内容来源：https://www.wired.com/story/nvidias-rubin-chips-are-going-into-production/

内容总结：

在拉斯维加斯CES科技展的新闻发布会上，英伟达首席执行官黄仁勋宣布，新一代人工智能超级芯片平台Vera Rubin已进入全面生产阶段，预计将于今年下半年开始向客户交付。该平台以揭示星系奥秘的天文学家薇拉·鲁宾命名，其核心芯片采用台积电3纳米制程与先进存储技术，通过第六代互联架构整合六类芯片，形成完整系统。

据英伟达透露，Rubin平台运行AI模型的成本可降至当前旗舰Blackwell系统的约十分之一，训练部分大模型所需芯片数量仅需Blackwell的四分之一。这一突破有望大幅降低高端AI系统的运营成本，进一步巩固客户对英伟达硬件的依赖。

微软与云服务商CoreWeave将成为首批部署Rubin芯片的企业。其中微软在佐治亚州和威斯康星州新建的两座数据中心将配备数千枚Rubin芯片。英伟达同时宣布与开源企业软件商红帽合作，为金融、汽车、航空及政府机构开发基于Rubin平台的产品。

尽管英伟达宣称Rubin已"全面投产"，行业分析指出，此类尖端芯片通常需经历小批量试产与测试验证阶段，大规模量产仍按原计划于2026年下半年推进。分析师认为，此次声明意在回应市场对研发进度的疑虑，表明其已突破关键开发节点。

面对AI行业爆发式增长，各大企业正激烈争夺英伟达先进芯片资源。Rubin预计将延续供需紧张态势，但部分公司已启动自主芯片研发以降低依赖，例如OpenAI正与博通合作定制下一代AI模型芯片。不过，分析师强调英伟达通过整合计算、网络、存储及软件协调的全栈系统架构，正构建起难以替代的生态壁垒。

此前，英伟达Blackwell芯片曾因设计缺陷导致交付延迟，该问题已于2025年中解决。随着Rubin平台推进，英伟达在加速计算领域的领先优势或将进一步扩大。

中文翻译：

英伟达首席执行官黄仁勋表示，公司下一代人工智能超级芯片平台Vera Rubin已按计划将于今年晚些开始向客户交付。黄仁勋在拉斯维加斯年度国际消费电子展（CES）的新闻发布会上表示："今天我可以告诉大家，Vera Rubin已进入全面生产阶段。"

英伟达在周日与分析人士和记者的电话会议中透露，Rubin平台运行人工智能模型的成本将降至该公司当前旗舰芯片系统Blackwell的十分之一左右。该公司还表示，Rubin训练某些大型模型所需的芯片数量仅约为Blackwell的四分之一。这些性能提升将大幅降低先进人工智能系统的运营成本，使得英伟达的客户更难有理由转向其他硬件平台。

英伟达在电话会议中称，现有合作伙伴微软和CoreWeave将成为今年晚些首批提供基于Rubin芯片服务的公司。该公司补充说，微软目前在佐治亚州和威斯康星州建设的两座大型人工智能数据中心最终将部署数千块Rubin芯片。英伟达表示，部分合作伙伴已开始在早期Rubin系统上运行其下一代人工智能模型。

这家半导体巨头还宣布，正与为银行、汽车制造商、航空公司和政府机构提供开源企业软件的红帽公司合作，共同开发更多基于新型Rubin芯片系统的产品。

英伟达最新芯片平台以美国天文学家薇拉·鲁宾命名，她的研究重塑了科学家对星系特性的理解。该系统包含六种不同芯片，包括采用台积电3纳米制程工艺和最先进带宽内存技术制造的Rubin GPU和Vera CPU。各芯片通过英伟达第六代互连与交换技术实现协同工作。

黄仁勋在公司CES新闻发布会上宣称，该芯片系统的每个组件都"具有彻底的革命性，是同类产品中的佼佼者"。

英伟达研发Rubin系统已有数年之久，黄仁勋最早在2024年的主题演讲中预告了该芯片的推出。去年公司曾表示，基于Rubin的系统将于2026年下半年开始交付。

目前尚不清楚英伟达所称"全面生产"的确切含义。通常此类先进芯片（英伟达与其长期合作伙伴台积电共同制造）的生产初期规模较小，需经历测试验证阶段后才会逐步提升产量。

创意策略公司分析师、《芯片战略》半导体行业通讯作者奥斯汀·莱昂斯指出："此次CES关于Rubin的公告意在向投资者传递'我们正按计划推进'的信号。"他表示华尔街曾有传闻称Rubin GPU研发进度滞后，英伟达此次声明旨在澄清其已突破关键研发与测试环节，并确信能在2026年下半年如期启动量产爬坡。

2024年，英伟达曾因设计缺陷导致Blackwell芯片在服务器机架互联时过热，被迫延迟交付当时的最新产品。至2025年中期，Blackwell的出货才重回正轨。

随着人工智能行业迅猛发展，软件公司与云服务提供商不得不激烈竞逐英伟达最新GPU的采购权。市场对Rubin的需求预计将同样旺盛。但部分企业也通过投资自研定制芯片来分散风险，例如OpenAI已宣布与博通合作，为其下一代人工智能模型开发定制芯片。这类合作凸显了英伟达面临的长期风险：自研芯片的客户能获得该公司无法提供的硬件控制权。

不过莱昂斯认为，今日的公告表明英伟达正从单纯的GPU供应商转型为"涵盖计算、网络、内存层级、存储和软件编排的完整人工智能系统架构师"。他补充道，即便超大规模企业持续投入定制芯片研发，英伟达高度集成的平台"正变得越来越难以替代"。

英文来源：

Nvidia CEO Jensen Huang says that the company’s next-generation AI superchip platform, Vera Rubin, is on schedule to begin arriving to customers later this year. “Today, I can tell you that Vera Rubin is in full production,” Huang said during a press event on Monday at the annual CES technology trade show in Las Vegas.
Rubin will cut the cost of running AI models to about one-tenth of Nvidia’s current leading chip system, Blackwell, the company told analysts and journalists during a call on Sunday. Nvidia also said Rubin can train certain large models using roughly one-fourth as many chips as Blackwell requires. Taken together, those gains could make advanced AI systems significantly cheaper to operate and make it harder for Nvidia’s customers to justify moving away from its hardware.
Nvidia said on the call that two of its existing partners, Microsoft and CoreWeave, will be among the first companies to begin offering services powered by Rubin chips later this year. Two major AI data centers that Microsoft is currently building in Georgia and Wisconsin will eventually include thousands of Rubin chips, Nvidia added. Some of Nvidia’s partners have started running their next-generation AI models on early Rubin systems, the company said.
The semiconductor giant also said it’s working with Red Hat, which makes open source enterprise software for banks, automakers, airlines, and government agencies, to offer more products that will run on the new Rubin chip system.
Nvidia’s latest chip platform is named after Vera Rubin, an American astronomer who reshaped how scientists understand the properties of galaxies. The system includes six different chips, including the Rubin GPU and a Vera CPU, both of which are built using Taiwan Semiconductor Manufacturing Company’s 3-nanometer fabrication process and the most advanced bandwidth memory technology available. Nvidia’s sixth-generation interconnect and switching technologies link the various chips together.
Each part of this chip system is “completely revolutionary and the best of its kind,” Huang proclaimed during the company’s CES press conference.
Nvidia has been developing the Rubin system for years, and Huang first announced the chips were coming during a keynote speech in 2024. Last year, the company said that systems built on Rubin would begin arriving in the second half of 2026.
It’s unclear exactly what Nvidia means by saying that Vera Rubin is in “full production.” Typically, production for chips this advanced—which Nvidia is building with its longtime partner TSMC—starts at low volume while the chips go through testing and validation and ramps up at a later stage.
“This CES announcement around Rubin is to tell investors, ‘We’re on track,’” says Austin Lyons, an analyst at Creative Strategists and author of the semiconductor industry newsletter Chipstrat. There were rumors on Wall Street that the Rubin GPU was running behind schedule, Lyons says, so Nvidia is now pushing back by saying it has cleared key development and testing steps, and it’s confident Rubin is still on course to begin scaling up production in the second half of 2026.
In 2024, Nvidia had to delay delivery of its then-new Blackwell chips due to a design flaw that caused them to overheat when they were connected together in server racks. Shipments for Blackwell were back on schedule by the middle of 2025.
As the AI industry rapidly expands, software companies and cloud service providers have had to fiercely compete for access to Nvidia’s newest GPUs. Demand will likely be just as high for Rubin. But some firms are also hedging their bets by investing in their own custom chip designs. OpenAI, for example, has said it is working with Broadcom to build bespoke silicon for its next generation of AI models. These partnerships highlight a longer-term risk for Nvidia: Customers that design their own chips can gain a level of control over their hardware that the company doesn’t offer.
But Lyons says today’s announcements demonstrate how Nvidia is evolving beyond merely offering GPUs to becoming a “full AI system architect, spanning compute, networking, memory hierarchy, storage, and software orchestration.” Even as hyperscalers pour money into custom silicon, he adds, Nvidia’s tightly integrated platform “is getting harder to displace.”