Odaily reports that according to market sources this evening, the latest research report from SemiAnalysis stated that "Nvidia's next-generation AI server cluster, Rubin NVL72, has undergone significant adjustments in memory configuration. To address supply chain tightness and ensure the timely delivery of Rubin racks, the per-rack capacity has been slashed from the original plan of 55TB to 28TB, a reduction of approximately 50%. It will adopt a scaled-down 96GB SOCAMM memory module, replacing the previous high-end 192GB module." This news may have pressured memory concept stocks, including Micron and SK Hynix, causing them to decline.
In response, Dylan Patel (@dylan522p), Founder, CEO, and Chief Analyst of SemiAnalysis, stated, "I love how people forward what we say but miss most of the content in the note. Our original report did not use this type of clickbait subscription-driven headline."
