Nvidia CEO Jen-Hsun Huang announced in Seoul on Friday that Samsung Electronics, SK Hynix and Micron Technology have all passed certification and will supply HBM4 high-bandwidth memory chips for Nvidia's next-generation artificial intelligence platform Vera Rubin.
Huang told the media when he arrived in South Korea for a visit that all three suppliers have passed qualification certification and have entered the mass production stage. They are currently making every effort to ensure the supply needs of the Vera Rubin platform. This is the first time that NVIDIA has officially confirmed that three memory chip manufacturers have obtained HBM4 supply qualifications at the same time. These three companies dominate the global memory chip market and have previously been competing for Nvidia's HBM supply share. SK Hynix occupied about 62% of the market share in the HBM3E era, and the fact that three companies have been certified at the same time means that the supply pattern of the HBM4 generation may change.
Jen-Hsun Huang revealed during the Taipei Computex this week that Vera Rubin has entered full mass production and the complete product will be officially shipped in the third quarter of this year. This AI system is built by NVIDIA Vera CPU and Rubin GPU cluster. A single server will be equipped with TB-level HBM4 high-bandwidth memory. HBM4 is the sixth generation of high-bandwidth storage products. Compared with the previous generation HBM3E, the interface width has doubled, and the data transmission speed has increased from approximately 1 TB/s to 2 TB/s.
Judging from the progress of each company, Samsung Electronics took the lead in launching HBM4 mass production and shipment in February this year; SK Hynix HBM officially entered full mass production in early 4; Micron announced the mass production of HBM4 in March, and its production capacity ramp-up rate is twice as fast as last year's mass production of HBM3E.
Arm CEO Rene Haas pointed out this week that the shortage of high-end storage is the most difficult capacity bottleneck to overcome in the current AI industry chain. This time, the three major manufacturers have obtained certification simultaneously, which is of positive significance to alleviating supply constraints.
Huang Renxun also revealed that Nvidia is setting up a new R&D center in South Korea and launching personnel recruitment to strengthen supply chain coordination with Korean partners.