing; both can do it |
| 5-axis simultaneous on a complex part | 3 | 4 | Local availability, not skill |
| 0.01 mm on a 300 mm part, CMM-verified | 4 | 3 | Shop-to-shop variance, not the average |
| Mill-turn and live tooling | 4 | 4 | A wash |
| Inconel, titanium, 45 to 52 HRC steel | 4 | 3 | Tooling cost and depth of experience |
| Material breadth on the shelf | 3 | 5 | Days off the schedule |
| Finishing under the same roof | 2 | 5 | Removes a vendor from your week |
| A design change discussed Thursday | 5 | 2 | The real cost of distance |
| Documented FAI and PPAP discipline | 4 | 3 | Your audit trail |
| A breakdown part for a running line | 5 | 1 | Downtime dwarfs part cost |
| Export-controlled or defence work | 5 | 0 | The only absolute in the table |
| Ten pieces or fewer, at any price | 4 | 2 | Fixed friction, not capability |
| Batch-to-batch repeatability in year two | 4 | 3 | Turnover and fixture retention |
| Landed cost at 250 pieces | 1 | 5 | The reason you are reading this |
| Total out of 70 | 52 | 45 | Meaningless on purpose |
Three rows decide your part. A titanium manifold for a defence programme lands on rows 5, 11 and 10 and never reaches price; a 6061 housing with a powder coat lands on rows 7, 14 and 13 and never reaches proximity.
The brochure rows deserve naming, in both directions. "ISO 9001" certifies that a management system exists; both sides hold one and neither tells you whether your bore is round. "Plus or minus 0.005 mm" is usually a machine specification, not a demonstrated capability, so ask for a capability statement on a feature like yours. "One-stop service" is real in China, because finishing sits in the same industrial park, but it also means your part passes through buildings you have never audited.
Four differences are structural: the labour rate, the 28 against 80 in every table above; the integration of finishing, which removes a logistics step; the proximity of an engineer, worth more than the price difference on a part still moving; and the depth of material stock, worth days. Our own 5-axis そして 精密加工 pages are written around those limits rather than the machine list, because the machine list is the part every supplier gets right.
Where do payment terms and IP risk enter the arithmetic?
Later and smaller than buyers expect, with one exception worth knowing precisely.
A typical China order runs 30 percent deposit and 70 before shipment. On the 2,500-piece bracket that is 8,175 on day zero, and carrying it 55 days at a 10 percent cost of capital costs about 123. Against a landed saving near 47,000, payment terms are not a decision variable; do not let a net-30 offer move a decision worth five figures. The exception is small orders, where wire fees are a cost per transaction: two transfers at 35 to 50 each is 70 to 100 on an order worth 402 in goods, or 17 to 25 percent of it.
What protects money offshore is the release condition, not the term: pay the balance against the dimensional report and packing list, not against a photograph of a pallet. A supplier who knows the final payment depends on a document produces the document.
IP is the same kind of question. An NDA is cheap to sign and slow to enforce across borders, so it changes behaviour at the margin rather than at the extremes. The controls that do real work are operational: strip title blocks, send a simplified STEP with the critical callouts on a separate controlled PDF, issue a uniquely watermarked drawing to each supplier so a leak identifies its source, and split the assembly across two suppliers when the product is the design. If the part is the whole product, that last control matters more than the other three.
The audit side is a checklist, not a feeling. Our China supplier audit checklist was written for moulding but the machine-shop sections transfer directly, and the sourcing guide on quality, cost and IP covers the contractual layer in more depth than this article needs to.
When should you keep the part local, and at what threshold?
This is the table we would use ourselves, and several rows send work away from shops like ours.
| 状態 | Threshold | ルート | The arithmetic behind it |
|---|---|---|---|
| Order size | Under USD 500 and under 20 kg | Local | 220 to 300 of fixed friction is 44 to 60% of the order |
| 時間 | Needed inside 14 calendar days | Local | Offshore air floor is 24 to 28 days with finishing |
| 重さ | Part over 4 kg, shipping by air | Local, or plan sea | Air turns a 68% unit gap into single digits |
| Adders | Duty plus freight over 55% of goods value | Re-quote or local | You are importing freight, not parts |
| 音量 | Under 50 pieces a year of a USD 15 part | Local | The entire annual saving is under 400 |
| コンプライアンス | ITAR, EAR, defence, nuclear, export licence | Local | The only compliant route at any price |
| 幾何学 | 5-axis simultaneous, 0.01 mm, under 30 pcs | Local, or a paid capability audit | The capability exists; finding it is the cost |
| Design maturity | Two or more changes a month | Local for the loop, China once frozen | Every offshore change is a 4-day round trip |
| Urgency | Line-down replacement | Local, always | Downtime dwarfs the part |
| Stable volume | Over 150 pcs a year, under 2 kg, 0.05 mm | 中国 | A 50 to 60% landed advantage that holds |
Read it as two questions: the first five rows ask whether the order can carry the fixed friction, the last five whether the part is stable and light enough to travel.
That is why the hybrid pattern beats either extreme. Iterate locally while the design is still moving, freeze it, then move the frozen revision offshore behind a signed first article. Or split the bill of materials: the two tight-tolerance parts machined twenty miles away, the eight brackets and spacers made in China, assembled at your dock. Buyers who insist on one source pay local prices where it does not matter, or accept offshore lead times where it does.
Three RFQ habits make the comparison real. Ask for a quantity ladder of 25, 250 and 2,500 in one message, because the volume curve is free. Ask for FOB as well as DDP, because a DDP price folds freight and duty into the unit price and makes the comparison unfalsifiable. And state your datum scheme, because the commonest first-article argument is not a dimension out of tolerance; it is two parties measuring from different datums and both being right.
How do you make the second order boring?
The first order is a project. The second order is where the money is, and it is boring exactly to the extent that you set four things up on the first one.
A signed golden sample. Three parts off the first accepted lot, measured and signed by both sides. Every future argument about "as good as last time" then has a physical answer.
A first-article protocol with a fixed content list. Three parts, every dimension measured to your datum scheme, a CMM report rather than hand-tool notes, material certification traceable to the heat number, finish thickness and adhesion on a coupon, and photographs of the packed carton. On the 250-piece order that package costs around 180 and is the cheapest insurance in the chain.
A change rule. Any change to the program revision, the fixture, the material source or a subcontractor triggers a new first article. Not a phone call. That clause prevents most of the slow drift that surfaces eighteen months later as a batch that will not assemble, the same failure mode covered in our piece on cost-control mistakes that cause delays and rework.
We machine aluminium, stainless, tool steel, brass, copper and engineering plastics on 3, 4 and 5-axis mills and on lathes with live tooling, including the CNC milling work that arrives with an anodise attached. When a buyer asks, we quote the landed structure rather than a unit price: send a drawing and your destination postcode and you get both numbers with the assumptions behind them, including the ones that make us look expensive.
Phone and WhatsApp: +86 156 0682 1288 · Email: [email protected]
よくある質問
Is China CNC machining still cheaper once freight and duty are included?
At what order size does a local CNC shop become cheaper?
How long does a China CNC order take door to door?
How do I protect my design when I send CAD files to a Chinese machine shop?
What should a first-article inspection include on a machined part?
著者について

アレックス・ユー
アイデアル・プロ・マニュファクチャリング
アレックスはIdeal Proの共同創業者ですが、重要なプロジェクトの監督や技術チームの指揮をとっていないときは、長距離サイクリングを楽しんだり、街で一番おいしいコーヒーを探し回ったり、次の旅の計画を立てたりしている姿を見かけるでしょう。 国際貿易の営業やプロジェクト管理に20年以上の経験を持つ彼は、厳しい納期や厄介な技術的課題など、ありとあらゆる状況を経験してきました。彼は、素晴らしい仕事は十分な休息と一杯の美味しいコーヒーから生まれると信じており、その好奇心と細部へのこだわりを、手掛けるすべてのプロジェクトに注ぎ込んでいます。 オフィスにいないときは、たいてい屋外で、新しい場所を探索したり、皆と共有できるエピソードを集めたりしています。.





