OpenAI is on track to unveil its custom ASIC chip, potentially in Q4 2025. This initiative aims to curb the escalating costs associated with training large language models. Currently, OpenAI heavily depends on Nvidia chips. This in-house chip development, codenamed Tigris, is expected to give OpenAI greater control over its AI pipeline and significantly reduce operating expenses. The chip, built in partnership with Broadcom and fabricated by TSMC, is designed for training efficiency, possibly surpassing current solutions in specific workloads. Despite potential challenges, the launch signals OpenAI’s ambition to reshape the vertical integration of AI development.