When companies talk about semiconductor risk in 2026, the conversation usually begins with AI: advanced processors, high-bandwidth memory, leading-edge fabs, advanced packaging. Those markets deserve the attention, but they are not the only ones getting tighter. The controllers, power-management chips, analog devices and other relatively mature components embedded throughout industrial equipment are experiencing their own pricing pressure, and in some cases AI is contributing to it indirectly. TrendForce reported in June that foundry prices for 8-inch mature-node production increased 5% to 15% between the first and second quarters of 2026, with average utilization among the world's 10 largest 8-inch foundries recovering to 88% and expected to reach 90% during the second half of the year. The reason is not a simple resurgence in demand for old chips. Foundries are deciding which products deserve limited manufacturing capacity, and that changes the economics for industrial buyers.
Mature Does Not Mean Easy to Replace
Mature-node semiconductors generally lack the performance specifications that dominate technology headlines. What they do is perform established functions reliably and repeatedly: power-management integrated circuits regulate electricity, microcontrollers execute control functions, analog chips translate physical inputs into electronic signals, and power semiconductors manage motors and other electrical loads. They can be found inside factory automation, HVAC systems, power supplies, drives, medical equipment, vehicles and countless other systems, often in products designed around that specific component years ago. Replacing it can require engineering work, testing and qualification, and depending on the application, changing a component can also affect certification, reliability testing or the performance of the larger system. That means a chip costing only a few dollars can carry considerably more commercial leverage than its purchase price suggests, a dynamic distinct from the broader capacity-timeline risk already documented across infrastructure projects: this is a pricing story, not primarily an availability one.
AI Is Competing for More Than Leading-Edge Capacity
TrendForce says increasing demand for AI servers, general-purpose servers and edge AI products is consuming mature-node capacity through components such as power-management ICs, power discrete devices, interposers and other supporting technologies, the same demand surge already reshaping how AI infrastructure buildouts strain the supply chains beneath them. At the same time, TSMC and Samsung have been cutting 8-inch capacity, directing limited production toward higher-margin AI-related products rather than lower-margin applications. The result is not necessarily an industrywide shortage of mature chips. It is competition for particular processes and production lines, which means a procurement team can look at total semiconductor capacity and see adequate supply while the process needed for one qualified industrial component is getting tighter.
The Cost Pressure Can Arrive Before the Shortage
This is where the current cycle differs from the disruption industrial companies experienced earlier in the decade. The first signal does not have to be a 40-week lead time or a production shutdown. It can be price: TrendForce expects the mature-node pricing trend to continue into 2027, and for some capacity-constrained 12-inch processes, prices were already showing increases of 5% to 10% between the second and third quarters of 2026. A facility may not buy microcontrollers or power ICs directly; it buys a variable-frequency drive, control panel, power supply, sensor system or piece of automation equipment from a manufacturer that does. The chip-price increase can therefore arrive as a higher equipment quote, a shorter price-validity period, a component substitution, a surcharge or a more expensive maintenance part, which makes semiconductor inflation harder to isolate inside an industrial budget.
Tariff Exposure Adds a Separate Layer
Trade policy complicates the picture further, and here too the distinction between advanced and mature semiconductors matters. In January, the administration imposed a 25% Section 232 tariff on a narrow category of advanced computing chips, but the proclamation specifically exempted qualifying covered chips imported for non-data-center civil industrial applications, among other uses. Then on August 6, the White House announced a separate Section 232 action covering polysilicon and its derivatives: beginning December 4, certain downstream products including solar wafers, cells and modules face an additional 15% duty alongside a new minimum-import-price program, though raw polysilicon and the mature industrial chips discussed here fall outside its scope. None of that means every industrial semiconductor is suddenly subject to a new tariff, but it does mean procurement teams are operating in an environment where semiconductor costs can be influenced by several layers at once, foundry pricing, upstream material costs, country of origin and changing trade treatment, and the cost that reaches an industrial customer may bear little resemblance to any single headline rate, a variability already showing up in how tariff pass-through clauses are being tested in other equipment-heavy contracts.
Procurement Needs to Know What Is Inside the Equipment
That creates a different question for industrial procurement. The question in the last supply crisis was often whether the chip could be obtained at all. The question now is increasingly how exposed the equipment being purchased is to chips whose economics are changing, which requires visibility below the finished product: which controls and power systems depend on mature-node components, which are single-sourced, which have qualified alternatives, and which suppliers are already seeing foundry increases move through their bills of materials. Those questions matter most for equipment expected to remain in service for 10, 15 or 20 years. A leading-edge processor may determine what the next generation of artificial intelligence can do. A much older semiconductor can determine whether a motor starts, a production line runs or a replacement control board remains economical. Those chips rarely attract the same attention. They only need to become more expensive.
