After Memory Stocks Plunge: Global X KOSPI 200 ETF

The KOSPI 200, the primary benchmark index for the South Korean equity market, is heavily concentrated in just two semiconductor giants: Samsung Electronics and SK Hynix, which together account for approximately 60% of the index’s total weight. Thus, their rapid share price appreciation over the past year has turned the index into a highly volatile index, as witnessed during the massive correction in July and October.

However, we believe the recent share price crash of these two Korean semiconductor leaders is overdone and presents a good trading opportunity for the Global X KOSPI 200 ETF (3408 HK).

[Table] Top 10 Holdings of 3408 HK

Name	Ticker	Weight (%)
SAMSUNG ELECTRONICS	005930 KS	33.6
SK HYNIX INC	000660 KS	24.6
SK SQUARE	402340 KS	2.4
SAMSUNG ELECTRO-MECHANICS	009150 KS	2.2
KB FINANCIAL GROUP INC	105560 KS	1.6
HYUNDAI MOTOR	005380 KS	1.4
SHINHAN FINANCIAL GROUP	055550 KS	1.3
SAMSUNG SDI	006400 KS	1.1
HANA FINANCIAL GROUP	086790 KS	1.0
CELLTRION INC	068270 KS	1.0

(Source) Mirae Asset Global Investments (HK), 15 Oct 2026

The Backdrop of the Crash: Supply Fears driven by CXMT’s Expansion

The primary driver behind the sharp sell-off in early and mid-October is an amplified market fear of a conventional DRAM oversupply. This panic was triggered by China’s Chang Xin Memory (CXMT), which announced its 5th-Generation (G5) platform in late September, followed by a 4F² cell architecture breakthrough in early October.

• 5th-Gen (LPDDR5X): new fabs will produce 3GB LPDDRX DRAM, which is a highly suitable memory for on-device AI smartphones (to run 10bn+ parameter AI models), AI PCs (24GB/48GB DRAM), humanoid robotics, and automotive driving cockpits. It implies CXMT will encroach into a larger addressable market of DRAM.
• 4F² Cell Architecture Breakthrough: This announcement caught global investors by surprise, as mainstream titans like Samsung Electronics and SK Hynix have not yet transitioned this architecture to commercial mass production. CXMT plans to launch its first 4F² product, a server DDR5 RDIMM by the end of 2026. While scepticism remains regarding manufacturing yields, this technology perfectly aligns with mainland China’s broader push toward 3D semiconductor architecture, a structural scaling method already deployed by YMTC and Huawei (Logic Folding) to bypass Western lithography constraints.

Historically, the global market was certain that advancing past legacy DRAM nodes would be physically impossible without Extreme Ultraviolet (EUV) lithography equipment. By executing a next-generation vertical 3D design, CXMT has demonstrated that it may bypass these geopolitical and technical equipment limits.

Why Conventional DRAM Matters More Than HBM

It is critical to recognize that CXMT’s recent technological breakthroughs are strictly confined to the conventional DRAM, with virtually no near-term impact on High Bandwidth Memory (HBM).

Independent industry tracking hints that CXMT is still struggling to achieve stable output for HBM3, and it is unlikely to achieve commercial mass production for advanced HBM3e until 2028, years behind global peers who achieved this milestone in 2025.

Nevertheless, equity markets remain hyper-sensitive to conventional DRAM pricing rather than HBM metrics for the following reasons:

• The Real Earnings Engine: Over the past 12 months, the true driver behind the massive earnings expansions and surging stock prices of memory chipmakers was conventional DRAM, not HBM.
• The 3x Price Spike: Conventional DRAM spot and contract prices skyrocketed 3 times over the last year. This explosive pricing leverage was the primary engine behind the record-breaking operating profits for the Korean semiconductor duo.

Assessing the Real Threat of Conventional DRAM Supply Glut

The critical question for investors is whether CXMT can genuinely flood the global market and crash conventional DRAM prices through these technical milestones. While this represents a credible structural threat over a long-term horizon, immediate operational realities present a severe bottleneck:

• The Yield Bottleneck: Securing stable, high-volume production yields for next-generation 3D vertical transistors is exceptionally difficult.
• Limited Supply Growth: Factoring in CXMT’s aggressive capacity expansion, global conventional DRAM bit growth is projected to expand by about 20% over the next two years (through 2028). This tightness is partially caused by the “crowding-out” effect, where tier-one makers continue to cannibalize conventional DRAM wafer capacity to feed HBM lines.
• Accelerating Demand: In contrast, underlying demand for conventional DRAM is projected to comfortably outpace this 20% supply ceiling, driven by the new demand from edge AI hardware, enterprise server updates, robots, autonomous driving, etc.

[Table] DRAM Supply (bit) Growth forecast for Conventional DRAM by 2028

Manufacturer	2026 Wafer Capacity	2028 Est. Wafer Capacity	WPM Growth Rate	Est. 2-Year Effective Bit Growth for Server/Mobile DRAM
Samsung	~720,000 WPM	~895,000 WPM	24%	12%~15%
SK Hynix	~590,000 WPM	~800,000 WPM	36%	10%~13%
Micron	~375,000 WPM	~605,000 WPM	60%	25%~30%
CXMT	~350,000 WPM	~500,000 WPM	43%	35%
Others	~100,000 WPM	~145,000 WPM	45%	~5%
Global Total	~2.13M WPM	~2.94M WPM	37%	~20%

(source) SemiAnalysis, TrendForce, Mirae Asset Global Investments (HK), Oct 2026
(Note) the table doesn’t include wafer for HBM, whose demand have been already secured by big tech till 2028.

Global X KOSPI 200 ETF

This implies that the fundamental downside for the KOSPI 200 may be limited, as these memory stocks account for 60% of the index’s total weight.

According to Bloomberg, the KOSPI 200’s 12-month forward P/E ratio has plummeted to a 6X PER, trading at a deep discount to its 20-year historical average of 10X. Ultimately, this sell-off means a decent trading opportunity via the Global X KOSPI 200 ETF(3408 HK).