Export controls on advanced computing are described as bans on selling chips. The operative mechanisms sit further up the supply chain and work quite differently.
The chokepoints are tools, not chips
Manufacturing leading-edge semiconductors requires lithography systems, deposition equipment and metrology tools produced by a very small number of firms worldwide.
Controlling those tools is more effective than controlling finished chips, because a fabrication plant cannot be built or maintained without them and there is no substitute available.
The same logic applies to design software, which is similarly concentrated and equally difficult to replicate without many years of accumulated engineering.
Rules reach beyond the issuing country
Controls commonly apply to foreign-made goods that were produced using the issuing country's technology, which extends their reach well past its own borders.
A plant in a third country using controlled equipment therefore falls within scope, even where no component physically crosses the controlling state's frontier.
This extraterritorial reach is what makes the measures effective and what makes them contentious with allied governments whose firms are affected.
Thresholds require constant revision
Controls are drawn around technical parameters such as processing performance or interconnect bandwidth, since a rule must be objective enough to administer.
Designers respond by producing variants that fall just below the threshold, which is lawful and which prompts regulators to redraw the line.
The result is a repeating cycle of specification and adjustment, with each round increasing complexity for the firms attempting to comply.
Licensing is where policy actually happens
Most controls do not prohibit outright. They require a licence, and the standard applied to licence applications is where the real restrictiveness is set.
A presumption of denial and a presumption of approval produce very different outcomes from identical published rules, and that presumption can change without any rule changing.
Processing capacity matters as well, since a licensing office with limited staff creates delays that function as restrictions regardless of the eventual decision on any individual application.
Effects are gradual and partly self-defeating
Restricted countries respond by investing heavily in domestic capability, which accelerates exactly the substitution the controls were meant to prevent.
Whether that matters depends on the timescale, since building an equivalent equipment industry takes far longer than designing around a single restriction.
Firms in the controlling country also lose sales, which reduces the research budgets that sustain their technological lead, and this trade-off is central to every debate about how broadly to draw the rules.