英伟达表示，旗下新一代 Rubin 资料中心加速器正接近在 2026 年内推出，客户很快就能试用以加速 AI 开发。公司称 6 颗全新的 Rubin 晶片已从制造伙伴端回到手上，并已通过部分里程碑测试，显示仍按计划可供客户部署。执行长黄仁勋在拉斯维加斯 CES 周一的主题演讲中强调进度，并称 AI 竞赛正在加速。
在效能数据上，英伟达宣称 Rubin 在训练方面较前一代 Blackwell 提升 3.5 倍、在执行 AI 软体（推论/运行）方面提升 5 倍。新款 CPU 具备 88 核心，效能约为替换元件的 2 倍。公司也强调 Rubin 系统将更便宜好跑：以更少的元件达到相同结果、降低运行成本。英伟达表示包含微软在内的大型云端业者，预计将在 2026 年下半年率先部署。
面对竞争与支出增速疑虑，市场有人担心对手与资料中心业者自研加速器会加剧压力、且 AI 投入无法长期维持现有节奏；但英伟达仍看好长期，称总市场规模可达「数兆美元」。目前英伟达系统的主要采购仍集中在少数客户的资本支出，例如微软、Google Cloud 与 AWS。产品层面，Rubin 硬体（含网路与互连）将纳入 DGX SuperPod，也可模组化单独采购，并搭配新工具扩大 AI 在机器人、自驾、医疗与重工业等领域的采用。
Nvidia says its new Rubin data-center accelerators are nearing release in 2026, and customers will soon be able to try them. All six Rubin chips have returned from manufacturing partners and have already passed key milestone tests, signaling they remain on schedule for customer deployment. CEO Jensen Huang highlighted this progress in a CES keynote in Las Vegas on Monday, framing it as part of an intensifying race to push AI to the next level.
Nvidia claims Rubin delivers 3.5× higher training performance and 5× higher AI software runtime performance than Blackwell. It also pairs with a new CPU featuring 88 cores and roughly double the performance of the part it replaces. Rubin systems are positioned to be cheaper to operate because they can achieve the same results with fewer components. Nvidia expects major cloud providers, including Microsoft, to be among the first deploying Rubin-based systems in the second half of 2026.
Huang’s early product disclosures—typically saved for Nvidia’s spring GTC event in San Jose—aim to keep industry attention as competition rises. Wall Street worries that rival chips, including in-house accelerators from data-center operators, could slow Nvidia’s growth and AI spending. Nvidia remains bullish, pointing to a total addressable market measured in trillions of dollars. Today, much of capex for Nvidia systems comes from a small set of buyers such as Microsoft, Google Cloud, and AWS, while Nvidia broadens into robotics, healthcare, and heavy industry with new autonomy tools.