As AI power and computing demands continue to rise, HBM4 has become a new battlefield for memory manufacturers. Whoever can be the first to pass NVIDIA certification and mass production will be expected to dominate the next stage of the AI ​​chip supply chain. Micron (Micron) is lagging in the development of HBM4 and may not be able to catch up with NVIDIA's supply schedule.

According to "서울경제", Micron still has difficulty meeting NVIDIA's performance and power consumption requirements for its HBM4 products, and there are rumors in the market that the company may need to redesign the chip architecture. If the manufacturing process is adjusted, the overall schedule will be delayed by up to 9 months, which means that mass production will be difficult to achieve within next year.

Industry insiders pointed out that HBM4 is a key component designed for high-power computing in AI and data centers, and its requirements for bandwidth, power consumption, and stacking integration are much higher than those of the previous generation. In comparison, SK hynix took the lead in completing the world's first HBM4 mass production line last month. Samsung Electronics also unveiled the actual HBM4 product at this month's "2025 Semiconductor Exhibition (SEDEX)" and launched mass production preparations. The technology gap is further widening.

Analysis points out that if Micron continues to lag behind, it will not only miss out on NVIDIA's recent major procurement opportunities, but may also be excluded from the Korean factory in the new round of AI server orders. The market expects that the HBM4 market will still be dominated by SK hynix and Samsung in the short term. Whether Micron can return to competition in 2026 will depend on the progress of its HBM4 design revision and yield improvement.