SK hynix Takes the Lead in HBM4 Supply to NVIDIA for Rubin AI GPUs
SK hynix has emerged as the leader in the HBM (High Bandwidth Memory) market once again, becoming the first company to supply NVIDIA with HBM4 modules for testing the Rubin AI GPU architecture. 
This move is a significant step forward in the evolution of computational technology, as HBM4 integrates both memory and logic components into a single package, marking a groundbreaking development in the industry.
According to a recent report by the Korean media outlet DealSite, SK hynix has already begun supplying HBM4 in small quantities to NVIDIA, paving the way for the official adoption of this technology. NVIDIA plans to launch the Rubin GPU by Q4 2025, with the customer qualification process set to begin in September. This is a critical milestone for NVIDIA, which is keen on sticking to its six-month product release cycle. The move also secures SK hynix’s position as NVIDIA’s primary supplier for HBM, with large orders expected to follow.
As for the competition, Micron is trailing behind SK hynix in the HBM4 race. Micron’s yield rates for HBM4 are reportedly lower, and the company is facing challenges in ramping up production. Samsung, while making strides with HBM4, has not yet revealed its yield rates. However, with its recent success in supplying HBM3E to AMD, Samsung’s chances of becoming an HBM4 supplier for NVIDIA remain in the mix, though nothing is confirmed.
Despite these challenges, NVIDIA is committed to a multi-sourcing strategy, meaning they will likely source HBM4 from both SK hynix and possibly Micron or Samsung. This will ensure a steady supply of HBM4 for the Rubin GPU and secure NVIDIA’s position as a leader in AI and high-performance computing markets. With SK hynix leading the charge, the future of HBM4 looks promising, and its adoption by NVIDIA is a crucial step for the industry as a whole.