September 9th, 2026 — The semiconductor foundry market is entering a new pricing cycle, with capacity tightening across both advanced and mature process nodes. What began as strong demand for leading-edge AI computing chips is increasingly affecting power management, analog, high-voltage and other specialty semiconductor processes. Industry data indicates that utilization at major 8-inch foundries has recovered to nearly 90% in 2026, while several foundries have already implemented wafer price increases.
The result is an increasingly tight supply environment extending from advanced nodes such as 4nm and 5nm all the way to mature 8-inch processes used for PMICs, BCD, analog and power semiconductor devices. For fabless semiconductor companies and ASIC developers, the implication is becoming clear: meaningful foundry price relief is unlikely before 2027.
Samsung Electronics has reportedly increased pricing for several advanced process technologies during 2026. Pricing for some new Samsung 4nm orders increased by approximately 10%–15%, depending on the customer and region. Similar increases have been reported for 5nm production, while 8nm pricing has risen by close to 10%. TSMC is also reportedly preparing broader pricing adjustments beginning in 2027.
Base wafer pricing for process nodes of 7nm and below could increase approximately 5%–10%. Customers requesting additional high-performance computing capacity may face further premiums of approximately 10%–15%. Under certain circumstances, this could make incremental HPC capacity substantially more expensive than today’s standard wafer pricing.
Importantly, the pricing pressure is not limited to advanced processes. TSMC’s 12nm, 16nm and 28nm processes could reportedly see increases of as much as approximately 10%.
| Foundry | Process / Segment | Reported Pricing Development |
|---|---|---|
| Samsung | 4nm / 5nm | Some new orders increased approximately 10%–15% |
| Samsung | 8nm | Pricing increased close to 10% |
| TSMC | 7nm and below | Reported 5%–10% base increase planned for 2027 |
| TSMC | Additional HPC capacity | Potential additional 10%–15% premium |
| TSMC | 12nm / 16nm / 28nm | Potential increases of up to approximately 10% |
| VIS | Mature nodes | Pricing adjusted from April 2026 |
| UMC | Mature nodes | Price increases introduced for H2 2026 |
| Nexchip | Foundry services | Approximately 10% increase from June 2026 |
| PSMC | 8-inch / 12-inch logic | Approximately 10%–15% increase |
| PSMC | Memory foundry | Approximately 45% increase reported from July |
Perhaps the more important development for the wider semiconductor industry is what is happening at mature nodes. For several years, semiconductor companies became accustomed to mature-node capacity gradually becoming easier to obtain following the severe shortages experienced earlier in the decade. That situation appears to be changing.
Average utilization among the world’s leading 8-inch foundries has reportedly recovered to nearly 90% during 2026.
Mature-node foundry pricing increased approximately 5%–15% between the first and second quarters of 2026, and upward pressure could continue into 2027.
Several factors are contributing simultaneously:
→ AI infrastructure is increasing demand for analog and power-management ICs.
→ Some major foundries have reduced or reallocated 8-inch capacity.
→ Capacity is moving away from weaker applications toward PMIC, BCD and power semiconductor products.
→ Automotive and industrial semiconductor demand remains significant.
→ Specialty processes cannot easily be transferred between foundries.
This combination makes the current capacity situation different from a temporary shortage caused by one particularly strong semiconductor segment.
The rapid expansion of AI data centers is normally associated with leading-edge logic processes. However, an AI server requires considerably more than GPUs and CPUs.
AI infrastructure also contains large numbers of:
→ Power-management ICs
→ Voltage regulators
→ Gate drivers
→ Analog ICs
→ Interface devices
→ Power discretes
→ Memory-related devices
→ Monitoring and protection ICs
Many of these devices are manufactured using process technologies ranging from approximately 180nm to 45nm, including BCD and other specialty platforms. SMIC has highlighted the scale of this effect. According to the foundry, an AI rack containing 72 GPUs can require more than 16,000 power-management devices. This creates an important multiplier effect.
Increasing GPU shipments do not only create demand for advanced wafers. Every additional AI accelerator can indirectly generate demand for numerous mature-node semiconductor devices.
One of the areas benefiting from this trend is BCD technology. BCD processes integrate bipolar, CMOS and DMOS technologies and are widely used for power-management and high-voltage applications.
The move toward 48V power architectures in AI data centers is particularly relevant.
Higher-power AI servers require increasingly sophisticated power conversion and distribution systems, supporting demand for high-current BCD processes and related power semiconductor technologies. Other specialty technologies experiencing similar pressures include:
→ High-voltage CMOS
→ RF SOI
→ Embedded non-volatile memory
→ Analog mixed-signal processes
→ Power discrete technologies
Unlike standard CMOS logic, these processes can be difficult to migrate from one foundry to another. Products may depend on foundry-specific devices, design rules, IP, models and manufacturing characteristics.
Adding mature-node capacity may appear easier than building a leading-edge semiconductor fab, but specialty manufacturing still has significant barriers. BCD, analog, high-voltage and other specialty processes often require long qualification and yield-learning cycles. New manufacturing capacity can take two to three years to become fully operational. Automotive applications create another constraint.
Once an automotive semiconductor has been qualified at a particular foundry, moving the product can involve redesign, engineering validation, reliability qualification and customer re-certification. Consequently, semiconductor companies frequently prefer accepting higher wafer prices rather than transferring production. This reduces the ability of the market to respond quickly when capacity becomes constrained.
Chinese foundries are also becoming increasingly important within the mature-node supply landscape. SMIC reported second-quarter 2026 revenue of approximately US$3.0 billion, representing growth of 36.1% year over year and 20% sequentially. Gross margin reached 25.3%, while capacity utilization remained high.
Hua Hong Semiconductor reported quarterly revenue of approximately US$717.5 million, with utilization reaching 102.8%.
| Indicator | SMIC Q2 2026 | Hua Hong Q2 2026 |
|---|---|---|
| Revenue | US$3.006 billion | US$717.5 million |
| YoY revenue growth | +36.1% | +26.8% |
| Gross margin | 25.3% | 16.5% |
| Capacity environment | High utilization | 102.8% utilization |
As Taiwanese foundries prioritize higher-value applications and capacity becomes increasingly constrained, some customers are moving additional wafer starts toward Chinese foundries to secure supply.
Foundry price increases are beginning to propagate through the semiconductor supply chain. A number of major analog, power and mixed-signal semiconductor manufacturers have implemented selective price adjustments during 2026. The industry therefore appears to be moving through a chain reaction:
AI infrastructure growth → higher advanced-node demand → increased power and analog demand → tighter specialty capacity → higher wafer prices → higher semiconductor component prices.
This is particularly important for customers purchasing power-management, analog and industrial semiconductor devices, where mature-node wafer costs represent a meaningful part of overall product economics.
For fabless semiconductor companies preparing products for 2027 and 2028, foundry selection may increasingly become a capacity decision rather than simply a wafer-price decision. Companies planning new tape-outs should consider several factors earlier in the development cycle:
→ Capacity commitments: Confirm whether sufficient production capacity will be available when the product reaches volume production.
→ Long-term wafer pricing: Evaluate pricing over several years rather than focusing only on initial engineering wafer costs.
→ Second-source strategy: Determine whether the product architecture allows qualification at another foundry.
→ Specialty process availability: BCD, HV, RF and embedded-memory capacity may become more constrained than standard CMOS.
→ Packaging capacity: Advanced packaging and high-performance substrates can create additional bottlenecks even when wafer capacity is available.
→ Forecast accuracy: Foundries are likely to prioritize customers providing credible long-term demand forecasts and capacity commitments.
→ Supply agreements: Strategic capacity agreements may become increasingly important for high-volume programs.
The semiconductor market does not appear to be returning to exactly the same shortage conditions experienced earlier in the decade. Instead, a more selective capacity imbalance is developing. At one end of the market, AI accelerators are consuming advanced-node capacity. At the other, the enormous number of power-management, analog and specialty devices surrounding those accelerators is increasing demand for mature-node production.
Meanwhile, automotive, industrial and consumer semiconductor demand continues to compete for many of the same specialty processes. This creates pressure at both ends of the semiconductor manufacturing spectrum.
With 8-inch utilization approaching 90%, foundries raising prices and new specialty capacity requiring years to bring online, semiconductor companies should not assume that today’s wafer pricing will remain available through 2027.
For ASIC and fabless companies, securing the right process technology, foundry capacity and long-term supply arrangement may become just as important as negotiating the lowest wafer price.