OpenAI to launch first in-house AI chip with Broadcom in 2026: Report

OpenAI is set to introduce its inaugural artificial intelligence chip in collaboration with the major semiconductor firm Broadcom, as reported by the Financial Times. This new silicon is slated for completion next year and will be utilized exclusively for internal purposes, rather than being made available to external clients. This strategic initiative underscores OpenAI's aim to lessen its reliance on Nvidia, whose GPUs currently dominate the market for training and deploying extensive language models. According to previous reports from Reuters, OpenAI has been working not only with Broadcom but also with Taiwan Semiconductor Manufacturing Co. (TSMC) for the chip's development, while also employing processors from AMD in addition to Nvidia to meet the surging demand for computational power. A source close to the project revealed that the design for OpenAI's first-generation chip is nearing finalization and is prepared to be sent to TSMC for production. Broadcom's CEO Hock Tan commented recently that the company anticipates a significant boost in AI-related revenue by fiscal 2026, following the acquisition of more than $10 billion in infrastructure contracts from a new client, which is widely speculated to be OpenAI. During an earnings call, Tan noted, "A new prospect placed a firm order last quarter, qualifying them as a customer," though he did not disclose the identity of the client. Earlier this year, he suggested that Broadcom is collaborating closely with four potential additional clients for custom chip solutions, alongside its existing major accounts. OpenAI's initiative mirrors similar moves by industry leaders such as Google, Amazon, and Meta, all of whom have developed their own proprietary chips to manage costs and enhance performance for AI tasks. By launching its own hardware, OpenAI aims to achieve greater adaptability in scaling future models while exercising more stringent control over expenses. This venture comes at a time when competition in the AI landscape is intensifying, necessitating substantial computational resources for training advanced models. OpenAI has previously indicated that its long-term vision involves diversifying its supply chain to mitigate dependence on a singular chip supplier.