According to South Korean media reports, Samsung Electronics has signed an MOU with Asan City in South Chungcheong Province, finalizing plans for the expansion of its Onyang plant. The new plant is scheduled to break ground in October of this year, with the goal of achieving mass production of HBM by May 2029. The project is reportedly the first of South Korea's three government-designated "Mega Projects" to materialize, with a total investment of 46 trillion won, of which 6 trillion won will be allocated to building the HBM production lines.

SK Hynix has increased its investment in the Cheongju P&T7 project to 7.0931 trillion won (approximately $4.8 billion) to proactively address the growing demand for AI memory.

According to South Korean media reports, Samsung Electronics is expanding its alliance with Nvidia into NAND flash supply. Industry sources indicated that Samsung Electronics has built V-NAND production capacity of more than 100,000 wafers per month, and is mass-producing its 10th-generation V10 product for supply to Nvidia. It is said that Samsung has allocated about 60% of its V-NAND production capacity to V9, which intended to meet the NAND demand for NVIDIA CMX in the second half of the year. In addition, trial production of the 500-layer-class V11 has begun earlier this year, and Samsung plans to double related volumes in the second half.

According to Korean media citing industry sources, Samsung Electronics' Memory Business has recently set up a dedicated backend process organization within its etching technology team.

According to South Korean media reports, Samsung Electronics has mostly confirmed the semiconductor equipment companies that will build the P5 Phase 1 line, and the process of selecting suppliers for additional lines after Phase 2 is progressing simultaneously. Equipment orders for P5 will begin in earnest. Orders for semiconductor equipment on the scale of tens of trillion won are imminent. Originally, the target for full operation of the P5 line was known to be 2028, but the first line (Phase 1) is expected to be operational as early as next year.

According to a report by ZDNET, SK hynix is pushing forward with its semiconductor cluster construction in Yongin on an accelerated timeline exceeding initial expectations. Recently, the company has placed orders with key partners for cutting-edge DRAM manufacturing equipment. The initial production scale under discussion is 20,000 wafers per month.

According to a report by South Korean media The Bell, SK hynix began shipping mass-production quantities of 12-layer HBM4 to NVIDIA at the end of last month, entering the "ramp-up" phase. This marks the first time HBM4 has been supplied to NVIDIA's next-generation AI platform "Vera Rubin" in its final spec, having completed all quality certifications. All previous shipments were classified as samples. It is also reported that SK hynix has completed price negotiations with NVIDIA for next year's HBM4 supply, with pricing coming in lower than Samsung's due to cost advantages. Following the shipments to NVIDIA in last month, SK hynix has also begun shipping 8-layer HBM4 products to AMD starting this month. SK hynix expects to gradually increase HBM4 shipment volumes from the second half of the year, with the company projecting that HBM4 will surpass HBM3E in its overall HBM sales mix by the fourth quarter.

As they advance next-generation High Bandwidth Memory (HBM) technology, Samsung Electronics and SK Hynix are facing a critical decision: whether to fully adopt hybrid bonding technology. Despite its significant advantages, the two giants have become increasingly cautious in their rollout, as the immediate demands for HBM "thinning" and "heat dissipation" have eased.

According to media reports, Nanya Technology stated at its investor conference that next-generation AI accelerator computing platforms such as Rubin Ultra and TPA are driving up HBM bit demand and continuing to crowd out traditional DRAM capacity. The company expects global DRAM to remain in short supply through 2028. Nanya Technology estimates that its capacity will expand by approximately 69% from 2026 to 2028. Its new 5A fab will introduce 1C/1D nm advanced process nodes and add about 45,000 wafers in new capacity. Bit shipments are expected to grow 8% year-over-year in 2027 and 53% in 2028, reaching 6.75 billion Gb and 10.33 billion Gb, respectively. In addition, because the average price per Gb in Q2 has not yet fully reflected the contract price increases, and there is a timing lag in contract renewals with major long-term customers, the company sees room for upward revision in its Q3 ASP.

Intel has unveiled a new patented technology called XBM, which is designed as an alternative to HBM4 and offers higher bandwidth capabilities. XBM stands for Extended Bandwidth Memory and is a DRAM memory solution. Its package size will remain consistent with HBM4, with each chip's capacity ranging from 0.5GB to 5GB, featuring a 32GT/s UCIe universal chiplet interconnect interface. The core breakthrough lies in the migration of the manufacturing process. Traditional DRAM memory cells are fabricated in the front-end-of-line (FEOL), which is the underlying silicon substrate natively used for transistor fabrication. In contrast, XBM moves the 1T1C memory cell to the back-end-of-line (BEOL), placing it within the metal via stack region above the transistors, and employs thin-film transistor (TFT) technology.

According to Korean media and industry sources, Samsung Electronics has decided to postpone the mass production of its CXL 3.1-based memory modules (CMM-D 3.0). This decision is primarily driven by the delayed release of CPUs supporting the CXL 3.1 standard by manufacturers such as Intel and AMD. In the short term, Samsung will continue to maintain the production of its CXL 2.0-based CMM-D 2.0 products.

According to media reports, Micron Technology held a groundbreaking ceremony on Saturday for the expansion of its plant in western Japan. The project involves an investment of 1.5 trillion yen (equivalent to $9.3 billion) to produce advanced memory chips, including high-bandwidth memory (HBM), primarily to meet demand from AI processors. Shipments of the related products are expected to begin around summer 2028. It is reported that Japan's Ministry of Economy, Trade and Industry will provide up to approximately 500 billion yen in subsidies. The project is part of Micron's global capacity expansion plan for AI memory, while the company is also advancing large-scale advanced-process investments in Idaho and New York to boost DRAM and HBM supply capabilities.

According to THE ELEC citing industry sources, SK hynix is in talks with equipment suppliers regarding the supply of semiconductor test equipment for its Cheongju P&T7 plant. Equipment suppliers are currently coordinating estimated delivery volumes for next year. It is estimated that the plant will require approximately 200 test equipment units, including HBM4 testers. Based on a unit price of KRW 1.5 billion to KRW 2 billion per machine, the total equipment value could reach as high as KRW 400 billion. Industry insiders indicated that SK hynix had initially planned to order around 200 test equipment units this year, but later reduced the order volume and deferred some orders to the second half of the year. The deferred demand may be concentrated in next year's P&T7 project.

According to media reports, Samsung Electronics' CTO and Head of the Semiconductor Research Institute, Song Jai Hyuk, stated at an internal DS Division business briefing that the reliability test yield for HBM4E has surpassed 70%. Regarding next-generation DRAM process development, Song Jai Hyuk indicated that the D1d process has achieved technological competitiveness ahead of rivals and has set a development goal of obtaining Production Readiness Approval (PRA) by November this year.

According to public information, Samsung Electronics recently filed a new patent for High Bandwidth Memory (HBM) packaging, aiming to address the reliability challenges faced by high-stack products such as HBM4E and HBM5.