TSMC, Samsung Cut Mature Chip Output, Price Hikes Loom

Leading Foundries Slash Mature Node Capacity as AI Demand Surges

TSMC and Samsung Electronics — the world's two largest contract chipmakers — are actively reducing production capacity at mature process nodes, setting the stage for a new wave of price increases across the semiconductor industry. According to a May 7 report from market research firm TrendForce, the supply squeeze is already pushing utilization rates sharply higher, with AI-driven demand for power management and power semiconductor chips accelerating the tightening.

The development marks a significant shift in a segment of the chip market that had been struggling with oversupply for much of 2023 and 2024. Unlike the cutting-edge nodes (3nm, 5nm) that dominate headlines, mature process technologies — typically 28nm and above — underpin a vast array of everyday electronics, from automotive controllers to display drivers to industrial sensors.

Key Takeaways at a Glance

• TSMC and Samsung are both reducing capacity at mature process nodes across 8-inch and 12-inch wafer lines
• 8-inch wafer utilization among the top 10 foundries has climbed from roughly 80% to nearly 90%
• AI infrastructure demand for power management ICs (PMICs) and power discrete devices is a primary driver
• Global 8-inch foundry capacity is projected to decline through the first half of 2027
• Some foundries are reallocating 12-inch capacity away from display driver ICs and image sensors toward higher-margin power chips
• TrendForce expects TSMC's 12-inch mature node reductions to proceed gradually over the next 1 to 3 years

Why Mature Process Nodes Still Matter in the AI Era

While the semiconductor industry's spotlight shines brightest on advanced nodes powering GPUs and AI accelerators, mature process technologies remain the backbone of global electronics. These nodes — spanning 28nm, 40nm, 65nm, 90nm, and larger geometries — produce chips that handle essential functions like voltage regulation, signal conversion, motor control, and sensor interfacing.

The current AI boom has paradoxically intensified demand for these older-generation chips. Every AI server packed with cutting-edge GPUs from Nvidia or AMD requires dozens of power management ICs (PMICs) to regulate voltage and current delivery. Edge AI devices, from smart cameras to autonomous vehicles, similarly depend on power semiconductors manufactured at mature nodes.

This dynamic creates a unique supply chain tension: as TSMC and Samsung prioritize their most advanced fabs for AI processors, they are simultaneously pulling resources away from mature lines — just as AI-driven demand for mature-node chips is rising.

8-Inch Wafer Market Faces Sustained Capacity Decline

The 8-inch (200mm) wafer segment is experiencing the most acute supply pressure. TrendForce's data shows that utilization rates among the world's top 10 foundries have rebounded sharply, climbing from approximately 80% at the start of 2025 to nearly 90% currently.

Several structural factors are driving this squeeze:

• TSMC and Samsung are both actively reducing 8-inch production volumes
• No major new 8-inch fab construction is planned globally, as most capital expenditure flows toward 12-inch and advanced-node facilities
• AI server deployments are driving incremental demand for power management and power discrete components
• Edge computing devices require analog and mixed-signal chips predominantly manufactured on 8-inch lines

Perhaps most striking is TrendForce's projection that global 8-inch foundry capacity will continue to decline through the first half of 2027. During this period, the major foundries' 8-inch lines are expected to maintain average utilization rates above 80%, providing a strong floor for pricing.

Compared to the 2022-2023 downturn — when 8-inch utilization at some foundries dipped below 60% — the current trajectory represents a dramatic reversal. This sustained tightening gives foundries significant pricing leverage over customers who have few alternative sources for legacy-node production.

12-Inch Mature Lines See Capacity Reallocation

The 12-inch (300mm) mature process segment presents a more nuanced picture. TSMC is reducing capacity here as well, but TrendForce characterizes the pace as 'moderate' over the next 1 to 3 years, suggesting the company is managing the transition carefully to avoid disrupting major customers.

What is already happening, however, is a significant reallocation of existing 12-inch capacity. Foundries are shifting production away from lower-margin applications — specifically display driver ICs (DDICs) and CMOS image sensors (CIS) — toward higher-value products tied to the AI supply chain:

• Power Management ICs (PMICs) for AI servers and data centers
• BCD (Bipolar-CMOS-DMOS) process chips for industrial and automotive power applications
• Power discrete devices used in voltage regulation modules
• Analog/mixed-signal ICs for edge AI and IoT deployments

This reallocation benefits not only the leading foundries but also second-tier players. As TSMC shifts its capacity mix, customers producing DDICs and CIS chips are seeking alternative foundries, generating a wave of order transfers that boosts utilization at competitors like UMC, GlobalFoundries, SMIC, and Hua Hong Semiconductor.

AI Infrastructure Is the Catalyst Behind the Price Pressure

The fundamental catalyst behind this supply shift is the explosive growth in AI infrastructure spending. Major hyperscalers — including Microsoft, Google, Amazon, and Meta — are collectively investing over $200 billion in AI-related capital expenditure in 2025 alone. Each new AI data center requires not just the headline-grabbing GPUs but also thousands of supporting power management and analog chips.

A single high-performance AI server rack can contain 50 to 100 individual PMICs, each responsible for converting and regulating power delivery to processors, memory modules, and networking components. As AI clusters scale from thousands to hundreds of thousands of GPUs, the aggregate demand for these mature-node chips grows proportionally.

This demand pattern differs fundamentally from the consumer electronics cycle that traditionally drives mature-node volumes. Smartphone and PC demand tends to be cyclical and price-sensitive. AI infrastructure spending, by contrast, is driven by long-term strategic commitments from well-capitalized enterprises, making it more durable and less price-elastic.

For foundries, this means they can command premium pricing for PMIC and power semiconductor production without facing the same pushback they might encounter from consumer electronics customers.

What This Means for the Broader Semiconductor Supply Chain

The emerging price increases at mature nodes carry significant implications across the electronics industry. Companies that depend heavily on mature-process chips — particularly in the automotive, industrial, and consumer electronics sectors — face rising component costs at a time when many are already navigating tariff uncertainties and macroeconomic headwinds.

Key implications include:

• Automotive OEMs may see higher costs for microcontrollers, power chips, and sensor interface ICs
• Consumer electronics brands could face margin pressure on products like TVs, monitors, and IoT devices that rely on DDICs and analog chips
• AI hardware companies will likely absorb PMIC cost increases given the high margins on AI systems
• Second-tier foundries stand to gain market share and improved pricing power from order transfers
• Fabless chip designers may need to redesign products to consolidate functions and reduce chip count

The situation echoes — though at a smaller scale — the 2020-2021 global chip shortage, when mature-node capacity constraints caused widespread production disruptions across automotive and industrial sectors. Industry observers note that the current dynamics are more targeted, affecting specific product categories rather than triggering a broad-based crisis.

Looking Ahead: A Structural Shift, Not a Temporary Blip

TrendForce's analysis suggests the mature-node supply tightening is not a short-lived phenomenon but rather a structural shift in how the world's leading foundries allocate their resources. As TSMC and Samsung continue converting mature-node fab space to advanced processes — or simply retiring older equipment — the global supply of mature-node capacity will contract steadily.

The timeline is telling: with 8-inch capacity projected to decline through at least mid-2027, and TSMC's 12-inch mature reductions extending over 1 to 3 years, customers face a multi-year period of constrained supply. Companies that lock in long-term supply agreements now will be better positioned than those relying on spot-market availability.

For the semiconductor industry at large, this trend reinforces a growing bifurcation. The leading-edge segment races toward 2nm and below, commanding massive capital investments and generating outsized margins. Meanwhile, the mature segment — long considered a commodity business — is quietly regaining pricing power as structural supply reductions collide with AI-driven demand growth.

Investors, supply chain managers, and chip designers would be wise to pay close attention to this often-overlooked corner of the semiconductor market. In the AI era, even 'legacy' chips have become strategically critical — and increasingly expensive.