
According to reports from Korean media outlets The Elec and Herald Economy, citing industry sources, Micron’s Global Operations Executive Vice President Manish Bhatia revealed at a JPMorgan investor conference that the company’s ramp-up progress for sixth-generation high-bandwidth memory (HBM4) is approximately twice as fast as the ramp cycle of 12-layer HBM3E in 2025. He also noted that yield improvements are accelerating alongside production scaling. In addition, Micron plans to begin mass production of its next-generation HBM4E products in 2027.
In terms of capacity expansion, Micron Technology is increasing investments to narrow the gap with leading competitors SK hynix and Samsung Electronics. Market forecasts suggest that by mid-2026, bit shipments from its 9th-generation NAND Flash and 1γ (1-gamma) process DRAM will account for more than half of total output. Notably, 1γ DRAM is expected to become the company’s largest single DRAM process node in terms of wafer capacity.
The faster-than-expected HBM4 ramp is driven by three key factors. First, manufacturing experience accumulated from earlier HBM3 and HBM3E production has significantly improved learning efficiency. Second, the core die of the product is based on Micron’s mainstream 1β (1-beta) process, which has already demonstrated stable yield performance and validated high reliability. Third, the integration of 1β DRAM with internally optimized logic base die design has further enhanced overall performance and product maturity.
For the upcoming HBM4E generation, Micron is adjusting its manufacturing strategy. The core die will transition to the 1γ process node, which is aligned with the 1c generation used by SK hynix and Samsung Electronics, and marks Micron’s first adoption of ASML’s extreme ultraviolet (EUV) lithography technology. Meanwhile, the base die will no longer be manufactured in-house and will instead be outsourced to TSMC.
Bhatia also confirmed that development of HBM4E is progressing smoothly, with mass production scheduled for 2027. The initial ramp-up phase will focus on JEDEC-standard products, while customized variants tailored to specific customer requirements are also under development. Although customized HBM4E solutions are expected to carry higher costs, they are designed to deliver enhanced performance and expanded functionality to meet differentiated application needs across advanced computing and AI workloads.