Anthropic has announced the formation of a new department dedicated to the development of its own artificial intelligence accelerators. The company aims to create both models and hardware concurrently, enhancing the efficiency and speed of its systems. This initiative comes amid a growing demand for its AI models, particularly the Claude series.
To support its chip production, Anthropic is reportedly considering Samsung as a potential manufacturer. In its quest for expansion, the company has recently secured partnerships with major players such as AWS, Google, Nvidia, and AMD for access to computing resources. However, these collaborations may not suffice to fully meet the increasing demand for Anthropic's services.
Anthropic is not the first AI lab to embark on the journey of building proprietary accelerators. OpenAI recently unveiled its Jalapeño accelerator, developed with Broadcom, which is tailored for AI inference workloads. Similarly, Google DeepMind has long utilized its TPU accelerators, and Meta is preparing to roll out its own MTIA chips.
Currently, Anthropic is on the lookout for engineers skilled in chip design to join its "custom silicon design department." The company has yet to release official comments regarding this initiative.
This strategic move underscores the intensifying competition in the AI infrastructure market, as companies strive to enhance their technological capabilities and meet the growing demands of AI applications.
Informational material. 18+.
