Old-Process NAND Confirmed in AI Data Centers, Three Greater China Manufacturers Favored

Overview

This industry research report starts from the information released by the Japanese company Kioxia at its Investor Day and conducts a cross-industry interpretation. The core conclusion of the report is that the high-performance solid-state drives launched by Kioxia, combining old-process NAND technology, are expected to create a vast new market in the AI data center field, which will benefit multiple old-process NAND suppliers in Greater China. Morgan Stanley explicitly expressed optimism about GigaDevice, Winbond, and Macronix International, and maintains an 'Overweight' rating on all three companies.

Core views

The core view of the report is that old NAND flash technology optimized for high-performance AI applications is about to see demand growth. The report first points out that the GP series SSD launched by Kioxia at its Investor Day possesses over 100 million IOPS, ultra-high speed, and low latency performance, aimed at accelerating next-generation AI applications. The key is that this product uses Kioxia's proprietary SLC chips, a relatively 'old' but stable NAND technology. This indicates that old-process NAND, due to ideal read/write speeds, has a concrete implementation plan for data center applications, and the potential total market size is expected to expand. Based on this external validation, the report believes this constitutes a positive signal for three Greater China suppliers that also have old-process NAND product portfolios—GigaDevice, Winbond, and Macronix International. The report reiterates its Overweight rating on these three companies, considering them the main beneficiaries of this trend.

Analysis framework

The report's analysis path is the 'Cross-Industry Mapping' method starting from specific events. The institution first focuses on an external market event (Kioxia IR Day), extracting an industry trend signal: old NAND technology finds new high-value applications in AI data centers. Then, this trend is correlated and mapped to covered Greater China companies with similar business structures (old-process NAND suppliers) to assess their benefit potential and degree. In the valuation section for specific companies, different methods are mixed based on each company's business characteristics and industry attributes: for Macronix and Winbond with strong cyclical attributes, the Price-to-Book (PB) method compared with historical cycles and peers is used; for GigaDevice, the Residual Income Model focusing more on long-term profitability is used, reflecting considerations of different drivers and industry characteristics.

Methodology notes

• Valuation MethodPB valuation

  Price-to-Book Ratio Valuation Method

  Price-to-Book ratio (stock price/net assets per share) is commonly used in strong cyclical industries (such as semiconductor memory), as asset value and supply-demand relationships greatly impact profits. The report uses this method for Macronix and Winbond, with target multiples (7.0x and 5.5x) far higher than historical averages, reflecting judgment on the industry cycle and potential supply shortages for specific product lines.

• Valuation MethodRIM Residual Income Model

  Residual Income Model

  This method measures the excess profit created by the company after deducting the cost of equity capital and discounts it for valuation. The report uses this model for GigaDevice, with key assumptions including a cost of equity of 8.9% (including beta 1.1) and a medium-term growth rate of 16.4%, reflecting an assessment of its ability to continue creating excess returns.

• Industry/Industrial Analysis FrameworkCross-Industry Mapping Method

  Cross-Industry Mapping Method

  Trend information released by industry giants (Kioxia) is used as a 'leading indicator', mapped to covered companies with similar business models (such as GigaDevice, etc.), to predict the business prospects of the covered companies. This is a common approach for event-driven industry analysis and opportunity discovery.

Asset mapping & comparison

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

• Macronix International (2337.TW)

  As a key supplier of old-process MLC/TLC NAND, it benefits most directly from the supply-demand gap expected in the report and is identified as the only supplier capable of filling the gap.

  Risks

  Demand for NOR Flash, Switch hardware, and game sales is weaker than expected; old NAND supply-demand gap is smaller than expected.

• Winbond Electronics (2344.TW)

  Benefits from DRAM price increases, NOR Flash price increases, and opportunities in SLC NAND and SiCap foundry business.

  Risks

  NOR Flash demand weakens; DRAM oversupply leads to prices below expectations; SLC NAND development slower than expected.

• GigaDevice (603986.SS)

  Benefits from the NOR Flash industry upcycle expected in the report, as well as development potential in the low-density NOR and DRAM fields.

  Risks

  NOR Flash demand weakens; chip design concentrates on medium/high-density NOR; DRAM development slower than expected.

Key data

• Macronix International 2026 Target PB Multiple7.0xThe target multiple is several times the historical average (since 2017) of 1.5x, based on the expectation that MLC/old TLC NAND shortages may intensify in the second half of 2026.
• Macronix International Potential Supply GapUp to 40%The report predicts that MLC and old TLC NAND may face huge shortages in the second half of 2026.
• Winbond Electronics 2026e Target PB Multiple5.5xHigher than the high end of its historical trading range (0.5-1.5x), based on opportunities in DRAM prices, NOR Flash prices, and SLC NAND/SiCap business.
• GigaDevice Cost of Equity Assumption (WACC)8.9%Includes Beta coefficient 1.1, risk-free rate 2%, and risk premium 6%

Impact & implications

The report believes that if the application trend of old-process NAND in AI data centers is confirmed, it will directly expand the potential market space for GigaDevice, Winbond, and Macronix International. This is not only reflected in revenue growth but is more likely to drive product price increases and company valuation restructuring due to tightness on the supply side (only a few suppliers like Macronix can fill the gap). The high PB target multiples given by the report for each company are based on strong expectations of supply-demand imbalance and industry upcycle.

Risks

• Supply-demand prospects for old-process NAND reverse, gap not as severe as expected; or demand growth falls short of expectations.
• NOR Flash memory industry enters a downward cycle, prices fall, demand weakens.
• DRAM price trend falls short of expectations, or falls due to oversupply.
• Development speed of relevant businesses for each company (such as SLC NAND, Switch hardware) falls short of expectations.
• Adjustments in chip design strategy may lead to increased product portfolio risk.

What to watch

• Actual adoption progress and market size changes of old-process NAND (SLC/MLC/TLC) in data centers, especially in AI applications.
• Changes in the overall supply-demand pattern and price trends of the memory industry (NAND, NOR, DRAM).
• Technical progress and customer expansion of GigaDevice, Winbond, and Macronix International in their respective key businesses (NOR/DRAM/SLC NAND).