OpenAI is making its own AI chips with Broadcom's help

Core Insights - OpenAI has signed a deal with Broadcom to develop custom chips and systems for its infrastructure and partners' data centers, aiming for deployment in the second half of 2026 and completion by the end of 2029 [1][2] Group 1: Partnership Details - The deal with Broadcom involves 10 gigawatts of chips and is valued at "multiple billions of dollars" [3] - OpenAI has previously established agreements with NVIDIA and AMD, with NVIDIA investing $100 billion for 10 gigawatts of AI infrastructure and AMD providing six gigawatts of compute power [4] Group 2: Future Plans and Financials - OpenAI CEO Sam Altman aims to build 250 gigawatts of compute power over the next eight years, a significant increase from the expected 2 gigawatts by the end of this year [5] - The estimated cost to achieve this capacity is around $10 trillion, with OpenAI needing to develop new financing tools [5] - OpenAI is projected to generate $13 billion in revenue this year, indicating a substantial gap in revenue compared to the required investment [6]