DDR4 Downgrade Rationale Amid a Bifurcating Legacy Memory Market
We are seeing a significant and potentially overlooked divergence within the legacy memory space. While our recent downgrade of DDR4 has faced investor pushback, we maintain our view that DDR4 fundamentals are poised to deteriorate due to accelerating supply additions and softening demand, creating a stark contrast with structurally tight legacy NAND. The market appears to conflate the dynamics of legacy and mainstream memory or misjudge the supply response timing for DDR4. Consequently, we shift our Top Pick to Macronix, a primary beneficiary of the legacy NAND shortage, while turning more cautious on DDR4-heavy names like Winbond and Nanya.
Market Misinterpretation of DDR4 Dynamics Within the Broader Memory Sector
Investors incorrectly read our DDR4 downgrade as a negative signal for the mainstream memory market. This is a key mispricing. Our call is specific to the legacy DDR4 market, which operates under a much smaller capacity base and different supply-demand drivers compared to mainstream DRAM. We remain constructive on mainstream DRAM, where supply continues to tighten due to agentic AI workloads demanding more data-intensive processing. AI spending is maxed out, and mainstream memory is a critical gating factor. The strength in high-bandwidth memory (HBM) and mainstream DDR5 does not extend to legacy DDR4, which lacks direct AI exposure.
Supply Acceleration and Lack of AI Demand Drive DDR4 Downgrade
Our downgrade is based on a clear inflection in DDR4 supply and demand. On the supply side, elevated prices have prompted aggressive capacity expansions. Winbond plans to double DRAM bit shipments in 2026, Nanya will bring new capacity online in 2H27, and CXMT intends to more than double its DDR4 capacity in 2027. We also believe PSMC will license advanced 1y/1z processes for DDR4, ramping from 1H27. These actions contrast with the continued supply discipline and bottlenecks in mainstream DRAM. On the demand side, high DDR4 prices have eroded consumer demand in smartphones and PCs, and the product cycle does not benefit from AI-driven memory consumption growth.
Structural Shortage in Legacy NAND Presents a Superior Opportunity
As we downgrade DDR4, we see a more compelling structural opportunity in legacy NAND, particularly MLC. This segment faces a deeper supply shortage, with a potential undersupply reaching ~40% in 2H26, versus a more transitory setup for DDR4. Major global suppliers continue to reduce legacy NAND supply, while demand remains resilient due to its stronger endurance and lower price sensitivity in industrial/enterprise applications. We forecast MLC and legacy TLC NAND pricing to rise over 200% from 1Q26 to 4Q26. This distinct dynamic makes legacy NAND the clear winner in the old memory space for 2026.
The Timing: Price Hikes Decelerate, Signaling Peak Momentum
The decision to downgrade now is driven by price momentum and inventory behavior. DDR4 8Gb contract prices were up 752% year-over-year as of February 2026. As hikes continued into March, smaller customers began pushing back, and suppliers showed a willingness to sell inventory, suggesting limited near-term upside. Our models forecast sequential price increases for key suppliers to decelerate sharply through 2026: we model Q/Q DRAM price increases for Winbond at 50%/50%/30%/10% across the four quarters of 2026. This decelerating year-over-year growth trajectory into the second half of the year is a key signal to turn cautious.
Investment Implications and Positioning
The investment implication is to rotate exposure within the legacy memory segment. Favor names leveraged to the legacy NAND shortage, notably Macronix International (2337.TW), our new Top Pick. Exercise caution on stocks with heavy DDR4 exposure, such as Winbond Electronics (2344.TW) and Nanya Technology (2408.TW), where expanding supply poses a risk to pricing power in 2H26. For Powerchip (6770.TW), monitor its DDR4 capacity plans, though its model includes other strategic partnerships. Our view on mainstream memory and related AI beneficiaries remains unchanged and separate from this legacy call.
Key Risks
The primary risk to our thesis is that DDR4 capacity expansions are delayed or less aggressive than indicated, prolonging the tight supply environment. Conversely, a sudden slowdown in AI demand could weaken the entire memory sector sentiment, negatively impacting even structurally sound segments like legacy NAND. For the legacy NAND bull case, a risk is that the supply shortage eases more quickly than expected if major producers reverse their output reduction strategies.