NVIDIA and Samsung reportedly in development talks for co-designed AI memory chips.

NVIDIA has captured the market with its AI accelerators, which require significant high-bandwidth memory capacity. Currently, SK Hynix supplies the bulk of this memory, while Samsung continues to pursue approval for its HBM3E chips. As Samsung's authorization process continues, both Samsung and SK Hynix are reportedly engaged in secret discussions to commercialize a new type of advanced AI memory chip known as SOCAMM, or system on chip advanced memory module. NVIDIA is also said to be in talks with SK Hynix regarding this memory chip, with prototypes currently being exchanged with NVIDIA.

The talks center around commercializing this new memory standard, which the industry views as an advancement over the existing high-bandwidth memory chips that form the backbone of AI accelerators today. These chips are designed to significantly improve the performance of artificial intelligence supercomputers. SOCAMM offers a higher cost-performance ratio compared to current solutions, and its module design features a higher number of ports, which helps reduce data bottlenecks—a remaining challenge in AI computing.

The detachable design of SOCAMM will enable data center operators to swap and upgrade memory modules and achieve ongoing performance improvements. Given its compact size, more SOCAMM modules can be installed within the same area for improved high-performance computing. Industry insiders highlight that NVIDIA and memory companies are viewing this development as a significant step forward in addressing the infrastructure demands of modern artificial intelligence systems.