UMC: Mature Node Pricing Power Emerges on AI Spillover and Tight Capacity

Core Conclusion

Mature-node foundry demand is better than feared, granting UMC pricing power for the first time in this cycle and driving material earnings upgrades. We raise our price target to NT$68 and upgrade the rating to Equal-weight from Underweight. However, the stock's recent rally appears to have priced in near-term optimism, warranting a balanced view as execution risks remain.

What The Market May Be Missing

The market underestimates the sustainability of demand from AI's ripple effects into mature nodes and UMC's ability to enact selective price hikes. While not a broad-based super cycle, overflow demand from AI infrastructure (PMICs, networking, controllers) is structurally absorbing weakness in consumer end-markets like smartphones and PCs. This creates a tight supply environment for specific applications at 40/55nm and 8-inch nodes, allowing UMC to exercise pricing power for the first time since the downturn, a dynamic not previously anticipated for 2026.

Evidence Chain

Pricing Power Returning on Improved Utilization. UMC is informing customers of potential wafer price hikes in 2H26 as fab utilization improves. This is a positive pivot from prior expectations of no price increases. We now model a 5-10% hike in 2H26 and another 5-10% in 2027, focused on applications like PMICs and MCUs rather than a broad-based increase. The investment implication is a direct boost to revenue growth and operating leverage, leading to a 20-29% increase in our 2026-28 EPS estimates.

Demand Driven by AI Spillover and Regional Tightness. Mature node utilization is recovering despite consumer electronics weakness due to strong AI-related peripheral demand, as evidenced by TSMC's recent commentary. Furthermore, PMIC capacity is tight across peers like Vanguard, Powerchip, SMIC, and Hua Hong. A concrete signal is UMC's PMIC customer Silergy placing rush orders, primarily for Chinese automotive and industrial demand. This indicates UMC is benefiting from both global AI trends and regional supply tightness, supporting sustained utilization in the high-70% to 80% range.

Earnings Trajectory Revised Upward, Margins Stable. Our revised model forecasts net income rising from NT$44.4bn in 2026e to NT$69.2bn in 2028e. Gross margins are expected to remain stable in the high-20% to low-30% range, as higher wafer prices are offset by rising production costs. The investment takeaway is a more robust and visible earnings recovery path over the next three years, improving the fundamental valuation floor.

Key Divergences & Risks

Price Hike Magnitude and Timing. Market speculation points to a possible 5-10% Q/Q price increase in 2Q26. We believe hikes will be more selective, apply to specific nodes/applications, and occur primarily in 2H26. Consequently, 2Q26 margin and revenue guidance could disappoint elevated expectations, serving as a near-term negative catalyst.

Competitive Capacity Landscape. Bullish narratives may overestimate the speed of capacity rationalization. TSMC clarified that winding down its mature-node fabs (2 and 5) will take several years, implying sufficient industry capacity will remain. A slower-than-expected exit of legacy capacity could cap the pricing power cycle.

Limited Impact from NAND Production. While UMC produces linking dies for 3D NAND, the wafer consumption is minimal compared to commodity memory. Investor perceptions of meaningful exposure to a memory recovery are likely overstated.

Valuation & Trade Implication

Our NT$68 price target is derived from a residual income model, factoring in the upgraded earnings forecast, a 9.2% cost of equity, and a 3.0% terminal growth rate. The stock currently trades near our price target at ~16.1x 2027e P/E and 2.2x 2027e P/BV, close to +1 standard deviation above its historical average. The upgrade to Equal-weight reflects improved fundamentals but balanced against full valuation. Investors should await a more attractive entry point or evidence of price hikes exceeding our conservative assumptions.

Appendix: Data Summary