This event has quite a significant impact on the stocks in the memory chip sector! 🧐

NVIDIA's next-generation AI server cluster Rubin NVL72 has made major adjustments in memory configuration. To cope with supply chain constraints and ensure timely delivery of the Rubin cabinet, the DRAM capacity of the next-generation AI cabinet has been cut in half!

The capacity of a single cabinet has dropped sharply from the original plan of 55TB to 28TB, a reduction of approximately 50%, using a slimmed-down version of the 96GB SOCAMM memory module, replacing the previous high-end 192GB module.

The cost of a single cabinet has decreased from 7.6 million dollars to 6.8 million dollars, saving 800,000 dollars, which accounts for about 11%.

Theoretically, NVIDIA dares to make such cuts because they should have a backup plan. It is very likely that they may compensate for the corresponding performance loss through other technologies, such as stronger NVLink interconnects, improvements with HBM4, or software optimization!

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。