Anthropic正在招聘芯片设计工程师，组建专门团队开发定制AI芯片，以支持其模型运行更快更高效[1]。该公司计划采用硬件与模型协同设计的方式，并已与AWS、Google、Nvidia和AMD达成AI计算硬件合作协议[1]。
此举使Anthropic加入了业界自主芯片研发的阵营[1]。OpenAI在6月推出由Broadcom制造的Jalapeño芯片用于推理工作负载[1]，Google DeepMind则使用Alphabet旗下的TPU芯片[1]，Meta正在开发MTIA加速器[1]。此前有报道称，Anthropic曾在评估三星作为潜在芯片制造伙伴[1]。
Anthropic is building a dedicated team to design proprietary chips optimized for running its AI models with greater speed and efficiency [1]. The company plans to pursue a co-design approach with hardware manufacturers, having already established partnerships with AWS, Google, Nvidia, and AMD for AI computing hardware collaboration [1].
This move aligns Anthropic with other leading AI companies pursuing custom silicon strategies. OpenAI introduced its Jalapeño chip, manufactured by Broadcom, for inference workloads in June [1], while Google DeepMind relies on Alphabet's TPU chips and Meta is developing its MTIA accelerator [1]. According to a report from The Information last month, Anthropic had been evaluating Samsung as a potential manufacturing partner for its chips [1].
The company is actively recruiting chip design engineers to staff its new "custom silicon team" [1], signaling a significant commitment to vertical integration in AI hardware development.