Meta extends Broadcom partnership to scale custom AI chip infrastructure

Meta Platforms has extended its partnership with Broadcom to accelerate the development of custom AI chips and expand its computing infrastructure, as part of its broader push into artificial intelligence.

The multi-year agreement, which now runs through 2029, focuses on co-developing multiple generations of AI accelerators tailored to Meta’s growing workload requirements. The collaboration includes an initial deployment exceeding one gigawatt of compute capacity, with plans to scale toward multi-gigawatt infrastructure over time.

At the center of the partnership is Meta’s strategy to design its own silicon, reducing reliance on third-party chip providers while improving efficiency for AI training and inference. Custom chips, including those from Meta’s in-house accelerator program, are expected to play a key role in powering next-generation AI models and large-scale data center operations.

Broadcom’s role extends beyond chip development to include networking technologies that support large AI clusters, enabling faster data movement and more efficient infrastructure scaling. The collaboration highlights the increasing importance of vertically integrated AI systems, where hardware and software are co-optimized for performance.

The move reflects a broader industry trend, with major technology companies investing heavily in custom silicon to meet rising AI compute demands. As enterprises scale AI capabilities, reliance on specialized chip architectures is becoming critical to managing cost, performance, and energy efficiency.

For Meta, the extended partnership reinforces its long-term commitment to building dedicated AI infrastructure, positioning the company to compete more aggressively in the evolving AI landscape. For Broadcom, it further strengthens its role as a key enabler in the global AI hardware ecosystem, supporting hyperscale companies in developing next-generation computing platforms.