Workflow
MONAKA CPU
icon
Search documents
富士通与英伟达联合开发AI半导体
日经中文网· 2025-10-04 08:51
Core Insights - Fujitsu and NVIDIA are collaborating to develop a semiconductor aimed at artificial intelligence (AI) applications, with a goal to connect their chips on the same substrate by 2030, enhancing computational efficiency and energy savings [1][2] - The partnership aims to tap into new markets such as data centers and robotics, leveraging NVIDIA's expertise in GPU technology and Fujitsu's CPU development [1][2] Group 1: Collaboration Details - The joint semiconductor development will utilize NVIDIA's technology to achieve ultra-fast interconnectivity between multiple chips, allowing them to function as a single chip [1] - Fujitsu's new CPU, named "MONAKA," is being developed based on Arm architecture, with a target of achieving double the power efficiency compared to competitors, and is expected to be operational by 2027 [2] - The collaboration is expected to enhance energy efficiency significantly, with Fujitsu's president stating it marks an important step towards an AI-driven society [2] Group 2: Market Expansion and Strategic Goals - NVIDIA is looking to expand its market presence in Japan by leveraging Fujitsu's extensive experience in system construction and its established customer base across various sectors, including government and finance [5] - The partnership also includes plans for future collaboration on the successor to the "Fugaku" supercomputer, aiming for higher computational capabilities by integrating CPU and GPU technologies [5] - Other Japanese companies, such as Hitachi and NTT, are also engaging in partnerships with international firms to enhance energy efficiency in AI data centers, indicating a broader trend of collaboration in the tech industry [4]
日本芯片大厂:Rapidus很好,我选台积电
半导体行业观察· 2025-06-24 01:24
公众号记得加星标⭐️,第一时间看推送不会错过。 Rapidus CEO在一篇文章中回顾道,公司试图生产的芯片采用了一种前所未有的结构,称为GAA (Gate All Around,环绕栅极)结构,这有助于实现2纳米半导体所需的超精细精度。IBM于2021 年5月发布了其基于GAA的2纳米工艺节点的基础研究成果,该成果是IBM多年潜心研究的成果。然 而,该技术尚未实现量产。 Rapidus 以此为契机,派遣了约150名工程师,主要集中在拥有多年半导体经验的人员,前往IBM位 于纽约的研发基地,开发实用技术。许多负责尖端技术的人员回到日本,开始进行试生产。 据介绍,在这个过程中,Rapidus面临很多技术挑战和问题,比如如何开发能够实现复杂3D晶体管结 构的技术以及如何结合不同的材料和成分。 来源:内容 编译自 nikkei 。 富士通社长时田贵人23日在横滨举行的定期股东大会上,就致力于尖端半导体国产化的Rapidus发表 了看法,他表示:"增加尖端半导体的供应源,对于确保我们供应链的稳定性极为有益。"富士通将继 续采购用于人工智能(AI)相关技术的尖端半导体。 富士通计划投资Rapidus。一位股东询问富士通 ...