Meta and Broadcom are joining forces to develop custom AI silicon.

Meta expands its collaboration with Broadcom for the joint development of a new generation of chips that support artificial intelligence.

Today, Meta announced the expansion of its partnership with Broadcom for the joint development of MTIA (Meta Training and Inference Accelerator) chips designed to support artificial intelligence across all Meta services. The main goal of this collaboration is to optimize costs and enhance the performance of cloud infrastructure.

Over the next two years, the companies plan to develop four generations of MTIA chips that will address ranking, recommendation, and generative AI workloads. Broadcom will also be responsible for the design, packaging of the chips, and providing the network infrastructure to scale computing capabilities.

The competition for leadership in AI silicon is intensifying, with giants like Google and NVIDIA actively developing their solutions. However, Meta is focusing on custom design and chip customization through its partnership with Broadcom, which could provide it with a competitive advantage and strengthen the company's AI ecosystem.

The agreement involves the introduction of over 1 gigawatt of custom silicon, which will serve as the foundation for Meta's strategic plan. This will enable the company to develop AI technologies on a global scale and ensure faster task execution for millions of users. The transition of Hock Tan, president of Broadcom, to other positions will also impact the development of the collaboration.