OpenAI and Broadcom announce OpenAI’s first custom artificial intelligence chip, developed in partnership with Broadcom as part of a plan to better tailor hardware for running its AI products. The companies describe the chip as designed specifically for AI inference—the stage of processing input to produce outputs—rather than general-purpose computing. Bloomberg and the Financial Post report that the unveiling is intended to help OpenAI improve performance and reduce operating costs by using dedicated silicon for its models. Quartz adds that Broadcom’s chief executive says the chip delivers around 50% cost savings compared with typical AI GPUs, indicating a significant reduction in cost per deployment. The reports characterize the effort as a step toward more efficient execution of AI workloads through custom hardware, while emphasizing that the partnership combines OpenAI’s model and product needs with Broadcom’s experience in chip design and production. All accounts agree on the same core details: the chip is custom-built, created with Broadcom, targeted at inference, and presented as a way to make running AI models faster and cheaper.