The semiconductor industry's most visible investments are chasing artificial intelligence and ever-shrinking transistor geometries, but the chips that control a car's window, read a medical sensor, or switch power in a grid inverter do not need the smallest transistors available. Instead, they need to survive harsh operating environments, remain available for years, and cost little enough to deploy by the millions. This is why the industry still depends on mature-node devices, even as the spotlight shines on cutting-edge technology.
'Legacy' describes the age of a manufacturing process, not the usefulness of the component made with it. These chips connect digital systems to the physical world: they regulate voltage, translate analog signals, drive motors, sense temperature and pressure, manage batteries, and execute modest control tasks. A missing part that costs a few dollars can still prevent a vehicle, medical device, or industrial machine worth thousands from shipping.
An older process can be the right technology
There is no universal nanometer boundary between leading-edge, current-generation, and mature chips. The answer changes with the type of device. A December 2025 U.S. Government Accountability Office report classified mature-node products across logic, analog, optoelectronic, sensor, and discrete categories, and noted that demand remains high in automotive and health care as well as long-lived infrastructure and defense systems.
Smaller transistors are valuable when a designer needs to pack more computing into less area or reduce the energy used by dense digital logic. Those gains do not automatically improve a power transistor that must handle substantial voltage, an analog circuit measuring a faint signal, or a sensor built around a particular material. The European Commission's Joint Research Centre found that shrinking some discrete, analog, and optoelectronic devices may deliver little performance or cost benefit; its semiconductor supply-chain study said manufacturing will continue to span everything from sub-5-nanometer logic to processes at 180 nanometers and above.
Mature processes also arrive with years of yield data, design libraries, packaging choices, and field experience. That does not make every old process superior. It means engineers can select for voltage, temperature range, reliability, cost, and supply life instead of paying for transistor density that the application cannot use.
The cost of changing a qualified chip
The purchase price understates the cost of replacement. Automotive integrated circuits, for example, face failure-mechanism and stress testing under the Automotive Electronics Council's AEC-Q100 framework. Changing a part, its process, or its factory can trigger engineering work, software changes, new samples, and requalification. In regulated or safety-critical products, that work may matter more than the saving on each chip.
Suppliers therefore sell continuity as part of the product. In September 2025, STMicroelectronics extended availability for its SPC58 automotive microcontrollers to 20 years, with several lines promised through at least 2038 and one through 2041. Such commitments reflect the commercial life of cars and industrial systems, not nostalgia for an old fabrication method.
Long service lives create a difficult inventory problem. Manufacturers cannot simply redesign every system when a component becomes scarce, yet foundries have little reason to keep lightly used lines open without dependable orders. The result is a market in which an inexpensive, highly specific part can become a production bottleneck long before a sophisticated processor does.
Supply-chain exposure is pervasive but shallow
A U.S. Commerce Department survey captured that mismatch. The Bureau of Industry and Security's mature-node assessment covered 97 end users in aerospace and defense, automotive, consumer products, industry, medical care, and technology. Forty-four percent could not determine whether their products contained chips made by China-based foundries. Based on the information available, 66% of respondents' product sales contained or probably contained at least one such chip.
Yet those China-fabricated chips represented only about 2.8% of the chips by count and 1.3% by value in the surveyed products. BIS called the exposure pervasive but shallow. The average car reported in the survey contained more than 1,700 chips, which explains how a small share of components can appear across most finished vehicles.
Visibility is part of the risk. An equipment maker may buy a module from a supplier, which buys packaged chips from a distributor, while a separate foundry performs the wafer fabrication. The brand printed on a package does not necessarily reveal where the silicon was made. Without that provenance, a company cannot know whether two nominal suppliers depend on the same factory or region.
Price complicates the resilience case. In the same BIS supplier survey, 72% of directly comparable wafers were cheaper at China-based foundries, with a median discount of 10%. For 31% of the products that surveyed suppliers made there, respondents said no alternative foundry was available. Buyers benefit from low prices today, while alternative capacity can become uneconomic before it is needed in a disruption.
New capacity is not only about AI chips
Governments and manufacturers are now funding both ends of the technology spectrum. As of July 2025, the GAO found that the U.S. Commerce Department had awarded $30.9 billion in direct incentives and $5.5 billion in loans across 40 semiconductor projects. Its project inventory includes mature-node logic and analog production at Texas Instruments and a mix of mature logic, analog, sensor, and discrete capacity at GlobalFoundries.
The spending is becoming physical output. Texas Instruments started production at its 300-millimeter SM1 fab in Sherman, Texas, in December 2025. The company says the site will ramp with demand and eventually make tens of millions of analog and embedded-processing chips a day. This expansion underscores that mature-node chips are not a relic but a strategic necessity, ensuring the resilience of the broader semiconductor ecosystem.
