Meta Partners Broadcom to Build Custom AI Chips

Meta has announced a partnership with Broadcom to co-develop multiple generations of custom AI silicon, targeting the compute infrastructure that underpins its AI applications and services. The collaboration extends Meta's existing in-house chip program, the Meta Training and Inference Accelerator, known as MTIA, which currently handles AI training and inference workloads across Meta's platforms. By bringing Broadcom into the development cycle, Meta is deepening its custom silicon strategy rather than relying solely on merchant silicon from vendors like Nvidia. The Broadcom partnership signals Meta's intent to control more of its AI compute stack across successive chip generations, a move that reduces long-term dependence on third-party GPU suppliers and could compress per-unit inference costs over time. Investors and supply chain analysts should watch for how this arrangement affects Meta's capital expenditure trajectory and whether it narrows procurement from existing GPU vendors as custom silicon matures.