AI Demand and Chip Price Increases Reinforce Each Other, Driving Asia Semiconductor YoY Growth to a New High

Overview

The report analyzes the Asia semiconductor cycle based on June WSTS data. Semiconductor revenue grew 134% YoY, accelerating for the tenth consecutive month, mainly driven by AI compute demand, tight memory supply-demand, and logic chip price increases. Revenue growth was significantly faster than shipment growth, indicating that pricing and product mix upgrades are the core drivers of the current upcycle. J.P. Morgan expects this AI-driven semiconductor upcycle to continue, supported by cloud service provider capex expansion, advanced-node capacity buildout, HBM demand, and mature-node price increases.

Core views

First, demand for advanced-node logic chips remains strong, with TSMC demand visibility extending to 2029 to 2030, and expectations for capex and capacity expansion continuing to be raised. Second, memory supply-demand is structurally tight; prepayments, bulk purchases and AI server product premiums in long-term agreements reflect strong supplier pricing power, and the shortage could worsen further in 2027. Third, AI peripheral applications are improving mature-node utilization, with PSMC and VIS already implementing price increases, and further price hikes are expected in the second half of 2026 and in 2027. Fourth, DRAM demand is spreading from GPUs to AI CPUs, and the market may be underestimating this incremental demand, leaving room for upward revisions to memory market size forecasts. Fifth, localized strength in industrial applications and resilient iPhone 17 demand can partially offset weakness in low- to mid-end smartphones.

Analysis framework

The report uses WSTS monthly industry data as its basis, decomposes semiconductor revenue growth into shipment volume and ASP changes, and analyzes the logic and memory segments separately. It also combines TSMC capex and advanced-node capacity, DRAM wafer starts and HBM capacity allocation, long-term supply agreement terms, mature-node utilization and price increase magnitude, as well as cloud service provider capex forecasts to assess industry supply-demand, pricing and earnings trends from 2026 to 2028.

Methodology notes

• Industry cycle analysisVolume-price decomposition

  Revenue growth is jointly determined by shipment volume and ASP

  The report cross-compares YoY changes in revenue, shipments and ASP for semiconductors, memory and logic chips, concluding that current growth is driven more by price increases and product mix upgrades than by shipment expansion alone.

• Supply-demand analysisCapacity and demand gap analysis

  Assess shortage persistence through capacity deployment cycles and demand growth

  The report combines DRAM wafer starts, HBM capacity share, the 2 to 2.5-year cycle from fab construction to mass production, and AI CPU demand growth to conclude that memory supply is unlikely to fully meet demand in the near term.

• Capex analysisCapex leading indicator

  Capex and node expansion reflect medium-term demand visibility

  By raising TSMC capex and N5 and below capacity growth forecasts, the report gauges the strength of advanced-node demand and medium-term opportunities for foundries, equipment and related supply chains.

• Industry chain analysisAI demand transmission

  AI demand spreads from core compute to memory and peripheral chips

  The report examines not only GPUs and advanced nodes, but also HBM, AI CPUs, mature-node peripheral chips, packaging and testing, equipment and server hardware to identify the broad beneficiary segments of AI capex.

Asset mapping & comparison

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

• TSMC (2330.TW)

  Core beneficiary of advanced-node foundry

  Strengths

  AI compute demand visibility extends to 2029 to 2030, with both capex and N5 and below capacity forecasts raised.

  Weaknesses

  Large-scale capacity expansion increases capital intensity, and the report does not provide stock-specific valuation or target price corresponding to this forecast.

  Comparison

  Compared with mature-node foundries, TSMC benefits more directly from advanced-node demand and high-end product mix upgrades.

  Risks

  Cloud service provider capex slowdown, advanced-node demand falling short of expectations, or capacity expansion returns below expectations.

• Samsung Electronics (005930.KS, SEC)

  Beneficiary of DRAM, HBM and the memory cycle

  Strengths

  Memory revenue and ASP grew significantly, while tight supply, long-term agreement prepayments and AI server premiums strengthened supplier pricing power.

  Weaknesses

  Earnings are highly sensitive to the memory price cycle, and the report does not elaborate on its product execution or competitive share differences.

  Comparison

  Compared with logic foundry names, its earnings leverage depends more directly on DRAM and HBM pricing and the supply-demand gap.

  Risks

  Memory supply expansion outpacing demand, ASP gains retreating, HBM competition, or product execution falling short of expectations.

• VIS (5347.TWO)

  Beneficiary of mature-node foundry price increases

  Strengths

  AI peripheral demand is lifting utilization and tightening capacity; the report believes it raised prices for 8-inch products by 5% to 15% in the first half of 2026.

  Weaknesses

  Technology and product mix are mainly concentrated in mature nodes, and long-term growth depends more on peripheral demand and pricing discipline.

  Comparison

  Compared with advanced-node foundries, VIS's main catalysts are improved mature-node supply-demand and ASP improvement.

  Risks

  Utilization decline, end of customer restocking, or new mature-node capacity causing price increases to become unsustainable.

• MediaTek (2454.TW) and Alchip (3661.TW)

  Beneficiaries of AI and advanced-node chip design

  Strengths

  Advanced-node demand is strong, and AI demand is spreading from accelerators to CPUs and a broader range of computing chips.

  Weaknesses

  The report does not provide a detailed breakdown of order books, earnings forecasts or valuation differences for the two companies.

  Comparison

  Compared with foundries, chip design companies' returns depend more on product competitiveness, customer adoption and order realization.

  Risks

  AI chip project delays, customer concentration, rising R&D investment and advanced-node cost pressure.

• ASE (3711.TW), Unimicron (3037.TW) and ASMPT (0522.HK)

  Beneficiaries in the advanced packaging, substrate and packaging equipment industry chain

  Strengths

  AI chip complexity, advanced packaging demand and upstream capacity expansion are expected to drive demand for packaging and testing, substrates and equipment.

  Weaknesses

  The report mainly provides industry direction and does not quantify each company's specific order leverage or capacity constraints.

  Comparison

  Compared with chip manufacturers, this group of stocks represents indirect beneficiaries of capex and high-performance packaging demand.

  Risks

  Customer capex delays, overly rapid packaging capacity expansion, equipment acceptance delays, or weakening substrate supply-demand.

• Tokyo Electron (8035.T), AMEC and Naura (002371.SZ)

  Beneficiaries of semiconductor equipment capex

  Strengths

  Raised TSMC capex forecasts and advanced-node capacity expansion support equipment demand.

  Weaknesses

  Equipment revenue is typically affected by order recognition, delivery cycles and fluctuations in customer capex.

  Comparison

  Compared with fabs, equipment companies benefit from capacity expansion spending but are more sensitive to capex inflection points.

  Risks

  Fabs cutting or delaying capex, trade and export restrictions, or equipment delivery or acceptance falling short of expectations.

• Hon Precision, Chroma, Accton and Delta

  Beneficiaries of AI server, testing, networking and power peripheral demand

  Strengths

  Rapid growth in cloud service provider capex and data center AI demand is favorable for server manufacturing, testing, networking and power supply chains.

  Weaknesses

  The report does not provide a unified comparison of each company's AI business exposure, margins and valuation.

  Comparison

  Compared with core chip companies, these companies gain broader but relatively indirect growth from peripheral demand for AI infrastructure.

  Risks

  Cloud service provider spending slowdown, customer concentration, product price declines, supply chain bottlenecks, or weakness in non-AI businesses.

Key data

• June semiconductor revenue YoY growth134%May was 128%, accelerating for the tenth consecutive month and reaching a new high in YoY growth.
• June logic semiconductor revenue YoY growth38%May was 34%, marking 34 consecutive months of YoY growth.
• June memory revenue YoY growth369%May was 353%, the main source of acceleration in overall revenue.
• June semiconductor shipment YoY growth8%May was 13%, showing that high revenue growth was mainly driven by ASP rather than shipment volume.
• June memory and logic chip shipment YoY growth44% / 6%May was 41% and 11%, respectively.
• June overall semiconductor ASP YoY growth117%May was 102%.
• June memory and logic chip ASP YoY growth225% / 30%May was 222% and 20%, respectively.
• TSMC capex forecastFY26/FY27/FY28: US$62bn/US$81bn/US$90bnPrevious forecasts were US$58bn/US$78bn/US$84bn.
• TSMC advanced-node capacity growth2026 to 2028 CAGR 24%Defined as N5 and below nodes.
• HBM share of DRAM capacity32% by end-2028The report expects DRAM wafer capacity to continue shifting toward HBM.
• DRAM monthly wafer start forecastFrom 1.9M at end-2025 to 2.85M at end-2028Even after capacity expansion, it may still fall short of expected demand, as new capacity requires 2 to 2.5 years from construction start to ramp-up.
• AI CPU unit market size growthCAGR 155%The spread of demand from GPUs to CPUs may not yet be fully priced in by investors.
• Cloud service provider capex growth forecast65% in 2027The report expects growth to remain strong at above 100% in 2026.
• Data center AI growth expectation2024 to 2029 CAGR 69%Driven by stronger accelerator demand and new categories such as agentic AI CPUs.
• DRAM ASP forecast2026E/2027E/2028E: +235%/+29%/+7%Flexible pricing mechanisms enable memory vendors to capitalize on the current shortage environment.
• Mature-node price increase magnitudePSMC 10% to 15%; VIS 5% to 15%PSMC covers 8-inch and 12-inch, while VIS price increases mainly target 8-inch; the report expects further price increases ahead.

Impact & implications

Industry revenue growth is led by pricing and product mix upgrades, implying significant earnings leverage for memory vendors, advanced-node foundries and mature-node foundries with pricing power. TSMC's capex upgrade will also transmit demand to semiconductor equipment, packaging and testing, substrates and AI server hardware supply chains. However, the beneficiary paths differ across segments: memory relies more on shortages and ASP, advanced nodes rely on AI compute demand and capacity expansion, while mature nodes rely on utilization and peripheral chip price increases. The report does not provide a unified target price or sector valuation framework, so improving industry conditions cannot be directly equated with the same upside potential for all recommended stocks.

Risks

• Cloud service provider capex or AI compute demand falling below expectations, weakening demand for advanced nodes, memory and peripheral hardware.
• New DRAM and mature-node capacity coming online faster than demand, leading to a narrowing supply-demand gap and reversal of ASP increases.
• Long-term supply agreements, prepayments and AI product premiums failing to persist, reducing memory vendors' pricing power.
• Continued weakness in low- to mid-end smartphones, with iPhone 17 or industrial demand unable to fully offset end-market pressure.
• Fab capacity expansion, equipment delivery or new capacity ramp-up falling short of expectations, affecting the transmission of capex to supply chain performance.
• High-growth forecasts in the report involve estimation error, and actual results may deviate significantly due to cycle, competition, exchange rates and regulatory changes.
• J.P. Morgan has market-making, shareholding, client or investment banking relationships with some discussed companies; investors should make independent judgments in conjunction with the disclosures.

What to watch

• Whether subsequent monthly WSTS data can maintain YoY revenue acceleration, and whether the contributions from shipment volume and ASP change.
• Whether TSMC further raises capex, advanced-node capacity or data center AI growth expectations.
• The progress of DRAM wafer allocation toward HBM and the target of a 32% capacity share by end-2028.
• The proportion of prepayments, purchase volume contribution and AI server product premiums in long-term memory agreements.
• New rounds of price increases by mature-node foundries such as PSMC and VIS in the second half of 2026 and in 2027.
• Whether the memory shortage intensifies as expected in 2027, and whether supply-demand improves only slightly in 2028.
• Whether AI CPU demand can realize the expected 155% CAGR and drive demand to spread from GPUs to CPUs.
• Marginal changes in cloud service provider capex, iPhone 17 demand and the low- to mid-end smartphone market.