Meta Platforms Inc. has extended its partnership with Broadcom Inc., focusing on the design and deployment of custom artificial intelligence processors. The social media giant is committing to an initial deployment of one gigawatt of its Meta Training and Inference Accelerators (MTIA), custom-designed chips for AI workloads.
The expanded collaboration will see Meta deploy multiple gigawatts of these chips, which are based on Broadcom's technology. Notably, the new MTIA chips will utilize a 2-nanometer process, a first for custom AI silicon in the industry. This move provides Meta with an alternative to the high-demand, expensive GPUs from Nvidia and AMD, offering smaller, more cost-effective chips optimized for specific AI tasks.
Meta CEO Mark Zuckerberg stated the MTIA chips leverage Broadcom's design, packaging, and networking technologies to build the "massive computing foundation" needed for "personal superintelligence." This development counters earlier reports of MTIA chip production challenges, with Broadcom CEO Hock Tan confirming the roadmap is "alive and well" and scaling to multiple gigawatts by 2027.
This partnership is part of Meta's significant investment in AI infrastructure, with plans to spend over $135 billion on capital expenditures in fiscal 2026. The company has previously announced deals for GPUs from AMD, chips from Nvidia, and custom processors from Arm Holdings, alongside commitments for rented AI infrastructure.