Photo: Camilo Concha/Shutterstock.com
OpenAI is set to diversify its supply chain by incorporating AMD silicon for its artificial intelligence (AI) systems and will work with Broadcom and TSMC to build its in-house AI chip. This move aims to reduce reliance on Nvidia’s GPU for its AI operations, bolster the supply chain, and cut infrastructure costs.
As reported by Reuters, OpenAI wants to gain power control over hardware amid growing demand for computing power. The company initially explored plans for a costly foundry network — a manufacturing chain solely dedicated to its chips.
However, high financial and time demands led OpenAI to abandon the foundry concept, focusing instead on a collaborative approach to chip development and diversification with established industry players.
Over time, OpenAI has become one of Nvidia’s biggest customers, deploying thousands of its high-performance Blackwell GPUs both for training large AI models and for inference, where the AI makes real-time predictions based on new data.
OpenAI has worked closely with chip supplier Broadcom to develop an inference-focused chip intended to improve performance and reduce dependency on Nvidia’s already strained GPU supply.
Broadcom, which has helped other tech giants like Alphabet fine-tune and streamline chip designs, has reportedly secured production capacity for OpenAI’s chip with Taiwan Semiconductor Manufacturing Company (TSMC), a globally respected semiconductor manufacturer.
OpenAI’s custom chip is expected to be ready for manufacturing by 2026, though this timeline could shift based on market and production factors. Meanwhile, the company is evaluating its options for acquiring or developing additional elements to enhance its custom chip design, with speculation that additional partnerships could be on the horizon.
Following reports of OpenAI’s collaboration with Broadcom, Broadcom’s stock surged by 4.5%, while AMD shares climbed 3.7%, signalling positive investor confidence in OpenAI’s hardware expansion.
Other companies, such as Meta, Amazon, and Google, are also venturing into custom chip development to gain an edge in the competitive AI market.
In the News: Star Health clears CISO of wrongdoings in data breach case
