Nomura initiates coverage on CXMT with a Buy rating and a target price of CNY 116

Overview

This is an initiation-coverage company research report by Nomura on ChangXin Memory Technologies (CXMT). The report positions CXMT as the largest DRAM maker in China and the fourth-largest DRAM maker globally, and argues that against a backdrop of global memory supply that is difficult to ease quickly and rapidly growing memory demand driven by AI inference and agentic AI, the company is well placed to continue gaining market share.

Core views

The core views are: first, global memory demand growth may continue to exceed supply growth, with AI, RAG, and agentic AI driving a triple upcycle in commodity DRAM, HBM, and SSD; second, through capacity expansion in Hefei, Beijing, and Shanghai, as well as node migration from 16-17nm to 14-15nm and other nodes, CXMT is expected to achieve 40-45% bit CAGR in 2026-2030F; third, CXMT's current pricing is 0-20% lower than overseas leaders, but as process upgrades and bit per wafer improve, wafer ASP is expected to rise from about USD14-15k in 2026F to USD21-25k in 2027-2028F; fourth, the U.S. MATCH Act, Entity List measures, or restrictions on key materials may weigh on the company's yield, process migration, and capacity expansion.

Analysis framework

The report uses a top-down DRAM supply-demand framework combined with bottom-up analysis of CXMT's capacity, process, ASP, and earnings forecasts. On the demand side, it focuses on assessing the impact of agentic AI task volume, KV cache, context expansion, multi-round tool calling, and memory-efficiency technologies on memory demand; on the supply side, it analyzes bottlenecks to expansion such as clean rooms, equipment, materials, and talent; at the company level, it evaluates capacity, nodes, yield, customers, domestic substitution, and valuation multiples.

Methodology notes

• Valuation methodTarget P/E method

  20x 2028F EPS

  The target price of CNY 116 is based on 20x 2028F EPS of CNY 5.8. The report believes that CXMT should enjoy a premium to Micron's historical P/E of about 10x, given its share gains and the higher valuation multiples in the China market.

• Industry supply and demandDRAM bit supply-demand model

  Demand CAGR exceeds supply CAGR

  The report assumes memory demand CAGR of about 60% in 2026-2030F, while industry supply bit CAGR is about 30-40%, so a supply-demand gap may persist.

• AI hardware demandAgentic AI memory hierarchy analysis

  KV cache and multi-round tasks increase memory pressure

  The report breaks the agentic AI workflow into eight stages: request, weight loading, prefill, inference planning, tool execution, context integration, multi-step iteration, and response generation, and believes the concurrent memory pressure is highest during stage 7, the multi-step iteration stage.

• Risk assessmentGeopolitical and export control scenario analysis

  MATCH Act and potential sanctions

  The report analyzes the impact of U.S. export controls, restrictions on overseas equipment services, photoresist supply, and the localization rate of equipment on CXMT's process migration, yield, and Shanghai expansion.

Asset mapping & comparison

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

• CXMT

  Core target

  Strengths

  China's largest DRAM maker and the world's fourth largest; benefits from domestic substitution, AI memory demand, capacity expansion, and node migration.

  Weaknesses

  Its mainstream process technology still lags overseas leaders; advanced photoresist and some critical equipment depend on overseas supply; HBM3 qualification still needs time.

  Comparison

  The report uses Micron as a valuation reference and believes CXMT can obtain a target multiple about 2x Micron's historical P/E due to share gains and China's market valuation premium.

  Risks

  Deteriorating end demand, intensifying domestic competition, insufficient supply of ICs and CPUs and other components, and escalating geopolitical tensions between the U.S. and China.

• Micron

  Overseas comparable company

  Strengths

  One of the world's major DRAM makers; its roughly 10x P/E over the past five years is used as a valuation reference.

  Weaknesses

  The report does not rate Micron and mainly uses it for valuation benchmarking.

  Comparison

  CXMT's target valuation multiple is 20x, about double Micron's historical roughly 10x.

  Risks

  Global DRAM cycles, price volatility, and changes in AI demand.

• YMTC

  Domestic competitive reference

  Strengths

  An important participant in China's domestic memory industry chain.

  Weaknesses

  The report lists it as a domestic peer that could intensify competition.

  Comparison

  Like CXMT, it belongs to China's memory industry chain, but the report focuses on CXMT's DRAM opportunity.

  Risks

  Intensifying competition could affect CXMT's market share, pricing, and profitability.

• GigaDevice

  Industry partner

  Strengths

  Cooperates with CXMT on DRAM-on-logic production and design, targeting China's edge AI and physical AI demand.

  Weaknesses

  The report does not elaborate on its financial impact.

  Comparison

  It is a cooperative ecosystem participant for CXMT in WoW and edge AI.

  Risks

  Uncertainty in technology mass-production timing, customer ramp-up, and application demand.

• Leadmicro

  Domestic equipment partner

  Strengths

  Mentioned in the report's ALD equipment section as a key domestic equipment partner.

  Weaknesses

  Affected by CXMT's expansion pace and localization validation.

  Comparison

  Positioned within the domestic substitution segment of CXMT's local supply chain.

  Risks

  Equipment performance, yield validation, and substitution pressure under export controls.

Key data

• RatingBuyNomura initiates coverage on CXMT.
• Target priceCNY 116.00Based on 20x 2028F EPS of CNY 5.8.
• IPO priceCNY 8.66The report table shows the IPO price as the current reference price.
• Implied upside+1,239.5%Calculated from the target price relative to the IPO price.
• 2026-2028F revenue CAGR63%Supported by capacity expansion, node migration, and ASP improvement.
• 2026-2028F attributable net profit CAGR74%The report forecasts rapid net profit growth.
• 2028F EPSCNY 5.8Valuation basis for the target price.
• Global DRAM shareAbout 10% currently, about 18% by end-2028The report expects CXMT's share to continue rising.
• 2026-2030F CXMT bit CAGR40-45%Higher than industry supply growth, but still lower than the demand growth assumed in the report.
• 2026-2030F memory demand CAGRAbove 60%Driven by agentic AI concurrent tasks and expansion of AI applications.
• Capacity by end-2028About 550kwpmThe report expects 280kwpm by end-2025, 350kwpm by end-2026, and an additional 100kwpm in each of 2027F and 2028F.
• Equipment localization rateAbout 40%The report's estimate; restrictions on overseas equipment and materials are key monitoring points.

Impact & implications

If the report's thesis plays out, CXMT will become a key beneficiary of China's domestic DRAM substitution and the tight global memory supply-demand cycle. AI inference, agentic AI, multi-round tool calling, and edge AI applications may expand demand for DRAM, HBM, SSD, and advanced packaging, while also benefiting upstream equipment, materials, and OSAT ecosystems. However, the investment case depends heavily on persistently tight supply-demand conditions, smooth capacity expansion, successful process migration, and manageable external regulatory risks.

Risks

• Deteriorating end demand may cause DRAM pricing or shipments to fall short of expectations.
• Intensifying competition from domestic peers such as YMTC may weaken CXMT's share and profitability.
• Insufficient supply of CPUs, ICs, and other key components may constrain downstream device demand and memory absorption.
• Escalating geopolitical tensions between the U.S. and China, the MATCH Act, or Entity List measures may restrict access to equipment, materials, or services.
• High-end photoresist is mainly dominated by Japanese suppliers, and restrictions could affect sub-15nm node migration.
• If memory-efficiency technologies deliver greater savings than assumed in the report, AI-driven memory demand may be lower than expected.
• If agentic AI adoption or end-user usage intensity is lower than expected, demand elasticity will weaken.
• If Shanghai expansion is constrained by equipment or material restrictions, the 2028F total capacity target may be delayed.

What to watch

• Funding after CXMT's listing, over-allotment execution, and the pace of capex deployment.
• Whether capacity expansion in Hefei, Beijing, and Shanghai proceeds as planned for 2026-2028F.
• Progress in node migration from 16-17nm to 14-15nm, sub-15nm, and 12-16nm nodes.
• Improvement trends in DDR5 yield at around 80% and DDR4 yield above 90%.
• HBM3 sample validation, customer qualification, and progress in HBM3e R&D.
• Whether DRAM prices stabilize after 2028F and whether long-term supply agreements expand.
• Changes in the U.S. MATCH Act, Entity List measures, restrictions on overseas equipment services, and photoresist supply.
• The actual rollout of agentic AI task volume, AI bot traffic, KV cache optimization, and memory-efficiency technologies.