OpenAI is partnering with Broadcom and TSMC to develop its first in-house chip for its artificial intelligence systems. The company is planning to integrate AMD chips alongside Nvidia chips to address its increasing infrastructure needs. OpenAI, known for ChatGPT, is exploring various options to expand chip supply and reduce expenses. Talks with Broadcom regarding this initiative have been reported.
According to sources, OpenAI has been collaborating with Broadcom for several months to create its inaugural AI chip focused on inference. This chip will play a crucial role in AI systems that require thousands of chips working together simultaneously. The company is currently deliberating on whether to create or acquire additional components for its chip design and is open to engaging more partners, as mentioned by insiders.
OpenAI has formed a team of around 20 individuals, led by experienced engineers who have previously worked on Tensor Processing Units (TPUs) at Google - such as Thomas Norrie and Richard Ho. Via Broadcom, OpenAI has secured manufacturing capacity with Taiwan Semiconductor Manufacturing Company to produce its first custom-designed chip by 2026.
The decision by OpenAI, a significant chip consumer, to diversify its chip suppliers while developing a personalized chip could have considerable implications for the tech industry. Major customers like Microsoft, Meta, and OpenAI have been exploring in-house or external alternatives due to chip shortages and escalating costs. OpenAI's integration of AMD chips through Microsoft's Azure illustrates AMD's attempt to compete with Nvidia in the chip market.
Although OpenAI initially considered constructing everything internally, including developing a network of chip manufacturing foundries, the company has, for now, abandoned these elaborate plans due to the associated costs and time constraints. Instead, it is focusing on in-house chip design efforts to optimize costs and ensure a stable chip supply.
OpenAI's strategy, which involves a mix of industry partnerships and internal and external approaches, aims to manage costs efficiently, similar to larger tech players like Amazon, Meta, Google, and Microsoft. Compute costs constitute a significant portion of OpenAI's expenses, prompting the company to enhance utilization and diversify its suppliers.
The company is cautious about poaching talent from Nvidia to maintain a positive relationship with the chipmaker, particularly for accessing Nvidia's new Blackwell chips. OpenAI heavily relies on substantial computing power to operate its systems effectively. AMD expects substantial sales from its AI chip following its launch, highlighting the growing demand for AI chips in the market.
OpenAI has forecasted a significant loss this year despite substantial revenue. Its move to incorporate AMD chips, alongside Nvidia chips, reflects the evolving dynamics in the AI chip market. Both AMD and Nvidia's market shares are seeing positive responses to these developments. OpenAI, AMD, and TSMC have chosen not to offer comments on these developments at this time.
