OpenAI, the parent company of ChatGPT, is creating its own custom AI chip in collaboration with Broadcom and TSMC, while also continuing to utilize Nvidia and AMD chips to meet its infrastructural needs, reports Reuters.
The report states that OpenAI has been exploring alternative chip supply sources to reduce expenses and is now considering full-scale in-house production. This initiative involves a costly plan to establish a manufacturing network for chip production, and OpenAI is evaluating funding options for its execution.
According to unnamed sources, this ambitious foundry project requires substantial financial and time investment, leading OpenAI to set it aside for now and focus on its own chip designs. To ensure a steady supply of chips and manage costs, OpenAI is adopting a hybrid approach involving partnerships with established chip industry companies and the development of proprietary chips.
Sources also indicate that OpenAI has been collaborating with Broadcom for several months to design AI chips tailored for inference tasks—central to AI decision-making processes.
OpenAI has assembled a 20-member team for chip development, led by top engineers including Thomas Norrie and Richard Ho, who previously contributed to Google’s Tensor Processing Unit (TPU) technology.
