Nvidia is poised to unveil a new processor tailored for AI inference computing, specifically catering to OpenAI’s needs. Scheduled for announcement at the upcoming GTC developer conference, this shift marks a pivotal change in Nvidia’s business strategy amid the ongoing AI boom. Historically, Nvidia has dominated the GPU market—holding over 90% share with its Hopper, Blackwell, and Rubin series, primarily designed for AI model training. However, the industry is shifting focus from model training to inference, necessitating a new architecture. The forthcoming chip will leverage technology from Groq, a startup specializing in language processing units (LPUs) for enhanced efficiency in inference tasks, rivaling Google’s TPUs. AI inference is crucial for responding to user queries, and many companies find Nvidia’s GPUs insufficient for these workloads due to cost and energy consumption. This innovation comes amid broader $110 billion funding announcements, including a $30 billion investment from Nvidia, positioning them strategically in the evolving AI landscape.
Source link
