注册并分享邀请链接,可获得视频播放与邀请奖励。

StatementDog | AI & Tech Supply Chains
@statementdog
Tracking the companies and technologies behind the AI buildout. Research informed by company interviews, site visits, analysts, and industry experts.
加入 June 2024
134 正在关注    4.7K 粉丝
CPO manufacturing is becoming more productive, with automation lifting throughput roughly five- to sixfold in optical-component assembly and testing, according to a SPIL executive cited by DIGITIMES. SPIL, part of ASE Technology $ASX, packages and tests chips and works with NVIDIA on co-packaged optics. The executive described early production lines processing fewer than 10 units per hour. Automation has since improved throughput, with a longer-term goal of reaching triple-digit hourly output. Current CPO demand remains concentrated in scale-out networks connecting servers. The executive said manufacturers can still accommodate that demand despite the remaining manual work in assembly and testing. Broader adoption in scale-up networks linking accelerators brings another challenge: whether today’s CPO architectures can meet those requirements. Designs vary across customers and product generations, making manufacturing processes difficult to replicate. For us, the next question is how much of the automation and production experience from today’s scale-out products can carry over to the designs customers choose for scale-up. $ASX $NVDA
显示更多