According to CoinMeta, NVIDIA announced that its new generation of rack-level AI computing platform Rubin is now entering full mass production, with Microsoft's first batch of devices having arrived. Microsoft CEO Satya Nadella shared images from the data center site, where the Rubin system integrates 72 Rubin GPU and 36 Vera CPU. NVIDIA stated that compared to GB200 NVL72, Rubin can reduce the inference cost per million token by about one-tenth. Meanwhile, several major NVIDIA customers have received price increase notifications, with the AI servers expected to be shipped in early 2027 facing general price hikes, with some increases exceeding 15%. This round of price increases is mainly due to rising memory costs. Although Samsung, SK Hynix, and Micron are expanding production, the DRAM supply still cannot meet the demands of AI data centers.