Foundry Mature-Node Prices Set to Rise on AI Chip Demand

Mature-node wafer foundry prices are poised for a significant rebound starting in late 2025, as major chipmakers like TSMC and Samsung scale back 8-inch wafer production while demand for AI-adjacent semiconductors — particularly power management ICs and power devices — surges to unprecedented levels. Industry analysts project that average 8-inch capacity utilization across the world's top 10 foundries will climb back to nearly 90% by 2026, marking a decisive end to the pricing downturn that has plagued the mature-process segment.

The shift is being driven by a dual dynamic: deliberate capacity reductions at the world's largest contract chipmakers and a structural increase in demand for the analog and mixed-signal chips that keep AI servers running. The implications ripple across the entire semiconductor supply chain, from Tier 2 foundries eyeing price hikes to system designers budgeting for higher component costs.

Key Takeaways

• TSMC and Samsung are reducing 8-inch wafer capacity from H2 2025 onward
• AI server growth is driving sharp increases in power management IC and power device demand
• Top 10 global foundry 8-inch utilization expected to recover to ~90% by 2026
• 8-inch pricing is bottoming out and set for a rebound
• 12-inch mature-node orders may shift to competitors as TSMC prioritizes advanced nodes
• Tier 2 foundries could signal price increases to customers by H2 2026

TSMC and Samsung Pull Back on 8-Inch Production

The global semiconductor industry is witnessing a strategic pivot at its highest levels. TSMC, the world's largest contract chipmaker, and Samsung Foundry, its closest rival, are both scaling down their legacy 8-inch (200mm) wafer fabrication capacity beginning in the second half of 2025. This move reflects a broader industry trend of reallocating resources toward more profitable advanced-node production — particularly the 3nm and 2nm processes that power cutting-edge AI accelerators and mobile processors.

For TSMC, the calculus is straightforward. With AI chip customers like NVIDIA, AMD, and Apple consuming massive volumes of leading-edge capacity, the company has strong financial incentives to shift capital expenditure away from mature nodes. Samsung faces similar pressures as it battles to win advanced-node market share.

The result is a tightening supply picture for mature-process wafers. Unlike the 2022-2024 period — when overcapacity and weak consumer electronics demand pushed utilization rates below 70% at some facilities — the market is now approaching a meaningful inflection point. Analysts estimate that global 8-inch utilization had bottomed out near 75-80% in early 2025 before beginning its recovery trajectory.

AI Servers Fuel Explosive Demand for Power Management Chips

While much of the AI semiconductor conversation focuses on GPUs and custom accelerators built on cutting-edge nodes, a less visible but equally critical demand wave is building in the mature-node segment. Every AI server rack requires dozens of power management ICs (PMICs), voltage regulators, power MOSFETs, and other analog components — nearly all of which are manufactured on 8-inch or 12-inch mature processes ranging from 28nm to 180nm.

The numbers tell a compelling story:

• A single NVIDIA HGP B200 server platform requires significantly more power delivery components than a traditional enterprise server
• Global AI server shipments are projected to grow 40-50% year-over-year through 2026
• Power management IC revenue tied to data center applications is expected to exceed $8 billion by 2026
• Each new generation of AI accelerators demands higher wattage, driving more PMIC content per server

This demand growth is not limited to server applications. The proliferation of AI-capable edge devices, smart vehicles with advanced driver-assistance systems, and industrial automation platforms all contribute to rising consumption of mature-node semiconductors. The automotive sector alone continues to absorb growing volumes of power devices and microcontrollers fabricated on 8-inch lines.

12-Inch Mature Nodes See Order Spillover Effects

The supply dynamics extend beyond the 8-inch segment. As TSMC reduces its focus on mature processes across both 8-inch and 12-inch (300mm) lines, a significant order transfer — or 'spillover' — effect is beginning to materialize. Customers who previously relied on TSMC for 28nm, 40nm, or 55nm production are increasingly evaluating alternative foundries.

This trend benefits several Tier 2 foundries positioned to absorb redirected orders:

• GlobalFoundries (U.S.-based, strong in automotive and IoT)
• UMC (Taiwan-based, significant 28nm capacity)
• SMIC (China-based, expanding mature-node output)
• VIS (Vanguard International Semiconductor) (TSMC affiliate, 8-inch specialist)
• HuaHong Group (China-based, power device focus)
• Tower Semiconductor (Intel-owned, analog and specialty processes)

While current market conditions do not yet reflect an outright supply shortage, the trajectory is clear. Industry sources indicate that several Tier 2 foundries are already in preliminary discussions with customers about potential price adjustments for the second half of 2026. The messaging is cautious but deliberate: foundries want customers to prepare for a higher-cost environment.

Pricing Dynamics: From Downturn to Recovery

The mature-node foundry market has endured a prolonged pricing downturn since mid-2022. After the pandemic-era chip shortage drove prices to historic highs — with some foundries imposing 15-20% surcharges — the subsequent demand correction sent prices tumbling. By late 2024, aggressive discounting had become common as foundries competed for a shrinking pool of orders.

That era now appears to be ending. The convergence of reduced supply and rising AI-driven demand is creating conditions for a sustained price recovery. Industry analysts point to several factors supporting this view:

First, the cost structure of operating legacy 8-inch fabs is rising. Older equipment requires more maintenance, and the labor and utility costs associated with running these facilities continue to climb. Foundries need higher average selling prices simply to maintain margins.

Second, the opportunity cost of mature-node production is increasing. Every square meter of cleanroom space dedicated to 8-inch production represents capacity that could theoretically be converted or repurposed. This gives foundries leverage in price negotiations.

Third, the demand floor established by AI infrastructure buildout provides foundries with confidence that volume commitments will hold. Unlike consumer electronics cycles that swing wildly with smartphone launches, data center power management demand follows multi-year infrastructure investment cycles.

What This Means for the Industry

The ripple effects of mature-node price increases will be felt across multiple sectors. For AI infrastructure companies — including hyperscalers like Microsoft, Google, Amazon, and Meta — higher component costs for power management and analog ICs will add to the already enormous capital expenditure required to build out AI data centers. While these costs represent a small fraction of total server BOM (bill of materials), they compound across millions of units.

For fabless semiconductor companies designing power management and analog products, the pricing shift creates both challenges and opportunities. Companies with diversified foundry partnerships and long-term supply agreements will be better positioned than those relying on spot-market capacity. Firms like Texas Instruments, Analog Devices, and Infineon — which operate their own fabs — may gain competitive advantages over purely fabless rivals.

System designers and procurement teams should begin planning for a 5-10% increase in mature-node component costs over the 2026-2027 timeframe. Dual-sourcing strategies and early engagement with Tier 2 foundries could help mitigate supply risk.

Looking Ahead: The 2026 Inflection Point

The semiconductor industry is approaching a structural shift in mature-node economics. The convergence of TSMC and Samsung capacity reductions, surging AI-related demand, and rising fab operating costs points toward a market that will look fundamentally different by the second half of 2026.

Several key milestones to watch include:

• Q3-Q4 2025: TSMC and Samsung begin executing 8-inch capacity reductions
• H1 2026: Global 8-inch utilization crosses 85%, pricing stabilizes
• H2 2026: Tier 2 foundries formally communicate price increase intentions
• 2027: Full impact of price increases reflected in downstream component pricing

The broader takeaway is that AI's appetite for semiconductors extends far beyond the headline-grabbing advanced nodes. The humble power management IC — fabricated on decades-old process technology — has become a critical enabler of the AI revolution. As the industry recognizes this reality, the economics of mature-node manufacturing are being repriced accordingly.

For companies building AI infrastructure, the message is clear: secure your mature-node supply chain now, before the pricing window closes.