Graphics processing leader Nvidia is developing a specialized processor aimed at helping companies like OpenAI create AI systems that operate with greater speed and efficiency, according to a Wall Street Journal report published Friday that cited sources with knowledge of the project.
The company is working on technology for “inference” computing, which enables artificial intelligence models to process and respond to user questions, the publication reported.
According to sources familiar with the development, Nvidia plans to reveal this new platform during its GTC developer conference scheduled for next month in San Jose, and the system will feature technology from startup company Groq.
Neither Nvidia nor OpenAI provided immediate responses when contacted for verification of the report.
Previous reporting indicated that OpenAI has expressed dissatisfaction with how quickly Nvidia’s current technology can generate responses for ChatGPT users, particularly for complex tasks like software development and AI-to-AI communication.
According to a source, OpenAI requires new hardware that could eventually handle approximately 10% of the company’s inference processing requirements.
The maker of ChatGPT had been exploring partnerships with emerging companies including Cerebras and Groq to obtain faster inference chips, sources revealed. However, Nvidia secured a $20 billion licensing agreement with Groq that ended OpenAI’s negotiations, according to one source.
Last September, Nvidia announced plans to invest up to $100 billion in OpenAI through an arrangement that provided the chipmaker with an ownership stake while giving OpenAI the funding needed to purchase advanced processors.
