Quick Summary
Covering the latest research from top Wall Street investment banks

TSMC’s capex upcycle becomes the core theme in Taiwan semiconductors

Institution
Goldman Sachs
Date
2026-04-29
Authors
Bruce Lu, Evelyn Yu, Ryan Huang, CFA
Company
TSMC; MediaTek; Aspeed; All Ring; GPTC; Hon Precision
Ticker
2330.TW/TSM; 2454.TW; 5274.TWO; 6187.TWO; 3131.TWO; 7769.TW
Industry
Semiconductors
Rating
Buy
BullishLow confidenceInvestor discussions remain constructive on AI demand, TSMC capex and advanced packaging. Goldman Sachs highlights large-cap value creation names and selected equipment/testing beneficiaries with Buy ratings.
AuthorsBruce Lu, Evelyn Yu, Ryan Huang, CFA
Target priceTSMC NT$2,750; TSM ADR US$550; MediaTek NT$2,454; Aspeed NT$20,000; All Ring NT$1,800; GPTC NT$4,500; Hon Precision NT$5,700
Asset classesEquity
Business segmentsleading-edge foundry、advanced packaging、AI ASIC、BMC SoC、semiconductor equipment、semiconductor testing equipment、mature foundry
Research firm divisions/subsidiariesGoldman Sachs(Other)

AI summary card

TSMC’s capex upcycle becomes the core theme in Taiwan semiconductors

After speaking with more than 50 investors in Hong Kong, Goldman Sachs believes capital is rotating back from small- and mid-caps that have risen strongly since the start of the year into large-cap value stocks with high value creation such as TSMC and MediaTek, while also remaining positive on the supply chains for advanced packaging, AI ASIC, BMC, and testing equipment.

Buy ratings are maintained on the main covered names; target prices are NT$2,750 for TSMC, US$550 for TSM ADR, NT$2,454 for MediaTek, NT$20,000 for Aspeed, NT$1,800 for All Ring, NT$4,500 for GPTC, and NT$5,700 for Hon Precision.
Taiwan technologySemiconductorsAI demandTSMC capexAdvanced packagingCoWoSSoICValue stocksEquipment supply chain
  • Goldman Sachs expects TSMC’s total capex to reach US$200bn in 2026-2028E, making it a key anchor for the entire semiconductor ecosystem.
  • The report believes AI/HPC demand is driving N3 and CoWoS capacity expansion, and that Taiwan’s equipment and materials supply chain is poised to benefit from higher dollar content, greater technical complexity, and share gains.
  • Key recommendations include TSMC, MediaTek, Aspeed, GPTC, All Ring, and Hon Precision, while mature-node foundries are seen as less attractive due to limited structural demand changes.
  • MediaTek is viewed as underowned and offering asymmetric upside from AI ASICs, with the ramp in AI ASIC revenue in 2027E as the key catalyst.
  • The SoIC theme remains underappreciated by investors, and Taiwan supply chain companies such as GPTC and Scientech may benefit from next-generation AI GPUs, optical engines, and high-performance computing architectures.

Report interpretation

Overview

This report is Goldman Sachs’ marketing feedback and multi-company research on Taiwan technology. After meeting with more than 50 investors in Hong Kong, the authors note that investors remain positive on AI demand, but capital flows are beginning to rotate back from small- and mid-cap stocks that have performed strongly since the start of the year into large-cap value stocks with strong value creation capabilities, such as TSMC and MediaTek. The report identifies TSMC’s capex upcycle through 2028 as the core anchor of Taiwan’s semiconductor ecosystem and extends that theme to advanced packaging, equipment, materials, BMC, AI ASIC, and testing supply chains.

Core views

The core views include: first, TSMC’s US$200bn capex cycle and expected revenue CAGR of about 30% through 2028E will drive stronger growth for Taiwan’s equipment supply chain; second, CoWoS remains the main near-term driver in advanced packaging, but the strategic importance of SoIC in next-generation AI GPUs, optical engines, and HPC architectures has not yet been fully priced in; third, MediaTek’s AI ASIC business, Aspeed’s BMC competitive positioning, All Ring and GPTC’s share advantages in advanced packaging equipment, and Hon Precision’s position in AI/HPC testing equipment are the key favored areas in the report; fourth, mature-node foundries such as UMC and Vanguard are seen as having limited appeal relative to AI-driven opportunities.

Analysis framework

The report combines investor marketing feedback, company investment theses, target price methodologies, earnings growth assumptions, market share, technology roadmaps, and valuation multiples in its analysis. The focus is not on single-company earnings commentary, but on mapping the expansion of TSMC capex and AI demand to high-beta beneficiaries across Taiwan’s semiconductor supply chain.

Methodology notes

  • Valuation methodsTarget P/E valuation method

    Derive the 12-month target price by multiplying the target price-to-earnings ratio by forward EPS

    TSMC’s target price is based on 22x 2027E P/E, MediaTek on 20x FY27E P/E, All Ring and GPTC on 35x FY27E P/E, and Hon Precision on 32x 2027E P/E.

  • Valuation methodsDiscounted target price method

    Discount forward valuation back to the target year

    Aspeed’s target price is derived using 40x 2028E P/E and discounted back to 2027E at a cost of equity of 13.6%.

  • Investment frameworkGS Factor Profile

    Comparison of growth, financial returns, valuation multiples, and composite factors

    Goldman Sachs’ factor framework generates percentiles using indicators such as forward sales, EBITDA, EPS, ROE, ROCE, CROCI, and valuation multiples to provide investment context for stocks.

  • M&A assessmentM&A Rank

    Assess the probability of a company becoming an acquisition target on a scale of 1 to 3

    A rank of 1 indicates high probability, 2 medium probability, and 3 low probability; if the rank is 1 or 2, Goldman Sachs may incorporate M&A factors into the target price.

Asset mapping & comparison

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

  • TSMC / 2330.TW / TSM
    Core anchor of AI demand and capex cycle
    Strengths
    Leading-edge process and CoWoS technology, more than 60% global foundry revenue share, and Goldman Sachs expects revenue CAGR of about 25% over the next few years while maintaining long-term gross margin above 56%.
    Weaknesses
    Highly capital intensive, with valuation and earnings sensitive to AI investment, customer migration, and execution quality.
    Comparison
    Compared with mature-node foundries, TSMC benefits more directly from long-term structural growth in AI/5G/HPC/EV.
    Risks
    Deteriorating end demand, slower customer node migration, slower AI investment, poor yield or execution, intensified competition, adverse FX, or cost headwinds.
  • MediaTek / 2454.TW
    Beneficiary of AI ASIC and value-stock rotation
    Strengths
    Viewed by Goldman Sachs as underowned, with faster AI ASIC revenue ramp in 2027E and long-term TPU projects that could expand ASP and earnings upside.
    Weaknesses
    Still affected in the short term by weak smartphone demand.
    Comparison
    The report emphasizes its transition from a traditional smartphone application processor supplier to an AI-related supplier.
    Risks
    Weaker-than-expected end demand such as smartphones, rising foundry costs, intensified competition, and slower ASIC ramp.
  • Aspeed / 5274.TWO
    Beneficiary of AI server and BMC demand
    Strengths
    BMC is a duopoly market, Aspeed has about 70%+ share, and AST2700 penetration, CPU server demand, and potential price hikes support an earnings upgrade cycle.
    Weaknesses
    Investors are concerned about the high valuation.
    Comparison
    Goldman Sachs believes the company can sustain a high valuation as long as the competitive landscape and BMC TAM growth outlook remain favorable.
    Risks
    Weaker-than-expected recovery in server demand, slower BMC penetration in AI servers, and intensified competition.
  • All Ring / 6187.TWO
    Beneficiary of CoWoS and advanced packaging equipment
    Strengths
    It holds close to 100% share in WoS underfill dispenser and TIM heatsink attach equipment, benefiting from advanced packaging expansion and the CPO business.
    Weaknesses
    Growth is highly dependent on the pace of advanced packaging capacity expansion.
    Comparison
    Compared with general semiconductor equipment makers, it has higher share and more direct leverage in specific back-end advanced packaging equipment.
    Risks
    Slower advanced packaging expansion, delayed adoption of new packaging technologies, and intensified competition.
  • GPTC / 3131.TWO
    Beneficiary of CoWoS and SoIC wet process equipment
    Strengths
    About 50% share in TSMC CoWoS wet clean, close to 100% at ASE/SPIL, and the sole supplier of SoIC wet clean at TSMC.
    Weaknesses
    Valuation and growth depend on advanced packaging penetration and increasing technical complexity.
    Comparison
    Compared with names driven only by CoWoS, GPTC also benefits from higher ASP associated with complex packaging technologies such as SoIC, CPO, and FOPLP.
    Risks
    Weak AI/HPC demand, slower adoption of new advanced packaging technologies, and intensified competition.
  • Hon Precision / 7769.TW
    Beneficiary of AI/HPC testing equipment
    Strengths
    More than 90% share in the FT handler market for AI/HPC applications, benefiting from growth in the number of AI GPUs/ASICs, larger package sizes, longer testing time, and higher power and thermal requirements.
    Weaknesses
    The market may take time to fully reflect its earnings growth outlook, and current rerating still depends on execution and demand delivery.
    Comparison
    The report believes its forward P/E valuation is lower than Taiwan and global peers, implying rerating potential.
    Risks
    Weak AI/HPC demand, slower adoption of SLT in AI ASICs, and intensified competition.
  • Mature-node foundries
    A relatively less favored segment
    Strengths
    May benefit from demand spillover from TSMC and long-term content growth such as AI PMICs.
    Weaknesses
    Still has high exposure to consumer demand, limited near-term structural demand change, and potential price increases are mostly low- to mid-single digit and mainly reflect rising costs.
    Comparison
    Compared with AI-driven advanced process and advanced packaging opportunities, the report sees limited appeal in mature-node foundries.
    Risks
    Insufficient demand recovery, limited pricing momentum, and constrained gross margin improvement.

Key data

  • Number of investor meetingsover 50 investorsGoldman Sachs spoke with more than 50 investors during its marketing trip in Hong Kong.
  • TSMC capexUS$200bn over 2026-28EThe report views this as an important anchor for the semiconductor equipment and materials ecosystem.
  • TSMC revenue growth30% revenue growth CAGR into 2028Goldman Sachs estimates this will drive a stronger growth outlook for the supply chain.
  • MediaTek AI ASIC revenueUS$2.0bn/US$12.3bn in 2026E/2027ERevenue contribution is expected to reach 10%/39%.
  • MediaTek revenue/earnings CAGR44%/85% in 2026-28EDriven by high-end 5G flagship SoC share, AI ASIC ramp-up, and new TAM from automotive and computing.
  • Aspeed BMC share70%+ global BMC market shareGoldman Sachs estimates BMC accounts for about 85% of Aspeed’s total revenue.
  • Aspeed revenue/earnings CAGR66%/72% in 2025-2028EMainly driven by AI server demand and changes in server architecture.
  • All Ring sharenear 100% in WoS underfill dispenser and TIM heatsink attach equipmentIts products are widely used in the CoWoS process.
  • All Ring revenue/earnings CAGR54%/60% in 2025-28EDriven by advanced packaging capacity expansion and rapid growth in the CPO business.
  • GPTC market share50% at TSMC CoWoS wet clean; near 100% at ASE/SPIL; sole SoIC wet clean provider at TSMCReflects its leading position in advanced packaging wet process equipment.
  • GPTC revenue/earnings CAGR31%/49% in 2025-28EDriven by advanced packaging capacity expansion and higher equipment ASP.
  • Hon Precision market share>90% in FT handler market for AI/HPC applicationsIt can benefit from growing testing demand for AI GPUs and AI ASICs.

Impact & implications

The investment implication of the report is that AI demand benefits not only leading foundries such as TSMC, but also spreads through pathways including N3, CoWoS, SoIC, CPO, FOPLP, AI ASIC, BMC, and longer testing time to Taiwan’s equipment and testing supply chains. Capital preference may rebalance from sharply rising small- and mid-caps toward large-cap or high-quality growth stocks with scale, value creation, and room for earnings upgrades. Mature-node foundries may relatively underperform AI-related supply chains due to weak consumer demand, limited pricing power, and insufficient structural change.

Risks

  • A slowdown in AI investment leading to lower-than-expected long-term semiconductor content growth.
  • A further deterioration in end-demand recovery, affecting capacity utilization and order visibility.
  • Slower-than-expected customer node migration, weakening demand for advanced process and advanced packaging.
  • Delayed advanced packaging expansion or adoption of new technologies such as SoIC/CPO/FOPLP.
  • Intensified competition putting pressure on ASP or profitability.
  • Poor yield, execution, or cost control leading to weaker-than-expected margins.
  • Adverse foreign exchange movements or cost increases above expectations.
  • Weak smartphone demand affecting MediaTek’s short-term performance.
  • Weaker-than-expected recovery in server market demand or lower-than-expected BMC penetration in AI servers.

What to watch

  • TSMC’s 2026-2028E capex guidance and the pace of N3 and CoWoS capacity expansion.
  • AI/HPC customer demand and order visibility across TSMC’s supply chain.
  • The pace of SoIC adoption in AI GPUs, optical engines, and high-performance computing architectures.
  • MediaTek’s 2027E AI ASIC revenue ramp, CoWoS allocation, and progress on next-generation TPU projects.
  • Aspeed AST2700 penetration, CSP server CPU demand, and price adjustments.
  • Order growth for All Ring in CoWoS, CPO, and PLP-related equipment.
  • GPTC’s share and ASP changes in CoWoS and SoIC wet clean equipment.
  • The extent of Hon Precision’s benefit from AI GPU/ASIC testing, SLT adoption, and longer testing time.
  • Whether mature-node foundries show a true structural demand inflection rather than merely low-single-digit price increases.
Zhejiang ICP No. 2022035445-5
Disclaimer: Market data, charts, indicators, research views, and other information provided on this website are intended solely for information display, research communication, and educational reference. They should not be regarded as personalized investment advice, securities recommendations, trading instructions, solicitations, or guarantees of return. While we strive to improve the reliability of our data and content, such information may still be subject to delays, errors, incompleteness, or untimely updates due to source differences, methodological limitations, system processing, or market volatility. Users should exercise independent judgment based on their own circumstances and bear all risks and responsibilities arising from the use of this website.

Settings

Sign in to view recent logins