[News] Micron Plans HBM4 Mass Production in 2026, Customized HBM4E to Launch in 2027-2028

---

---

A report from TechNews, citing Tom’s Hardware, indicates that in response to the rapidly growing demand driven by AI, SK hynix, Samsung, and Micron are all accelerating the development of HBM products, and recently Micron has revealed plans for its HBM4 and HBM4E. Micron shows in its AI memory portfolio that its next-generation HBM4 is expected to enter mass production in 2026, followed by HBM4E in subsequent years, in 2027-2028.

Notably, Micron’s plans to begin high-volume production of HBM4 in 2026 coincide with the expected launch of NVIDIA’s Vera Rubin GPUs and AMD’s Instinct MI400-series GPUs, as the report from Tom’s Hardware points out.

The report notes that its HBM4E not only offers higher data transfer rates but also introduces the option of customized base dies. These custom logic dies are anticipated to be manufactured by TSMC using an advanced node, enabling them to incorporate more caches and logic for enhanced performance, as highlighted in the report.

The exact details of Micron’s plans to customize its HBM4E base logic dies remain unclear, and it is uncertain whether these customizations will align with the JEDEC standard. Nonetheless, Micron has announced that its HBM4E development is progressing smoothly and that it is actively collaborating with multiple customers, as noted by the report.

The report notes that these collaborations are expected to result in several configurations tailored to clients’ specific base die requirements, representing a significant step toward personalized memory solutions for bandwidth-intensive applications such as AI, HPC, and networking. Also, Micron’s HBM4E may also support Marvell’s customized HBM computing architecture, enabling more effective customization of various XPU and HBM solutions.

Different Process Technology for HBM4 Production Compared to Samsung and HK hynix

Micron’s HBM4 will use DRAMs manufactured with its proven 1β (5th Generation 10nm-class) process technology, as noted by the report. However, both Samsung and SK hynix are rumored to utilize 6th Generation 10nm-class manufacturing technology for their HBM4 products. This raises questions about the market competitiveness of Micron’s HBM4, according to the report.

Aside from its work on HBM4, Micron also disclosed great progress on shipments of its HBM3E 8H, which is set to be used in NVIDIA’s B200 and GB200 platform. Furthermore, the report from Tom’s Hardware indicates that its HBM3E 12H is set to be used by AMD’s Instinct MI325X and MI355X accelerators, and NVIDIA’s Blackwell B300-series compute GPUs.

Meanwhile, a previous report by Maeli Business Newspaper indicates that Samsung aims to begin mass production of HBM4 promptly upon completing development by the end of 2025. As for SK hynix, the Korea Economic Daily reports that the company plans to produce its sixth-generation HBM4 chips using TSMC’s 3nm process technology, and these chips are slated for delivery to NVIDIA in the second half of 2025.

Read more

• [News] Micron Clouded by Weak Q2 Outlook; High-volume HBM3E 8H Ships to Second Major Customer this Month
• [News] Marvell Develops Custom HBM with Micron, Samsung, and SK hynix

(Photo credit: Micron)