ABF substrate‑in‑package substrates: The supply‑demand gap is widening, and we are raising the target prices for the three major players.

Overview

Morgan Stanley’s report highlights that the ABF substrate market is entering the early stages of a new upward cycle. Driven by robust demand from AI servers, GPUs, and ASICs, the firm has revised upward its forecast for the ABF substrate supply‑demand gap over the next few years, projecting a 22% shortfall by 2030—up from the previous estimate of 15%. In light of this, the research note has raised earnings estimates and target prices for Unimicron, NYPCB, and ZDT, while upgrading ZDT’s rating from “Equal Weight” to “Overweight.”

Core views

Structural shifts on the demand side are the core driving force. According to the research report, the demand mix for ABF substrates is undergoing a fundamental transformation: the share attributable to the PC market has plummeted from roughly 70% in 2015 to 24% by 2025 and is expected to fall below 15% by 2030. Meanwhile, demand for ABF substrates tied to servers, AI GPUs, AI ASICs, and networking equipment is surging, accounting for 60% of total demand by 2025 and projected to exceed 75% by 2030. This shift toward higher‑value, larger‑size, multi‑layer substrates not only increases the number of substrates required per chip but also raises technological barriers. The supply‑demand balance is tightening, bolstering pricing power. The latest bottom-up model indicates that by 2030, the supply gap for ABF substrates will widen to 22%. Although Unimicron has boosted capacity by approximately 40% through bottleneck‑elimination measures as of 2026, this will only provide short‑term relief; new large‑scale capacity expansions typically take more than two years to come online. Consequently, supply constraints from 2027 to 2030 appear unavoidable. Such shortages will empower suppliers with stronger pricing leverage: the report forecasts ABF substrate prices to rise 15–20% year over year in 2026, more than 20% in 2027, and potentially 25–40% in 2028. The ecosystem is expanding, benefiting second‑tier players. With supply tight, end customers are diversifying their supplier base, creating opportunities for smaller vendors and new entrants. For instance, Kinsus is developing an expanded portfolio of substrates for Nvidia, while Zhen Ding Technology (ZDT) is actively pursuing certification for Google TPU substrates. ZDT’s Shenzhen plant achieved its first quarterly profit in Q1 2026, with utilization climbing above 60%, and the company now holds a pivotal position in China’s AI‑chip substrate market. Additionally, ZDT maintains high‑margin operations in optical‑module PCBs, which are poised to become a key driver of earnings. Specific stock‑selection rationale: 1. Unimicron (3037.TW): As a leading supplier of substrates for AI ASICs and server CPUs, the company stands to benefit from robust demand in the premium segment. EPS estimates for 2026–2028 have been revised upward by 19%, 21%, and 41%, respectively, with the target price raised to NT$1,225. 2. Nanya Circuit Board (8046.TW): Dominant in the network‑chip substrate space and pursuing an aggressive pricing strategy, Nanya’s EPS forecasts for 2026–2028 have been upgraded by 37%, 59%, and 81%, respectively, with the target price lifted to NT$1,275. 3. Zhen Ding Technology (4958.TW): In addition to being a critical substrate supplier for Chinese AI chips, ZDT is expanding into international markets—such as Google TPU—and venturing into AI‑server PCBs. Given its growth potential and attractive valuation, the firm’s rating has been upgraded to Outperform, with the target price significantly raised to NT$570.

Analysis framework

The research report employs a classic “supply-and-demand framework” combined with a bottom-up modeling approach. First, by disaggregating end-use applications (PCs vs. AI/Server), it identifies the long-term trend in demand structure: a shift from PC‑centric to AI/HPC‑driven. Second, it constructs a detailed supply-and-demand model that incorporates the latest demand forecasts from major hyperscalers for ASICs (e.g., Amazon Trainium, Google TPU) and CPUs, as well as growth expectations for China’s AI chip market, thereby quantifying future aggregate demand. Third, it assesses supply-side capacity expansion plans and technological bottlenecks—such as T‑glass supply constraints and the lead times for new capacity additions—to calculate the projected supply‑demand gap. Finally, based on this gap, it derives expectations for pricing power and margin improvement, updates each company’s EPS estimates, and applies the Residual Income Model (RI Model) to value the stocks, arriving at revised target prices.

Methodology notes

• Industry/ Sector Analysis FrameworkSupply-and-Demand Framework

  By analyzing the lag in capacity expansion on the supply side and the structural boom on the demand side (driven by AI), we assess the turning point in industry sentiment.

  The core logic of the research report is as follows: Demand has surged due to the AI boom, while new capacity expansion on the supply side requires more than two years to come online, resulting in a significant supply-demand gap (22%) between 2027 and 2030, which in turn supports higher prices and margin expansion.

• Valuation MethodologyRIM Residual Income Model

  Valuation is conducted using the residual income model (RI) rather than the conventional P/E multiple.

  For the capital-intensive and highly cyclical ABF substrate industry, the RI model more accurately captures equity cost and long-term excess return potential, thereby mitigating the distortions inherent in relying solely on PE multiples during earnings troughs or peaks.

• Cyclical and Economic Cycle FrameworkAnalysis of the Economic Turning Point

  We judge that the industry is in the early stages of an upward cycle.

  The research report reviews the ABF industry’s historical cycles—downward from 2012 to 2018, supply shortages from 2020 to 2022, and a correction from 2023 to 2024—and notes that the sector has now passed its trough as of mid-2026. With the broadening of AI-driven demand, the industry is entering a new upward phase that could last for an extended period.

Asset mapping & comparison

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

• Unimicron Technology (3037.TW)

  Benefits: As a globally leading supplier of ABF substrates, the company holds a dominant position in the AI ASIC and server CPU substrate markets, directly benefiting from price hikes driven by supply-demand imbalances and increased market share.

  Strengths

  Technologically advanced, with a high-quality customer base (including key AI chip manufacturers), and strong execution in capacity expansion.

  Comparison

  Compared with NYPCB, it holds a larger share in extra-large‑size AI substrates; and compared with ZDT, its ABF business has a longer track record, warranting a reasonable valuation premium.

  Risks

  Persistent weakness in PC demand; risk of substitution by new technologies.

• South Asia Circuit Board (8046.TW)

  Benefits: The company holds a dominant position in the ABF substrate segment for network chips, and its BT substrate business is also poised to benefit from the rebound in memory demand.

  Strengths

  The company holds a significant share of the network chip substrate market, employs an aggressive pricing strategy, and maintains strong cost control.

  Comparison

  It outperforms its peers in niche segments of the network equipment market, with overall growth expectations exceeding those of ASE.

  Risks

  Fluctuations in network equipment investment; rising raw material costs.

• Zhen Ding Technology (4958.TW)

  Benefits: The company is a key supplier of substrates for AI chips in China, while also expanding into international major customers—such as Google’s TPU—and the high-margin optical module PCB business.

  Strengths

  The Shenzhen plant is already profitable, the company has secured a strong foothold in the Chinese AI market, and its optical module PCBs boast a high gross margin of over 40%.

  Weaknesses

  ABF has a relatively short operating history, and the progress of international major-client certification remains uncertain.

  Comparison

  Valued below that of pure ABF manufacturers, the company holds revaluation potential if it successfully secures a foothold in the supply chains of major international OEMs.

  Risks

  iPhone demand fell short of expectations; TPU substrate certification failed; domestic competition has intensified.

Key data

• 2030 ABF Supply-Demand Gap Forecast22%This represents a significant upward revision from the previously forecast 15%, primarily driven by stronger-than-expected AI-related demand.
• ABF Demand Share for AI/Server Applications in 2030>75%In 2015, it was only 10%, and the PC’s share will decline from 70% to below 15%.
• Xinxing Electronics target priceNT$1,225Implied upside of 49%, corresponding to a 2027E PEG ratio of 0.48x.
• South Asia Circuit Board Target PriceNT$1,275Implied upside of 62%, corresponding to a 2027E PEG ratio of 0.50x.
• Zhen Ding Technology target priceNT$570Implied upside of 42%; rating upgraded from EW to OW.
• Price Growth Forecast for ABF, 2026–202815-40% YoY15–20% in 2026, over 20% in 2027, and 25–40% in 2028.

Impact & implications

For ABF substrate suppliers, this translates into a golden period of both rising volumes and prices over the next three to five years. Leading players such as ASE and Unimicron, leveraging their market‑share advantages in high‑end AI chip substrates, are poised to see substantial margin expansion. For manufacturers like Zhen Ding Technology, which operate across both PCB and substrate businesses, their strategic positioning in AI server PCBs and China‑domestic AI chip substrates makes them particularly resilient and attractive investment targets. Across the broader semiconductor value chain, the scarcity of ABF substrates could emerge as one of the key bottlenecks constraining the shipment of high‑end chips, prompting chip design firms to secure long‑term supply agreements at an earlier stage.

Risks

• Demand for PCs, general-purpose servers, and AI servers has been weaker than expected.
• Major suppliers’ substantial capacity expansions have accelerated the restoration of supply-demand equilibrium.
• The supply constraints on key raw materials such as T‑glass have exceeded expectations.
• The adoption of advanced packaging technologies such as CoWoS, which do not require ABF substrates, is accelerating.

What to watch

• The actual procurement volumes of AI ASICs and CPUs by major cloud service providers (CSPs)
• The progress of new capacity commissioning and yield rates at leading manufacturers such as Xinxing and Nanya.
• Progress in certification at major international clients, including Google’s TPU project, for Zhen Ding Technology.
• ABF substrate‑on‑board quarterly price trends and gross margin changes