OpenAI's custom AI chips set for 2026 production: report

OpenAI, the company behind ChatGPT, is moving forward with plans to develop its own AI chip, aiming for production by 2026, according to a recent report by Reuters. Working with Broadcom and TSMC, and adding AMD chips to its current lineup alongside Nvidia's GPUs, OpenAI is diversifying its chip suppliers to meet the increasing demand for its AI technology and reduce costs, states the report.

The company recently decided to abandon the idea of building its own chip manufacturing plants —a costly, complex project—and is instead focusing on designing a custom chip with support from Broadcom and TSMC. Broadcom, a leader in chip design, will help develop OpenAI's chip, which will be tailored for inference (real-time decision-making). TSMC is set to handle the manufacturing, says the Reuters report.

OpenAI has brought in experienced engineers, including former members of Google's Tensor Processing Unit (TPU) team, to lead its in-house chip design, states Reuters. At the same time, OpenAI is leveraging its partnership with Microsoft Azure to bring in AMD's new MI300X chips, aiming to control soaring operational costs that could lead to a $5 billion loss this year, further adds the report.

As a major buyer of GPUs, OpenAI remains cautious about disrupting its relationship with Nvidia, given the importance of accessing Nvidia's newest Blackwell chips, suggests Reuters. Nvidia, AMD, and OpenAI declined to comment to Reuters on the developments.