Changxin Technology rises over 6%

Serenity released a report stating that Samsung and SK Hynix’s memory inventories have fallen to less than 10 days of supply, with a potential severe shortage next year. KB Securities forecasts the memory market could face "the tightest supply conditions in history" in 2027. In this regard, Serenity pointed out that KB Securities projects the proportion of memory costs in AI infrastructure spending will rise from around 40% this year to 57% next year, while TrendForce’s forecast is even higher at 68%. Meanwhile, related storage stocks have dropped roughly 38% from their highs over the past three months, and their price-to-earnings ratio has fallen to around 3 times based on next year’s expected earnings. On the NAND front, market commentary surrounding Phison also points to extremely tight supply in 2027. Serenity noted that the surge in NAND prices has forced some Chinese consumer-grade storage module makers to clear inventories, creating an opportunity for Phison to purchase stock at low prices—illustrating that localized price declines and overall supply shortages can coexist. Serenity said that against the backdrop of growing AI demand and storage’s rising cost share in AI infrastructure, South Korean storage stocks and U.S.-listed related companies including Micron (MU) and SanDisk (SNDK) may benefit from 2027’s supply-demand tightness and growth in infrastructure spending.

TrendForce released its latest memory industry research, showing that in Q2 2026, driven by a sharp rise in contract prices for conventional DRAM, overall DRAM industry revenue rose 59.5% quarter-on-quarter to approximately $154.73 billion. On the demand front, as LLM model training and AI inference continue to fuel demand for AI servers, shipments of HBM3e, LPDDR5X, and high-capacity RDIMM have grown in parallel; agentic AI applications have further driven procurement demand for RDIMMs of various capacity specifications. On the supply side, TrendForce stated that DRAM manufacturers currently hold low inventories, and new production capacity is prioritized for the server market, leading to only a modest increase in overall DRAM bit shipments in the quarter. The pace of supply expansion remains slower than demand growth spurred by AI and other applications.

Per Blockaid's monitoring, DeFi risk management protocol Cozy Finance’s deployment on the Optimism blockchain is facing a sustained attack. Attackers have so far stolen around $170,000 in assets. Blockaid has since released preliminary attack details, flagging the malicious transactions and linked attacker addresses. Multiple attacker addresses have been confirmed, plus one token address suspected to have been exploited in the breach. It remains unclear whether the attack is still ongoing, and users are advised to exercise caution when interacting with Cozy Finance-related smart contracts.

According to Bitget market data, South Korea’s KOSPI index closed up 308.19 points on Monday, September 7, with a 4.61% gain to end at 6995.4 points. SK Hynix rose 8.26% in closing, while Samsung Electronics gained 5.68%.

According to HTX market data, Ethereum (ETH) has fallen below the $2500 threshold, currently trading at $2497.3, with a daily gain of 0.79%.

Jefferies released a report noting that Zhipu AI’s 2026 annual recurring revenue (ARR) guidance of $2.4 billion exceeds expectations, though its sustainability is in doubt due to a high August base, uneven growth in computing power supply, high customer concentration, and low switching costs. The report adds that while Zhipu Cloud’s gross margin improved in the first half of 2026, it may decline in the second half amid the launch of new domestic GPU clusters and the resumption of its Coding Plan. Jefferies cut Zhipu’s target price to HK$1,183.79 from HK$1,299.8, maintaining a “Hold” rating. The bank revised Zhipu’s 2026–2029 revenue forecasts upward by 37% to 119%, reflecting accelerated growth in its cloud segment, and lowered net loss projections by 14% to 21%. In its sum-of-the-parts valuation, the cloud segment’s multiple was reduced from 50x its 2026 forecasted ARR to 30x to better align with overseas peers. Jefferies still views China’s large language model (LLM) sector as overcrowded, and favors full-stack cloud service platforms with advantages in computing power, data, and monetization over independent AI labs, such as Alibaba (09988.HK) and ByteDance.