China's CXMT Tops Market Cap on Debut, Chases Samsung and SK in DRAM

*Editor's Note: 'AI PRISM' (Personalized Report & Insight Summarizing Media) is an artificial intelligence (AI)-based customized news recommendation and summary service developed with support from the Korea Press Foundation. It selects and provides six customized news items by reader type.

1. CXMT Seizes Technology and Capital... "Could Overtake Micron Within 4 Years"

- Key Summary: CXMT, China's largest DRAM maker, listed on the Shanghai Stock Exchange's STAR Market on the 27th and closed at 49 yuan, up approximately 465.82% from its offering price, overtaking Industrial and Commercial Bank of China to take the top spot by market capitalization on the mainland market. This IPO was the largest in Asia's stock markets this year, and CXMT plans to invest up to 66.61 billion yuan (about 14.435 trillion won) raised in expanding production lines and advancing DRAM technology. Nomura Securities forecast that CXMT's market share expansion would accelerate further, while Goldman Sachs projected that CXMT's production capacity could double from current levels by 2030 and overtake Micron. However, some assessments note that CXMT remains at the 16-nanometer-class DRAM mass production stage, putting it 2 to 3 years behind the 10-nanometer-class processes of Samsung Electronics and SK hynix (000660.KS).

2. Naver Secures GPUs and Investment Simultaneously... Moves to Preempt Global AI Factory

- Key Summary: Naver announced on the 27th that it has established an AI infrastructure cooperation framework totaling $10 billion with Nvidia and Brookfield. Nvidia will acquire a 4.5% stake in Naver through a third-party allotment paid-in capital increase, while Naver signed a $9 billion AI infrastructure procurement contract with Brookfield, marking its largest external procurement since its founding. Based on its full-stack AI capabilities of directly operating data centers, GPU clusters, and AI platforms, Naver plans to trial-operate a 55-megawatt (MW) AI factory in the first half of 2027 and expand it to 200 megawatts by 2028. Naver CEO Choi Soo-yeon said the company would accelerate its efforts to attract AI factory clients on the back of this investment.

3. Samsung Overhauls 20-Year-Old Cleanroom, Integrates Front-End Equipment... Boosts Wafer Output by 200,000 per Month

- Key Summary: Samsung Electronics is converting a 20-year-old cleanroom at Complex 1 (H1) of its Hwaseong site into a dedicated end-fab for commodity DRAM, consolidating end-fab lines scattered across Complex 2 (H2) in Hwaseong and Cheonan into one location. According to market research firm DRAMeXchange, the average trading price of commodity DRAM (based on DDR4 8Gb) rose more than 80% in the first half of this year, while DDR5 (based on 64GB modules) recorded an increase of up to 190%. With Apple, Samsung Electronics' key customer, beginning to review alternative suppliers such as CXMT, Hwaseong H1 is assessed to offer high investment efficiency as it can utilize existing infrastructure. The industry interprets the unification of Hwaseong production facilities as a measure to increase short-term shipments and profitability while responding to competition with Chinese firms.

4. Broadcom, Once Blocked from Qualcomm Takeover by Trump... Grows into 'Key Anti-Nvidia Player'

- Key Summary: Broadcom is drawing attention again following an AI summit held in San Francisco, U.S., last weekend. As of the 26th, Broadcom's market capitalization stood at $1.82 trillion, ranking second among U.S. semiconductor firms after Nvidia (excluding TSMC). Broadcom CEO Hock Tan saw the Qualcomm takeover fall through during the first Trump administration, but has since forged custom chip (ASIC) development partnerships with Google, OpenAI, Anthropic, and Meta, growing into a key partner in the anti-Nvidia front. According to Forbes, custom tensor processing units (TPUs) involved with Broadcom are known to be lower in price and 67% more energy-efficient than Nvidia's top-spec GPUs.

5. 38% of Office Workers "Use Personal AI"... 'Shadow AI' Risk Hits Companies

- Key Summary: In a survey conducted by Samsung SDS of 1,750 employees at companies and public institutions, 38% of respondents said they personally subscribe to generative AI not approved by their company and use it for work. This is 14 percentage points higher than the proportion (24%) who use enterprise versions contracted and provided by companies. The proportion of respondents who said they use generative AI for more than two hours a day quadrupled from 18% late last year to 72% this year. Unlike enterprise versions, personal AI does not block data training or control access rights, raising concerns about information leaks. The industry points out that strengthening security education and adopting solutions are necessary.

6. LG Electronics (066570.KS) Liquid Cooling Solution Wins First Domestic Nvidia Certification... Targets AI Data Centers

- Key Summary: LG Electronics announced on the 27th that it has become the first Korean company to obtain Nvidia's quality certification for a high-efficiency liquid cooling solution for AI data centers. The certified 600-kilowatt (kW)-class coolant distribution unit (CDU) implements a direct-to-chip method that directly absorbs heat from high-bandwidth memory (HBM) and high-performance graphics processing units (GPUs). With the global AI data center market expected to grow to about $133.5 billion (about 185 trillion won) by 2034, LG Electronics plans to pursue expanded certification across its full lineup of ultra-large CDUs, including 1-megawatt (MW), 2.5-megawatt, and 4-megawatt units. Lee Jae-sung, President and head of LG Electronics' ES Business Division, emphasized that the company would become a total solution provider leading the global market based on group-wide integrated infrastructure capabilities.