OpenAI, the company behind ChatGPT and the GPT-4 large language model, is reportedly considering the creation of custom silicon to power its next-generation AI models. According to Reuters, insider sources have suggested that the company is also exploring the acquisition of chip design firms. While a final decision has not been made, discussions from last year revealed OpenAI's struggle with the increasing scarcity and costs of AI chips, with NVIDIA as its main supplier. OpenAI CEO Sam Altman has expressed concerns about the shortage of GPUs, which NVIDIA dominates with an 80% market share in AI-optimized chips.
In 2020, OpenAI relied on a massive supercomputer provided by Microsoft, a major investor in the company, which utilized 10,000 NVIDIA GPUs. This setup is crucial for the functioning of ChatGPT, but it comes with a significant cost. Analyst Stacy Rasgon estimates that each interaction with ChatGPT costs around 4 cents. If ChatGPT queries were to reach just a tenth of Google's search volume, the initial GPU investment would amount to a staggering $48.1 billion, with an annual expenditure of approximately $16 billion for ongoing operations. OpenAI declined to comment on the matter. The potential development of custom silicon indicates a strategic move towards greater self-sufficiency and cost optimization to support further AI advancements.
