- What happened: Taiwanese IC design companies are reportedly considering price increases from late 2026 to early 2027, with the size of the hikes said to range from 5% to double digits.
- Why it matters: Mature-node capacity is being diverted to AI-related components, and higher manufacturing costs could spill over to general-purpose ICs.
- What to watch: The formal terms of any price revisions, demand once front-loaded ordering runs its course, and how much new-fab capacity is actually available to use.
Taiwanese IC design companies are considering a new round of price increases from late 2026 to early 2027. According to an October 5 report by Taiwan's Economic Daily News, the products named include power management ICs, display driver ICs, and microcontrollers, with hikes ranging from 5% to double digits. However, no company has formally announced a price revision; for now, this is an outlook based on the newspaper's reporting. The background is rising foundry manufacturing fees, but the burden on actual customers will differ depending on when they signed their contracts and whether they can shift production to other fabs.
After foundry fees, IC prices could rise too
The Economic Daily News report describes the revisions IC designers are weighing as a second round of price increases. For companies that design circuits and outsource manufacturing to foundries, higher fees push up product costs. Even if inexpensive wafers or existing contracts cushion the impact for a while, once fees rise on the next orders, the question becomes how much to pass on in selling prices.
The paper also reported that TSMC's mature-node foundry fees are expected to rise 3% to 10% in January 2027. However, TSMC has not disclosed specific revisions and has said it sets prices strategically. Other Taiwanese foundries have not commented on the expected 2027 increases. All of this remains at the reporting stage, and it has not been decided that foundry fees and IC selling prices will each rise by the reported ranges.
A 10% increase in foundry fees does not mean IC selling prices will rise a uniform 10%. Product costs also include packaging, assembly, and testing, and the manufacturing cost per IC changes with the number of good dies obtained from each wafer. Selling prices also depend on each product's margin and on negotiations with customers, so it is hard to judge from the reported figures alone how much prices for home appliances or automobiles will rise.
AI-related demand spreads to mature nodes
Research firms have also pointed to the possibility that mature-node price increases will continue. TrendForce's independent research, published on June 30, forecast that mature-node foundry fees would keep rising in 2027, against a backdrop of capacity being redirected to AI-related components and some manufacturers cutting production capacity. AI-related demand is not concentrated only at fabs producing leading-edge compute chips.
Power management ICs regulate voltage and current to supply power to each component. Growing power demand from AI servers also feeds demand for these ICs and for power semiconductors. In addition, some interposers, the intermediate substrates that connect multiple chips, are manufactured on mature nodes. TrendForce reports that demand is concentrating on processes such as 55nm interposer manufacturing on 12-inch wafers, while capacity for lower-margin image sensors and display driver ICs is shrinking.
This reallocation of capacity affects IC design companies that do not make AI products. Because they still need to secure capacity at the same fabs, their manufacturing fees can rise even if demand for their own products is not surging. According to the research, at fabs using 8-inch wafers, growing demand for power management ICs and power semiconductors coincided with capacity reductions and reallocations by TSMC and Samsung. Older process generations no longer necessarily have spare capacity.
Foundry earnings also include figures that show how strong current demand is. According to UMC's earnings release of July 29, company-wide utilization in the second quarter was 85%, and it expected over 90% in the third quarter. The company explained that demand is strong for power management ICs and microcontrollers made on 8-inch wafers, and it raised its 2026 capital expenditure plan to $2 billion. Note, however, that 85% is not the utilization rate of 8-inch fabs alone, and the 90%-plus figure was an outlook as of the July announcement. These are not formal price revision announcements, but they hint at why negotiations over capacity are getting tougher.
Long-term contracts and manufacturing options change customers' burden
Even for the same IC, the timing and treatment of price increases differ by customer. A microcontroller supplier interviewed by Economic Daily News said major customers who have placed long-term orders will not be subject to this additional increase. However, it said 2027 pricing has already been negotiated, and customers had accepted the increases. This means no additional hike, not that previous prices are being held.
Even if customers try to switch foundries to cut costs, the fabs they can choose from are limited. According to a related report by the same paper, customers' procurement conditions requiring manufacturing to avoid China constrain IC designers' choice of foundry. This constraint is not a blanket ban applied uniformly to all products; it stems from individual customers' sourcing requirements. Switching to a U.S. foundry will not necessarily be cheaper either, and changing foundries alone may not resolve the cost increase.
Where customers do allow manufacturing in China, a decision cannot rest on the destination fab's quote alone. According to the paper, major Chinese fabs are also raising prices, and moving production to another fab can lower yields and raise the cost per IC. Yield refers to the share of manufactured products that are usable good units. Even at the same manufacturing cost, fewer good units mean a higher cost per sellable IC.
Just because they are all mature-node fabs does not mean a product already in volume production can be made at the same quality and cost immediately after a transfer. Customers need to consider the regions they permit for manufacturing, the yield after transfer, and the capacity they have already secured. In these price negotiations, whether stable volume production is possible at the destination fab, not only the fees a fab quotes, shapes customers' options.
When will new fabs add capacity, and how much?
VSMC, a joint venture between VIS and NXP, plans to begin volume production at its new Singapore fab in the first quarter of 2027. The target of securing monthly capacity of 44,000 12-inch wafers, however, is set for 2029. NXP has the right to use 40% of the fab's capacity. Based on the fab opening announcement of September 28 and NXP's 2025 annual report, three points need to be distinguished when estimating the supply increase from the new fab.
| Point to check | What has been disclosed | Caveat for assessing supply and demand |
|---|---|---|
| Start of volume production | Planned for Q1 2027 | Monthly capacity at the start of production has not been disclosed |
| Capacity expansion target | Target of 44,000 12-inch wafers per month in 2029 | This capacity will not be usable from early 2027 |
| Rights to use capacity | NXP holds the right to use 40% of the fab's capacity | The contract status and open capacity for the remaining 60% cannot be determined from this information |
As of the September announcement, the new fab had already begun trial production. It plans to use 130nm to 40nm processes to make power-related semiconductors, analog semiconductors, and other products. The plan could add supply in the very segments where price increases are being reported, but there is a time lag between the fab's opening, the start of volume production, and the expansion of capacity.
Furthermore, at a joint-venture fab, not all capacity is necessarily opened to external customers. NXP's usage rights indicate that part of the added capacity will support its own procurement. At the same time, nothing in these documents suggests that all the remaining capacity is already under contract. When assessing 2027 supply and demand, the published capacity target cannot simply be treated as open capacity available to outside customers.
Gauging demand after front-loaded ordering runs its course
Rising orders at present do not necessarily mean consumption is recovering. According to a related Economic Daily News report, customers are front-loading orders out of concern over future price increases and supply shortages, but the IC design companies interviewed do not see a broad demand recovery. If orders are rising because customers are securing future purchases early, subsequent orders could fall.
The recovery in fab utilization shown by UMC and the IC designers' cautious view of end demand are not necessarily contradictory. Utilization can rise as capacity is redirected to AI-related components, while for parts used in consumer products, front-loaded orders can build up customer inventories. The paper also reports that IC designers are comparing orders with the previous year and are wary of over-ordering and rising customer inventories.
To gauge price pressure in early 2027, it will be necessary to look at the price revision terms IC design companies actually present together with order trends after front-loaded ordering runs its course. Even if orders fall, some customers will find it hard to avoid cost increases if the fabs they can move production to are limited. Conversely, if volume production at new fabs progresses and the capacity actually available grows, the conditions for price negotiations would change. How far the increases spread will depend not only on the strength of demand but also on who can use how much capacity, and when.