For three decades the organising principle of manufacturing was to place each step wherever it was cheapest.
The resulting network was efficient, and almost invisible to the consumer.
Two shocks made it visible, and the political response has outlasted the disruption that prompted it.
Governments now speak of , friend-shoring and strategic autonomy, terms that describe an intention rather than a policy.
The data so far show at the margin rather than the reversal that the language implies.
Final assembly has moved; the components feeding that assembly frequently have not.
A country that imports and exports finished goods appears in the statistics as diversification without being it.
Measuring dependence therefore requires tracing rather than counting shipments.
On that measure concentration has fallen only slightly, and in some critical inputs it has increased.
The explanation is not political reluctance but industrial reality.
A supplier network takes a decade to build and depends on that cannot be shipped with a machine.
Subsidies can pay for a factory; they cannot rapidly produce the four hundred firms that make it viable.
There is also a cost that industrial policy documents rarely quantify.
capacity across blocs raises prices for everybody and lowers the return on each plant.
Whether that is worth paying is a judgement about probability, not a technical calculation.
Framed honestly, the question is what insurance a state wishes to buy and at what annual cost.
Framed dishonestly, it becomes a promise that supply chains can be made both cheap and safe.
The second framing wins elections, which is why the first appears mainly in the footnotes.