SK hynix heralds the next generation of AI GPUs with HBM4 mass production from 2026.

Micron and Samsung have announced their next generations of HBM4 memory products for 2025-2026. SK Hynix is also on board. Due to the increasing use of AI, improved computing power is required, with HBM playing an important role in the production of AI accelerators.

During a presentation at SEMICON Korea 2024, SK Hynix Vice President Kim Chun-hwan announced that the company will start mass production of HBM4 by 2026. He emphasized that this would be a significant boost for the AI markets. Kim emphasized the importance of reaching the next generation while managing the high demand in the HBM industry. Therefore, it is crucial to develop a solution that ensures a continuous supply and is innovative at the same time. According to Kim, the HBM market is expected to grow by up to 40% by 2025 and SK Hynix has positioned itself early to benefit the most from this.

The expectations for HBM4 are illustrated by a timeline from Trendforce, which states that the first HBM4 samples are expected to offer up to 36GB of capacity per stack. Full specifications are expected to be published by JEDEC in the H2 2024-2025 period. First customer samples and availability are expected in 2026. So it will be some time before we see the new high-speed memory AI solutions in action. It is unclear what kind of AI products will use the new process, so it is difficult to make predictions at the moment.

With SK Hynix entering the market, it seems as though the HBM markets will see significant competition in the future. It will be exciting to see which company will ultimately take the lead.

Source: Trendforce