OpenAI和博通联合开发部署10吉瓦AI加速器

Core Insights - OpenAI and Broadcom have announced a strategic partnership to jointly develop a custom AI accelerator with a computing power of 10 gigawatts [1] - The project is set to begin in the second half of 2026 and is expected to be completed by the end of 2029 [1] - The collaboration emphasizes the importance of custom accelerators and Ethernet in building scalable network architectures for AI data centers [1] Group 1 - OpenAI will be responsible for the design of the AI accelerator and system architecture, while collaborating with Broadcom for development and deployment [1] - The system architecture will utilize Broadcom's high-performance Ethernet solutions to enable vertical and horizontal scaling of AI data centers [1] - The custom accelerator will be deployed in OpenAI's own facilities and partner data centers to meet the rapidly growing demand for AI computing power globally [1] Group 2 - OpenAI's CEO, Sam Altman, stated that this partnership is a crucial step in building the necessary infrastructure to unlock AI's potential, benefiting individuals and businesses [1] - The development of proprietary accelerators will enhance the partner ecosystem and collectively build the capabilities needed to advance AI [1] - Broadcom's semiconductor solutions group president, Charlie Kawas, noted that the combination of custom AI accelerators and Ethernet solutions will optimize performance and cost for next-generation AI infrastructure [2]