OpenAI explores custom AI chip with this NVIDIA rival
OpenAI, currently utilizing NVIDIA's GPUs for its artificial intelligence models, is contemplating a change in its hardware approach. CEO Sam Altman is reportedly discussing with silicon designer Broadcom and other chip manufacturers about creating an OpenAI-specific chip. This potential shift aligns with Altman's broader vision of reducing the company's reliance on NVIDIA's GPUs.
OpenAI's ambitious vision beyond proprietary AI chip
Altman's ambitions go beyond just creating a proprietary AI chip for OpenAI. He aims to improve the supply of components and necessary infrastructure, including power infrastructure and data centers, to run powerful AI models. As part of this initiative, OpenAI is recruiting former Google employees who have experience working on Google's tensor processing unit (TPU).
OpenAI's potential partnership with Broadcom
Earlier this year, reports indicated that Altman was seeking to raise billions of dollars to establish a network of semiconductor factories. A potential partnership with Broadcom aligns well with this vision, given the company's extensive experience in designing custom AI accelerators. Notably, Broadcom has previously collaborated with Google on the TPU project and specializes in creating custom ASIC solutions.
OpenAI's potential utilization of Broadcom's offerings
Broadcom, a fabless chip designer, offers a wide range of silicon solutions crucial for data center operations. These include networking components, PCIe controllers, SSD controllers, and custom ASICs. OpenAI could leverage Broadcom's complete vertical stack of products to fulfil its data center needs. Additionally, Broadcom's expertize in inter-system and system-to-system communication technologies could provide OpenAI with a solution for their AI infrastructure requirements.
