Discussions come as AI firms increasingly look beyond Nvidia for specialized hardware
AI company Anthropic is looking into developing a proprietary silicon chip. The company has reportedly been in discussions with Samsung regarding collaboration opportunities. The company has also been contemplating developing its own chips to alleviate ongoing issues regarding the availability and supply chain constraints of AI technology. Details on the proposed silicon design remain minimal at this time.
Details Still Unclear
Per the report, Anthropic has not yet settled on the chip’s intended purpose, how it would be integrated into servers, or what performance targets it might aim for. The company declined to comment on any potential partnership with Samsung, and reiterated that its infrastructure strategy will continue to rely on chips supplied by Google, Amazon and Nvidia, adding that it has nothing further to share regarding the Samsung talks.
Part of a Broader Industry Shift
If confirmed, Anthropic would be among a growing number of artificial intelligence companies who are looking for custom chips in order to optimize their workloads and reduce dependence on Nvidia, which continues to lead the market in AI accelerators. Last week, OpenAI developed its first custom inference chip called Jalapeno, in collaboration with Broadcom, claiming superior performance-per-watt compared with its competitors’ products. Amazon continues to use its Trainium for AI training, while Google continues to expand use of its Tensor Processing Units via its cloud computing services.
Samsung’s Growing AI Role
Samsung already holds a strong position in the AI supply chain through its semiconductor manufacturing capabilities and pre-existing partnerships with Nvidia. In addition, the technology giant from South Korea is also partnering with many other companies in the AI space, including Google, as businesses continue to increase demand for specialized AI hardware.
