New York\u00a0\u2014\u00a0\n <\/p>\n
\n OpenAI on Wednesday announced its first custom AI chip<\/a> in a step towards expanding beyond consumer products to become a player in AI infrastructure.\n <\/p>\n \n The processor, called Jalape\u00f1o, was developed with Broadcom to more efficiently handle the computing needs for ChatGPT and OpenAI\u2019s coding agent, Codex, the company said in a blog post. It\u2019s also designed to handle workloads for future AI developments, OpenAI claims.\n <\/p>\n \n \u201cWhile OpenAI is still measuring final performance, early testing shows that Jalape\u00f1o will deliver performance per watt substantially better than current state-of-the-art,\u201d the company said.\n <\/p>\n \n OpenAI and Broadcom announced plans<\/a> to develop custom chips to power 10 gigawatts worth of computing last year, but only unveiled the first chip from that partnership on Wednesday.\n <\/p>\n \n Tech giants like Google and Amazon have increasingly shifted toward developing their own AI chips. Custom chips allow companies to better tailor processing performance for their own models and reduce reliance on existing chip companies like Nvidia. OpenAI said Jalape\u00f1o is designed specifically for modern large language models instead of being a \u201cgeneral-purpose\u201d chip.\n <\/p>\n \n \u201cNobody wants to be beholden to Nvidia,\u201d Ben Barringer, global head of technology research at investment firm Quilter Cheviot, told CNN last November<\/a> when asked about Google potentially challenging Nvidia in AI chips<\/a>. \u201cThey are trying to diversify their chip footprint.\u201d\n <\/p>\n \n The announcement comes as OpenAI is gearing up for an initial public offering<\/a> that could potentially value it at a trillion dollars \u2013 putting more pressure on the company to prove it can generate revenue to match that valuation.\n <\/p>\n \n And there\u2019s big money to be made in AI infrastructure; chipmaker Nvidia became the world\u2019s most valuable company<\/a> largely because its chips and systems have become critical components to the data centers that run AI models.\n <\/p>\n