In a strategic move to advance its artificial intelligence capabilities, OpenAI has partnered with Taiwan Semiconductor Manufacturing Company (TSMC) and Broadcom to develop an in-house AI chip, according to sources reported by Reuters. While initially considering the establishment of dedicated chip manufacturing facilities, OpenAI has reportedly set aside those plans, choosing instead to concentrate on designing a proprietary AI chip internally. This new chip, tailored for running advanced AI models, could be ready by 2026.
For months, OpenAI has collaborated closely with Broadcom on this chip design project, marking a shift toward greater independence in the AI hardware sphere. The collaboration with TSMC, one of the world’s leading semiconductor foundries, could be instrumental in the eventual production and scalability of OpenAI’s chip once the design phase is complete.
In parallel with its hardware design initiative, OpenAI has diversified its chip usage by incorporating AMD processors for AI model training through Microsoft’s Azure cloud infrastructure. Previously, OpenAI primarily relied on Nvidia’s GPUs to power its extensive model training operations. However, the rising costs and limited availability of Nvidia GPUs have prompted the company to seek alternative solutions to meet its increasing computational demands.
This decision to leverage AMD’s technology is part of a broader trend within the AI industry, as chip shortages and long lead times for Nvidia GPUs have pressured tech companies to diversify their hardware portfolios. This shift may also contribute to driving down training costs and enhancing OpenAI’s flexibility as it pursues ambitious AI research and product development goals.
These strategic partnerships underscore OpenAI’s commitment to building scalable, efficient AI systems while adapting to the rapidly evolving semiconductor landscape.
Source: TechCrunch
