Samsung discusses collaboration, SK Hynix expands production, Kioxia samples NAND, Japanese and Korean stock markets respond to computing power surplus panic with a V-shaped rebound.
After experiencing two consecutive days of intense sell-offs, Japanese and Korean tech stocks saw a strong rebound. Market sentiment was boosted by multiple positive developments in the industry, which led to a temporary reversal of the previous trading logic around "AI computing power surplus".
On July 3, South Korea's KOSPI index plunged over 3% during early trading, then rapidly surged to rise 5%, triggering the circuit breaker for algorithmic trading. Japan's semiconductor sector also climbed sharply, showing an overall deep V-shaped recovery. Among individual stocks, both Samsung Electronics and SK Hynix jumped over 8%, while Kioxia surged more than 10% intraday, driving a comprehensive rebound in Japanese and Korean semiconductor sectors. By market close, the Nikkei 225 rose 1.5%, TOPIX index gained 1.2%, and Korea's Seoul Composite Index soared 5.8%.
The direct catalyst for the rebound came from several new developments in the AI industry chain. It is reported that AI startup Anthropic is negotiating customized AI chip cooperation with Samsung Electronics, boosting market expectations for Samsung’s wafer foundry business. Meanwhile, Samsung and SK Hynix continue to advance AI semiconductor capacity expansion plans and have announced new investments; Kioxia declared its tenth-generation 3D NAND chip has been sampled to AI data center customers, indicating demand remains strong.
However, as AI investment enters a new phase focused more on return and capital efficiency, the market’s cautious assessment of supply-demand balance and capacity expansion pace in the industry chain persists, and related divergences have not been fundamentally resolved.
AI Catalysts Lead the Rebound: Samsung's Strategic Cooperation and Performance Expectations Resonate
The current market rebound first stems from new catalysts in the AI industry chain.
According to reports, Anthropic is discussing cooperation with Samsung Electronics to jointly develop customized AI chips. Although the collaboration is still at an early stage, the market believes this means Samsung’s wafer foundry business has the opportunity to further participate in the AI chip ecosystem.
Jung In Yun, CEO of Fibonacci Asset Management Global, said the short-term profit contribution of this cooperation is limited, but its strategic significance is positive, helping reinforce Samsung’s important position in the AI chip sector and highlighting Asia’s role in the AI semiconductor supply chain.
At the same time, the market is also beginning to trade ahead of Samsung’s upcoming preliminary quarterly results. Analysts generally expect the company’s second-quarter earnings to continue growing sharply, and investors are paying closer attention to management’s assessment of sustained demand for AI memory.
Korean Leaders Continue Capacity Expansion, Responding to Market Concerns over AI Demand
In sharp contrast to previous worries from the capital market, Korean semiconductor companies are still expanding their investments.
On July 2, Samsung Electronics and SK Hynix once again announced new plant construction plans. Public information shows Samsung plans to invest about 140 trillion won in the Chungcheong region of Korea, covering HBM wafer plants, high-performance packaging, OLED, and next-generation batteries. SK Hynix plans to invest about 100 trillion won, focusing on NAND and advanced packaging.
Meanwhile, the Korean government announced it will promote business investments of over 312 trillion won in southeastern regions, focusing on semiconductors, AI, and aerospace industries. Major companies involved include SK Group, Samsung, Hanwha, and Hyundai Motor.
From the capital market’s perspective, although there is short-term volatility, Korean memory chip leaders are opting to continue investing, aiming to stabilize expectations for long-term AI growth.
Industry insiders also believe Meta’s recent leasing of idle computing power should be interpreted as resource optimization, rather than a turning point in AI infrastructure demand. As AI companies put greater emphasis on cost control in future, the trend toward self-developed chips may become even stronger.
Samsung Drives DRAM Price Increase, Long-Term Contracts Reinforce Profit Expectations
Aside from demand expectations, price has also become a key focus for the market.
According to South Korea's ZDNet, Samsung Electronics is negotiating third-quarter DRAM prices with customers, aiming to raise the average selling price of generic DRAM by up to 20% compared to the second quarter, while server and mobile LPDDR product price increases may exceed 20%.
Industry insiders indicate that ongoing AI infrastructure construction keeps demand for server DRAM, high-bandwidth memory (HBM), and LPDDR tight, and supply pressures are unlikely to ease in the short term.
Notably, long-term supply agreements (LTA) are becoming an important foundation for stable industry profitability.
Micron recently disclosed it has signed 16 long-term supply agreements with customers. These agreements not only lock in procurement quantities but also set price floors, helping reduce future risk of sharp price drops. With the proportion of long-term contracts rising, insiders believe the likelihood of a significant downward trend in the DRAM market next year is low.
However, some insiders note that Samsung is taking a tough stance in price negotiations, and whether customers can fully accept the price increase plans remains to be seen.
Kioxia Launches New-Generation NAND, Betting on AI Data Center Demand
Japanese memory maker Kioxia chooses to respond to market doubts with new products.
The company announced its tenth-generation BiCS FLASH 3D NAND has begun sampling to AI data center clients and plans to start mass production in 2027. The new product adopts a 332-layer stacked architecture and proprietary CBA technology, achieving about 60% higher storage density than the previous generation, with interface speeds of 4.8 Gbps.
Reports say Kioxia believes that compared to over-400-layer designs, the 332-layer architecture offers a better balance among cost, power consumption, and reliability.
CEO Hiroo Ota stated the company sees no signs of weakening demand from data centers and will continue actively responding to market growth, not ruling out further increases to capital expenditure. He believes that as AI agents and robotics applications develop, the flash storage market will still have significant growth opportunities.
However, competition in the market is also intensifying.
According to calculations by Omdia analyst Akira Minamikawa, in 2025, Samsung Electronics will hold about 40% of the data center NAND market, SK Hynix about 30%, and Kioxia about 10%. Korean manufacturers, relying on one-stop sales capabilities built around HBM products, remain a major competitive pressure for Kioxia.
Meanwhile, SK Hynix plans to invest in new NAND production facilities, and Samsung is also planning new NAND production lines. The three leading manufacturers are expanding capacity simultaneously, which means future market attention will continue to focus on supply-demand relations and price trends.
Market Begins to Reevaluate AI Investment Returns
Although this rebound has eased market panic, industry consensus is that the logic behind AI industry investment is changing.
Analysts state that the industry is shifting from the previous "bottomless purchasing of AI chips" to paying more attention to investment returns. In the future, cloud service providers will procure memory chips and other products more carefully, and AI infrastructure investment will enter a new stage prioritizing efficiency.
After this week’s dramatic volatility, whether the AI industry chain can continue to maintain high growth expectations will remain a core variable for market focus.
Risk warnings and disclaimersThe market carries risks, and investment requires caution. This article does not constitute personal investment advice and does not account for the specific investment objectives, financial situation or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article suit their particular circumstances. Investing based on this article is at your own risk.