Meta has significantly expanded its collaboration with Broadcom, committing billions of dollars to jointly design and manufacture custom chips for its artificial intelligence operations. Under the agreement, the two companies will co-develop multiple generations of training and inference accelerators, with the partnership framework extending through 2029.

The deal includes plans to develop chips capable of delivering more than 1 gigawatt of computing capacity. This agreement represents the first phase of what is expected to be a multi-billion-dollar deployment program, with future expansion potentially reaching several gigawatts. For perspective, 1 gigawatt of power is enough to supply electricity to approximately 750,000 U.S. households.

At the same time as this major business development, Meta’s board structure is undergoing a key change. Broadcom CEO Hock Tan announced that he will not seek re-election to Meta’s board. Tan joined the board in early 2024 to help strengthen Meta’s hardware infrastructure and supply chain capabilities.

However, as the scale of commercial cooperation between the two companies in custom chip development has grown rapidly, Tan’s decision to step down is aimed at avoiding potential conflicts of interest and ensuring transparency in corporate governance. He will instead serve as a strategic advisor on Meta’s custom chip roadmap, continuing to provide guidance on technological development.

Broadcom has already been working with Meta on custom chip development, a collaboration that has become increasingly critical as Meta invests tens of billions of dollars in infrastructure to compete in the AI race. By developing its own chips, Meta aims to reduce costs and lessen its reliance on third-party suppliers such as Nvidia and AMD, even as it continues to spend heavily on purchasing AI processors and related hardware from those companies.

On the technology front, the partnership between Meta and Broadcom is advancing toward cutting-edge innovation. The two companies plan to develop one of the industry’s first AI accelerators based on a 2-nanometer process node, significantly improving data center efficiency and performance. Meta has outlined a roadmap that includes multiple chip generations — MTIA 300, 400, 450, and 500 — designed to optimize recommendation algorithms, content ranking, and large-scale inference workloads across its platforms.

Market Interpretation:

For Broadcom, this deal marks a major expansion of its custom AI chip business, which has become a key driver of revenue growth and investor enthusiasm. The partnership also positions Broadcom as a potential alternative to Nvidia in the AI semiconductor market.

With Meta as a core customer, the deepening relationship between the two companies is likely to have a significant impact on the competitive landscape of the global AI chip industry.