OpenAI is advancing its hardware development by partnering with Broadcom and TSMC. it will help to create its first custom AI chip. This initiative aims to lessen its dependence on external suppliers like Nvidia. The project marks a shift from OpenAI’s earlier ambitions to establish a network of chip foundries.
Now, it focuses on strategic partnerships and in-house chip design. It will help to manage costs and ensure a steady supply of chips amid increasing demand for AI infrastructure. OpenAI is incorporating AMD chips alongside Nvidia chips into Microsoft’s Azure cloud for its AI systems.
This move reflects OpenAI’s response to escalating compute costs and global chip shortages. It can affect its core operations, including AI model training and inference. Reports indicate that the company has considered various strategies to lessen its reliance on Nvidia’s GPUs. Currently, Nvidia’s GPUs hold over 80% of the market share.
The choice to pursue internal chip design underscores OpenAI’s goal to optimize its technology stack. It remains competitive in a rapidly growing market. Broadcom is assisting OpenAI in refining the architecture for these new AI chips. It could eventually support a variety of inference-focused applications that generate real-time predictions based on fresh data.
Broadcom is expected to help OpenAI realize its vision by incorporating fast data-movement components crucial for large-scale AI operations. OpenAI has assembled a team of around 20 engineers. It includes former Google experts like Thomas Norrie and Richard Ho. They are focused on designing chips that could establish OpenAI as a significant player in the AI hardware industry.
