Samsung's HBM4E yield rate exceeds 70%, the development of the seventh generation AI memory enters a stable stage.

date
15:01 01/07/2026
avatar
GMT Eight
In the industry, usually more than 80% is considered the threshold for "mature yield" of a process. Currently, HBM4E is still in the reliability testing phase, and a level of over 70% is considered to indicate that the development process has officially entered a stable range.
Samsung Electronics has successively achieved phased results in AI memory technology. After successfully achieving mass production of the sixth generation high bandwidth memory HBM4, Samsung has reported positive progress in the development of the seventh generation HBM4E and next generation DRAM technology, laying the foundation for consolidating its technological leadership in the next round of AI memory competition. According to the semiconductor industry news, Samsung Electronics Chief Technology Officer and Semiconductor Research Institute Director Song Jai Hyuk stated at an internal management briefing for the Device Solutions (DS) department on June 30 that the reliability test yield of HBM4E has now increased to over 70%. The industry generally considers above 80% as the threshold for "mature yield" in terms of process stability, and HBM4E is currently still in the reliability testing phase, with a yield of over 70% being seen as a sign that the development process has officially entered a stable range. At the same time, Song Jai Hyuk revealed at the same event that the next generation 10-nanometer seventh generation DRAM technology (D1d) has already gained a competitive edge over competitors in terms of technological competitiveness, and plans to complete Production Readiness Assessment (PRA) in November this year. These two developments combined have led the market to have a more optimistic expectation for Samsung's competitiveness in the next generation AI memory race. HBM4E development accelerated, sample evaluation progressing smoothly Samsung first achieved mass production of HBM4 in February this year, and on May 29, further detailed technical specifications of the HBM4E 12-layer product were publicly announced, with samples sent to key customers. HBM4 will be mounted on NVIDIA's AI accelerator "Vera Rubin" to be released in the second half of this year, and HBM4E as its successor product is planned to be assembled on NVIDIA's next-generation AI accelerator "Vera Rubin Ultra" to be released next year. With the steady progress of sample evaluations by major customers, the mass production development of HBM4E is proceeding smoothly in sync. In this context, the increase of reliability test yield to over 70% is of significant importance. Although this number has not yet reached the industry-recognized threshold for mature yield, considering the current development stage of the product, it is generally interpreted as a positive signal that HBM4E is accelerating toward the conditions for mass production. D1d process advancing, aiming to pass production certification in November Regarding the development of the next generation DRAM technology, Song Jai Hyuk stated that the technical competitiveness of the D1d process is ahead of competitors, with the goal of obtaining Production Readiness Assessment (PRA) in November this year. PRA is the final internal quality evaluation procedure before product shipment, comprehensively verifying yield, performance, and production efficiency to determine if mass production conditions are met, and once passed, can formally switch to mass production mode. The strategic significance of the D1d process goes beyond just DRAM itself. Samsung plans to fully introduce this process from the eighth generation HBM5 onwards. The industry believes that if the development of D1d proceeds as planned, it will have a chain positive effect on the overall competitiveness of the next generation DRAM and subsequent HBM5 products. Contradictions in R&D personnel compensation distribution emerge Alongside the announcement of technological progress, another layer of voices has emerged within Samsung's DS division. It is reported that after this briefing, there has been a rise in dissatisfaction within the R&D organization regarding the identification of roles and the compensation structure of R&D personnel, with some members calling for a more active recognition of the contribution of the R&D organization by the company. Samsung had previously reached an agreement with the DS division to allocate funds from 10.5% of operating profit to establish a "Special Operational Performance Bonus" system. However, within the DS division, there is a significant gap in performance bonuses between the storage business and the public divisions including the research institute, as well as non-storage businesses (System LSI and foundry outsourcing), and the calls for improving the compensation structure are continuing to grow. This article was reprinted from "Wall Street News", edited by GMTEight: Chen Siyu.