Broadcom logo / Yonhap News Agency

OpenAI, the developer of ChatGPT, plans to mass-produce its own artificial intelligence (AI) chips in collaboration with U.S. semiconductor corporations Broadcom starting next year, the Financial Times (FT) reported on the 4th (local time). This is interpreted as an effort to respond to the surging demand for AI computing and to reduce dependence on NVIDIA.

FT cited sources familiar with the matter as saying that the AI chip jointly designed by OpenAI and Broadcom is scheduled to be shipped next year. Hawk Tan, CEO of Broadcom, noted after the company's second-quarter earnings announcement that "a new customer has promised orders worth $10 billion (about 13.9 trillion won)," and sources indicated that the new customer is OpenAI.

OpenAI plans to use the AI chips produced with Broadcom for internal purposes, not for external sales. This is a strategy to meet the surging computing demands required for the training and operation of the latest models like GPT-5. Sam Altman, CEO of OpenAI, mentioned last month that "we will double our computing resources over the next five months."

Broadcom is developing and supplying customized AI chips (XPU) targeted at global big tech corporations such as Google. Tan stated that the contract with the new customer known as OpenAI has generated "immediate and significant demand," improving the company's growth outlook.

The current AI infrastructure market is trending toward diversification from GPUs produced by NVIDIA and AMD to XPUs designed in-house by various corporations.

FT reported that "XPUs will continue to expand market share in the AI infrastructure market," noting that Broadcom has secured its fourth major customer in the customized AI chip business through its contract with OpenAI. HSBC analysts projected that Broadcom's customized chip business would record a higher growth rate than NVIDIA's graphics processing unit (GPU) business by 2026.

Currently, NVIDIA essentially monopolizes the AI accelerator market. As dependence on a single supplier increases and AI investment expenses rise, big tech corporations such as Google, Amazon, and Meta are seeking alternatives to NVIDIA AI chips.

※ This article has been translated by AI. Share your feedback here.