Broadcom and Meta have entered into a multi-generation partnership to scale Meta’s AI compute infrastructure through 2029. Under the agreement, Broadcom will provide the foundational technology for the Meta Training and Inference Accelerator (MTIA) chips, beginning with an initial deployment exceeding one gigawatt. The collaboration aims to support a sustained, multi-gigawatt rollout of custom silicon designed to power generative AI features across Meta’s global platforms, including WhatsApp and Instagram.
The partnership centers on Broadcom’s custom XPU accelerator platform, which allows for the tight integration of logic, memory, and high-speed I/O. This co-design approach is intended to optimize performance and total cost of ownership as Meta deploys the industry’s first 2nm AI compute accelerators. To manage the massive data demands of these clusters, Broadcom is also delivering advanced Ethernet networking solutions to eliminate bottlenecks across tens of thousands of compute nodes, ensuring low-latency connectivity for intensive inference workloads.
As part of this expanded alliance, Broadcom CEO Hock Tan will transition from Meta’s Board of Directors into a specialized advisor role. In this capacity, Tan will provide direct guidance on Meta’s custom silicon roadmap and infrastructure investments. “This partnership will give us greater performance and efficiency for everything we’re building,” stated Meta Founder and CEO Mark Zuckerberg, noting that the infrastructure is a prerequisite for delivering “personal superintelligence” to billions of users worldwide.