Amid Memory Shortage, Nvidia Weighs Cutting HBM4E Capacity for Rubin Ultra GPU

Nvidia is reportedly considering lowering the HBM4E capacity to be mounted on its next-generation GPU Rubin Ultra below the originally planned 1TB. According to reports from The Information, the samples currently under testing amount to only 192-256GB, falling short of even the maximum HBM4 capacity (288GB) of Rubin, which is currently in mass production. This means the memory supply shortage is severe enough to prompt even the largest customer to change its design, suggesting stronger bargaining power for memory manufacturers and the possibility of continued price strength. Nvidia CEO Jensen Huang first unveiled Rubin Ultra at a developer conference last year, stating that each GPU would be equipped with 1TB of HBM4E consisting of 16 memory stacks. However, the samples currently being tested have fewer stack layers and less memory capacity per die. Some samples use HBM4, the previous-generation version. Nvidia is considering lowering the memory specifications because memory suppliers may struggle to produce enough HBM4E in time for Rubin Ultra scheduled release. Lowering the memory specifications could affect performance but may lower the price. Nvidia signed a 500 billion dollar infrastructure cooperation partnership with SK Group late last month to secure HBM supply, leading to interpretations that the memory supply shortage is shaking the negotiating structure across the entire semiconductor value chain.