The ABF substrate pricing cycle is accelerating, with AI demand widening the supply-demand gap

Overview

This report focuses on the ABF substrate cycle within the Greater China technology hardware supply chain. Morgan Stanley believes that, with upward revisions to AI ASIC, GPU, server CPU, and network chip demand, the degree of tightness in ABF supply and demand is intensifying, the pricing upcycle is accelerating, and earnings and valuation for Unimicron, Nan Ya PCB, and Zhen Ding will be revised upward.

Core views

Key views include: first, 2Q profitability was stronger than expected, mainly due to higher utilization, a more favorable product mix, and stronger pricing; second, BT and ABF substrate pricing momentum is continuing to strengthen, with further room for hikes in 2H26; third, upward revisions to demand for Amazon Trainium, Google TPU, Chinese AI chips, Nvidia/AMD GPUs, and Nvidia server CPUs are driving ABF demand; fourth, in the next two years, new capacity is unlikely to be released quickly, so short-term analysis focus will be more on end-demand changes.

Analysis framework

The report updates all three covered companies by combining supply chain work, 2Q operating previews, MSe versus consensus comparisons, a bottom-up ABF substrate supply-demand model, and company earnings forecast and target price valuation models.

Methodology notes

• Supply-demand analysisBottom-up ABF substrate supply-demand model

  ABF substrate supply-demand imbalance

  The report splits end demand into AI ASIC, GPU, server CPU, network chips, and traditional PCs, and includes new capacity assumptions from players such as AT&S, resulting in an assessment of about a 25% ABF substrate supply shortfall in 2030.

• Valuation methodsResidual Income valuation model

  Residual income model

  Unimicron’s target price is based on a residual income model, with key assumptions including 9.2% cost of equity, 1% risk-free rate, 8.7% equity risk premium, 1.0 beta, a mid-term growth rate of 15%, and terminal growth rate of 3%.

• Earnings previewMSe vs. Consensus

  Analyst forecast versus consensus comparison

  The report compares 2Q revenue, gross margin, operating profit, non-operating income, net profit, and EPS, emphasizing that operating performance for all three companies generally benefits from pricing and product mix.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Unimicron (3037.TW)

  Preferred beneficiary, Overweight reaffirmed

  Strengths

  Has a relatively strong supply position in high-end ABF substrates related to AI ASIC and server CPUs, with the target price raised to NT$1,465, implying about 50% upside.

  Weaknesses

  2Q EPS estimate is below consensus, mainly due to more conservative assumptions on non-operating income.

  Comparison

  The report’s preferred play among the three companies.

  Risks

  PC, general server, or AI server demand below expectations; expansion of new capacity; T-glass constraints; CoWoP replacing ABF substrates.

• Nan Ya PCB (8046.TW)

  Beneficiary with stronger repricing leverage, Overweight reaffirmed

  Strengths

  BT and ABF substrate price increases contribute clearly to earnings; the report believes the company has historically been proactive in implementing price hikes, and 2Q EPS estimates are above consensus.

  Weaknesses

  Sensitive to the pricing cycle and tightness in supply-demand; if demand or pricing falls short of expectations, earnings leverage may compress.

  Comparison

  Second in the preference ranking after Unimicron, with a target price of NT$1,550 and implied upside of about 35%.

  Risks

  Demand weakening, industry capex expansion, material constraints, or substitute-technology shocks.

• Zhen Ding (4958.TW)

  Cyclical beneficiary, Overweight reaffirmed

  Strengths

  Revenue and core operating profit benefit from stronger demand, margin improvement, and product mix, with a target price of NT$690.

  Weaknesses

  2Q EPS estimate is about 5% below market consensus, due to more conservative assumptions on non-operating income.

  Comparison

  Has lower implied upside among the three companies, at about 15%.

  Risks

  Non-operating contribution below expectations, demand slowdown, capacity expansion, or ABF substitution risk.

Key data

• NYPCB 2Q BT substrate priceup about 20–30%The report considers this a major contributor to NYPCB’s stronger-than-expected 4Q and 5Q profits.
• ABF substrate 2Q priceup about 10%Supply-chain work suggests ABF price momentum is still strengthening.
• ABF supply-demand gap in 203025% supply shortfallPreviously estimated at around 22%, expanded this time due to revised demand expectations.
• ABF demand growth2025–2030e CAGR 24.4%Above the previous forecast of 22.2%.
• PC-related ABF demand shareabout 70% in 2015, expected below 10% by 2030Shows end demand shifting from PCs to AI, servers, and networking.
• Server, AI GPU, AI ASIC, and networking-related ABF demand shareabout 60% in 2025, expected above 80% by 2030The report views this as the core structural driver of ABF market growth over the next five years.
• Unimicron target priceNT$1,465Raised from NT$1,285 previously, about 14% up, implying roughly 50% upside.
• Nan Ya PCB target priceNT$1,550Implying about 35% upside.
• Zhen Ding target priceNT$690Implying about 15% upside.

Impact & implications

If the report’s thesis holds, the ABF substrate industry will enter a new upcycle led by AI diffusion, with price and margin improvement becoming the main earnings leverage for covered companies. Unimicron is the preferred name due to its stronger supply position in high-end ABF substrates for AI ASIC and server CPUs. NYPCB benefits clearly from a more aggressive pricing strategy and tighter supply conditions, while Zhen Ding also benefits from stronger demand and operating margin improvement, though its upside is relatively lower.

Risks

• PC, general server, and AI server demand below expectations.
• Meaningful industry-wide new capacity expansion that alters the supply-demand tightness landscape.
• T-glass constraints are tighter than expected, affecting capacity and costs.
• Substitution risk from technologies such as CoWoP, weakening long-term demand for ABF substrates.
• Export controls or related legal restrictions could affect certain entities, securities, or supply-chain activities.

What to watch

• Whether ABF and BT substrate price increases can continue to be delivered in 2H26.
• Subsequent quarter changes in gross margin and operating margin for Unimicron, Nan Ya PCB, and Zhen Ding.
• Whether demand forecasts for Amazon Trainium, Google TPU, Chinese AI chips, Nvidia/AMD GPUs, and server CPUs continue to be revised upward.
• The schedule for AT&S Malaysia AI/HPC IC substrate expansion and other new capacity plans.
• The pull from AI training-to-inference diffusion on large-size, high-layer-count ABF substrate demand.