Chinese DRAM maker ChangXin Memory Technologies, or CXMT, officially listed on Shanghai’s STAR Market on July 27, 2026, reaching an intraday market capitalization of about RMB 3.35 trillion, according to 8Marketcap. The valuation is a sharp jump from its IPO value of RMB 579.2 billion and comes despite accumulated uncovered losses of RMB 36.65 billion as of the end of 2025. The company’s turnaround story traces back to the 2016 launch of the “506” domestic DRAM project in Hefei under the leadership of GigaDevice founder Zhu Yiming, backed by Hefei state capital and later a mix of brokers, market investors, bank-affiliated AIC funds, and internet companies. After years of heavy spending on fabs, process development, and yield improvement, CXMT posted a first-quarter 2026 revenue of RMB 50.8 billion and attributable net profit of RMB 24.762 billion, helped by a surge in DRAM contract prices and AI-related memory demand. The report also notes that the valuation carries substantial domestic substitution and AI premium, while process gaps, higher bit costs, customer concentration, and remaining historical losses still weigh on the company’s longer-term fundamentals.

Nomura has initiated coverage on Chinese DRAM maker ChangXin Memory Technologies, or CXMT, with a Buy rating and a target price of RMB 116, according to a report discussed in a ChainCatcher article by “The Dream of the Fourth Dimension.” The front page of the report also showed an IPO price of RMB 8.66 and implied upside of 1,239.5%, making it, in the author’s view, the most aggressive target price so far from a foreign institution on the company. The target is based on 20x projected 2028 earnings per share of about RMB 5.8. Nomura anchors that multiple to Micron’s historical forward valuation range, then adds what it sees as an A-share premium using ACM Research Shanghai versus its U.S.-listed parent as a reference. The report also projects a steep jump in revenue, net profit attributable to shareholders, gross margin, cash, and return on equity through 2028. Its broader thesis rests on rising memory demand from agentic AI, supply bottlenecks across cleanrooms, tools, materials and engineers, and CXMT’s own capacity ramp, yield improvement, and share gains. The original article, however, argues that the model’s most aggressive assumption is gross margin reaching 83.7% in 2026 and more than 90% by 2028, and suggests the figures are better read as an upside case than a long-term midpoint.

Anthropic is negotiating with Microsoft to lease its in-house AI inference chip Maia 200, potentially becoming a key external customer. The firm has committed over $330 billion to three cloud giants, with 2026 compute costs expected to exceed $20 billion, pursuing a multi-chip strategy unlike OpenAI's Nvidia reliance.

A TrendForce research note says the memory hierarchy behind AI systems is being redrawn as the industry shifts from training to inference. High Bandwidth Memory, or HBM, dominated the training era because it solved a bandwidth problem. In inference, the pressure point has changed. Key-value cache growth during prefill and decode is pushing capacity to the forefront, creating room for three different approaches: HBF for larger-capacity memory between HBM and SSD, SSD POD for offloading cache into lower-cost storage tiers, and SRAM for ultra-fast on-chip access in decode-heavy workloads. The report walks through how each technology fits into the stack rather than framing them as direct substitutes. HBF, being developed by SanDisk and SK hynix, is aimed at much larger capacity than HBM but has not reached mass production. NVIDIA’s SSD POD concept, tied to its Dynamo framework and NIXL transport library, is designed to move KV cache from limited GPU memory into CPU RAM, local SSDs, and remote storage without stopping inference. SRAM, used aggressively by companies such as Groq and Cerebras, trades capacity for speed by keeping memory on chip. The piece also points to recent product moves from NVIDIA, Google, SambaNova, Cerebras and Etched as evidence that inference-focused hardware design is accelerating.

SK Hynix is set to list on Nasdaq next week in a $29 billion deal aimed at drawing U.S. capital into the AI memory trade, with the offering seen as a potential record foreign IPO.

South Korea's KBS documentary gains rare access to SK Hynix's M16 fab, revealing fully automated HBM production lines. One FOUP of 25 wafers is valued at 25 Genesis luxury cars. The company's success has made it the No.1 dream employer, with college graduates even hiding degrees to apply for technician jobs. A corporate vest has become a top dating accessory.

Melius Research initiated coverage of Micron and SanDisk with Buy ratings, setting two-year price targets of $700 and $1,350 respectively. Analyst Ben Reitzes highlights that AI-driven demand is shifting the memory industry to a 'reverse SaaS' long-term contract model, boosting revenue visibility.

Samsung's largest union NSEU demands management respond by Friday 10am or face an 18-day strike starting May 21. The walkout could disrupt DRAM and HBM supply for AI chips and memory markets.