Samsung Is Already Developing HBM5 With 2nm Technology

Samsung is already working on the next generation of high-bandwidth memory — HBM5 and HBM5E — and this could become a key part of the future AI hardware race.

According to reports from South Korean media, Samsung plans to use its own 2nm process technology for the base die in HBM5 memory stacks. The DRAM chips inside HBM5 are expected to use Samsung’s sixth-generation 10nm-class process, known as 1c. For HBM5E, Samsung may move the DRAM dies to an even newer 1d process, while keeping the 2nm base die.

This is important because HBM memory is one of the most critical components for modern AI accelerators. Nvidia, AMD, Google, Microsoft and other companies need massive memory bandwidth to train and run large AI models. As a result, the competition between Samsung, SK hynix and Micron is becoming more intense.

Samsung’s relationship with Nvidia also appears to be getting stronger. Jensen Huang recently praised Samsung’s HBM4 memory, and Samsung has reportedly secured important AI-related manufacturing work, including chips connected with Groq technology.

Samsung is also using a step-by-step upgrade strategy. HBM4 uses a 4nm base die, while HBM5 moves that base die to 2nm. This may help reduce technology risks while still pushing performance forward.

In this Xenon Lab News episode, we look at Samsung’s HBM5 roadmap, why 2nm base dies matter, how HBM memory affects AI performance, and why Nvidia’s trust could become a major win for Samsung in the next wave of AI computing.

Subscribe to Xenon Lab for more news about AI chips, memory, semiconductors, processors and future technologies.

#Samsung #HBM5 #HBM4 #HBM5E #Nvidia #AIHardware #AIMemory #Semiconductors #2nm #DRAM #SKHynix #Micron #Groq #TechNews #XenonLab

Видео Samsung Is Already Developing HBM5 With 2nm Technology канала Xenon Lab | Science & Technology