Why is it still cheaper to manufacture in China, and when isn't it?

The ecosystem, not the wage

Wages in Chinese manufacturing regions have risen for years; the advantage persisted anyway, because it was never mostly about labour. It's density: within a single region you can find the moulder, the toolmaker, the platers and anodisers, the PCB house, the packaging plant and forty component suppliers, often within an hour of each other. That density compresses lead times, crushes logistics friction between process steps, and creates competition at every link. Add deep tooling ecosystems that make moulds faster and cheaper, and scale that amortises everything across enormous volumes, and you get the gap: structural, not cyclical. Automation on the line doesn't dissolve it either: a robot cell still needs moulds, fixtures, components and someone nearby who can fix it, so density keeps paying as the labour share shrinks.

Where the gap is biggest and smallest

The advantage compounds for products that touch many processes: consumer electronics, connected devices, anything with tooling plus electronics plus finishing plus packaging. It shrinks for products that are mostly one process (simple fabrication, moulding you can do anywhere), and it can invert entirely once freight, duty, inventory carrying and coordination overhead join the landed-cost model. A unit price advantage that evaporates in the freight line isn't an advantage; it's a spreadsheet illusion.

When "make it here" wins

Local manufacture earns its case more often than the folklore admits: low volumes where tooling amortisation dominates, heavy or bulky products where freight punishes distance, fast-iteration programs where a supplier an hour away beats one an ocean away, products with regulatory or procurement preferences for local supply, and prototypes, almost always. This is why our position is that China is a capability, not a toll gate: the decision belongs to evidence, product by product, and some of our recommendations are "don't import this one."

The agent-native lens: who runs the comparison

When the buyer sends an assistant. Ask an AI assistant where a product is cheapest to make and it will tend to answer the folk question: wages, headline unit prices, a country ranking stitched together from articles like this one. That's the comparison this article argues against. Brief it with the product instead: the bill of materials, the processes each part needs, the volumes you'll really order and where you sell. Then ask for both landed-cost models side by side, every figure with its source, and a list of what it couldn't price. The other side of that table matters too. A local manufacturer that publishes its processes, capacity and typical lead times plainly gives an assistant something to put in the second column; one that only says "call us" may never make the comparison.

When the business runs on agents. The make-or-import answer stops being a decision made once and filed. Agents can keep both paths current: update freight and duty assumptions from the latest shipments, re-run the comparison when volumes change, gather fresh local quotes on a schedule you set, and flag when the gap narrows enough to matter. What they can't do is price the unquantified lines. Iteration speed, IP posture and how much supply-chain risk you'll carry are judgements, so a person writes them down as explicit weights before the model runs. Moving a product, in either direction, is signed off by someone who owns the margin.

The decision, done properly

Build the landed cost for China, quote the local alternative at your real volumes, and weigh the unquantified lines (iteration speed, IP posture, supply-chain risk appetite) explicitly instead of pretending they're free. Then add a line most comparisons still leave off: supplier data. More customers and regulators now expect a record of materials, origins and who made each part, and a dense ecosystem with many sub-suppliers is cheap to buy from but can take real work to document. Budget for that work rather than discovering it. The cheapest factory is still the one that's cheapest after everything lands, including the things that never appear on a quote.