hbm

According to South Korean media ZDNet Korea, Samsung Electronics is adjusting the original purpose of its NRD-K Line 2 at the Giheung campus from a research and development facility to a wafer foundry line, with the main target product being the 2nm base chips required for NVIDIA's HBM. This production line is expected to be operational in the second half of 2028, adopting a "Send-Fab" model, which involves receiving wafers from other mass production lines and performing specific processes.NRD-K is the most advanced composite research and development campus built by Samsung Electronics to seize future semiconductor technology, with a total investment of approximately 20 trillion won, planning for three production lines, of which the first line is scheduled to be completed by the end of 2024. This adjustment aims to proactively respond to the expansion of HBM demand driven by investments in AI infrastructure. Samsung plans to first apply the 2nm process to custom HBM and HBM5 to maintain performance advantages.Industry insiders point out that the scale of NRD-K Line 2 is relatively small, making it suitable for operation in the Send-Fab format. However, there are still about two years until the production line is operational, and the final purpose may be adjusted again based on market conditions. Samsung had previously considered using this production line for DRAM production, but after modifications to the relevant plans this month, DRAM capacity will be more concentrated in the Pyeongtaek campus. Equipment procurement orders have not yet been officially issued.

Micron Technology stated to Deutsche Bank at the FMS 2026 conference that AI is reshaping the memory industry landscape, with the proportion of memory in the total value of systems rising from about 10% thirty years ago to nearly 50% currently, and this trend is accelerating. Currently, both DRAM and NAND are in a state of supply shortage. Due to the need for physical disassembly of systems for memory upgrades, the demand for memory in AI systems is more rigid, and price elasticity is lower than market expectations. The Strategic Customer Agreement (SCA) is expected to cover about 40% of sales, providing a price stability mechanism for both parties.Micron's management described the demand for CPU-side AI Agents as "pre-season warm-up," viewing it as a new pillar of growth outside the GPU ecosystem. For emerging technologies like SRAM, Micron believes their scalability is limited and cannot challenge the core position of HBM, with only about 5% of systems currently running such specific workloads. Deutsche Bank maintains a "Buy" rating, believing Micron has a unique advantage of "growth without sacrificing profits." During FMS, Sandisk and SK Hynix jointly released the first HBF technical specifications, but Micron believes that its actual value and commercialization prospects remain controversial.

According to the latest research by TrendForce, the tight supply of DRAM is expected to continue until 2027, and the uncertainty of the HBM4e verification schedule is prompting AI chip manufacturers to consider reducing the HBM configuration of their next-generation products. Starting from Q3 2026, NVIDIA has begun evaluating the HBM solution for Rubin Ultra, expanding from the original 12-Hi HBM4e to alternative options such as 8-Hi HBM4e, 12-Hi HBM4, and 8-Hi HBM4, with the final specifications yet to be determined. Previously, due to the expected supply bottleneck of LPDDR5X lasting until 2027, NVIDIA decided to halve the SOCAMM capacity of the next-generation Vera Rubin Superchip module.TrendForce pointed out that the main reasons for the downgrade of Rubin Ultra from 12-Hi HBM4e include: the overall DRAM shortage in 2027 will limit the allocation of wafer capacity to HBM, and there is still uncertainty regarding the mass production schedule and yield ramp-up of 12-Hi HBM4e. It is expected that the HBM bit shipment volume in 2027 will increase by approximately 50-60% year-on-year, which will still be insufficient to meet the growth in demand. HBM suppliers will maintain pricing power throughout 2027, and HBM prices are expected to rise significantly. The final configuration will depend on the wafer allocation decisions of memory suppliers.

According to DIGITIMES, industry insiders revealed that the three major memory manufacturers have completed the capacity allocation negotiations for the entire year of 2027 in advance, with both DRAM and HBM capacities sold out. In terms of NAND Flash, although it is not as tight as DRAM, it is expected that the capacity will also be fully booked by the end of August 2026. ADATA Chairman Chen Libai confirmed that applications related to HBM and AI servers will account for nearly 70% of DRAM capacity, while the supply of DRAM for PCs and mobile phones will be relatively limited.Industry assessments indicate that Samsung, Micron, and SanDisk's NAND capacity for the entire year of 2027 has also been pre-sold, with Kioxia and SK Hynix expected to confirm by the end of August. SK Group Chairman Chey Tae-won previously stated that the demand for AI semiconductors in 2027 is expected to increase by 60-100% compared to 2026, with storage demand expected to grow by 50-60%, leading to a potential widening of the supply-demand gap and possibly facing "the most severe shortage in history." Industry insiders pointed out that the capacity allocation model is more orderly compared to 2026, but the amount that manufacturers can provide is usually only 60-70% of customer demand, leaving some small and medium buyers still facing the dilemma of having no products to purchase. The price increase of memory may slow down in 2027, but remaining high could become the norm.

Samsung Electronics and SK Hynix are accelerating their layout of HBM and next-generation storage technologies to maintain their leading advantage. Samsung Electronics stated that its R&D investment reached a record 16 trillion won (approximately 11.6 billion USD) in the second quarter and plans to expand its layout of high-end storage products such as HBM4, HBM4E, and DDR5.SK Hynix is developing next-generation technologies such as HBM5, hybrid bonding, and iHBM, with capital expenditures expected to reach 40 trillion won (approximately 32.6 billion USD) this year.

According to Jinshi reports, on Friday evening (July 10) Eastern Time, SK Hynix CEO Kwak Noh-jung predicted that 2027 will be the year with the tightest supply in the history of the storage industry."Now customers are coming to SK Hynix seeking long-term supply agreements. Demand continues to grow, while our production capacity is limited. Although we are doing our utmost to expand capacity, even after 2030, customer demand may still exceed the company's supply capability," he said.However, domestic investment professionals who have had contact with South Korean storage manufacturers indicated that, in addition to their own development, executives from overseas storage giants are also paying attention to the expansion progress of Chinese storage manufacturers. The domestic "two storage" (Changxin Storage and Yangtze Memory Technologies) may become the biggest variable in the future storage market.