On October 13, 2025, OpenAI unveiled a strategic partnership with Broadcom. The two entities will collaborate to develop custom AI chips and roll out an inference infrastructure boasting a capacity of up to 10 gigawatts. The deployment of this infrastructure is set to occur in phases, spanning from 2026 to 2029. In this partnership, OpenAI will take charge of designing the chip architecture, whereas Broadcom will be in charge of the development and deployment processes. This move represents a significant stride for OpenAI in its quest for autonomy within the AI infrastructure realm.