\n<\/p>\n
On Wednesday, OpenAI unveiled<\/a> its first custom-built inference processor, designed and manufactured in collaboration with Broadcom. Named Jalape\u00f1o, the new processor was designed specifically for the unique needs of OpenAI\u2019s inference systems. OpenAI\u2019s own AI models assisted in the development of the chip, the company said.<\/p>\n While the chip is still being tested, OpenAI says early results show significantly better performance-per-watt than current state-of-the-art alternatives.<\/p>\n The partnership was officially announced in October<\/a>, but OpenAI\u2019s chip plans have long been rumored<\/a> as a way to reduce the company\u2019s dependence on Nvidia\u2019s GPUs. Google<\/a> and Amazon<\/a> have both built custom chips to serve a similar purpose, often called \u201cAI accelerators\u201d \u2014 silicon designed specifically to speed up machine learning workloads.<\/p>\n OpenAI president Greg Brockman explained the company\u2019s approach to chip development on its in-house podcast<\/a>, shortly after the Broadcom partnership was announced.<\/p>\n \u201cWe have a deep understanding of the workload,\u201d Brockman said in the episode. \u201cWe\u2019ve really been looking for specific workloads that are underserved, [and asking] how can we build something that will be able to accelerate what\u2019s possible?\u201d<\/p>\n \n