OpenAI's First In-House AI Chip to Tape Out at TSMC in H1
OpenAI is reportedly preparing to tape out its first proprietary artificial intelligence chip at Taiwan Semiconductor Manufacturing Company (TSMC) during the first half of this year. This strategic move aims to reduce the company's heavy reliance on NVIDIA and other external AI chip suppliers, thereby strengthening OpenAI's bargaining power in the competitive hardware market. The new chip is specifically designed for training large language models. According to sources cited by Reuters, OpenAI targets mass production of this in-house silicon by 2026. The tape-out process, which marks the final stage of design before manufacturing, typically costs tens of millions of dollars and requires approximately six months to produce a finished product, unless expedited through premium payments. However, success is not guaranteed; if the initial chip fails functional tests, OpenAI will need to diagnose the issues and restart the development cycle. This initiative highlights a growing trend among major tech firms to develop custom silicon to optimize performance and control supply chains amidst increasing demand for AI computing resources.
Wire timeline
OpenAI's First In-House AI Chip to Tape Out at TSMC in H1
OpenAI is reportedly preparing to tape out its first proprietary artificial intelligence chip at Taiwan Semiconductor Manufacturing Company (TSMC) during the first half of this year. This strategic move aims to reduce the company's heavy reliance on NVIDIA and other external AI chip suppliers, thereby strengthening OpenAI's bargaining power in the competitive hardware market. The new chip is specifically designed for training large language models. According to sources cited by Reuters, OpenAI targets mass production of this in-house silicon by 2026. The tape-out process, which marks the final stage of design before manufacturing, typically costs tens of millions of dollars and requires approximately six months to produce a finished product, unless expedited through premium payments. However, success is not guaranteed; if the initial chip fails functional tests, OpenAI will need to diagnose the issues and restart the development cycle. This initiative highlights a growing trend among major tech firms to develop custom silicon to optimize performance and control supply chains amidst increasing demand for AI computing resources.
TechNode