Meta Extends Custom AI Chip Deal to 2029 with 1 GW Commitment and 2nm MTIA Chips
Meta and Broadcom have extended their custom AI chip deal through 2029 to cover MTIA generations with an initial commitment of over 1 gigawatt of compute capacity. The MTIA chips will be the first 2nm AI silicon and use Broadcom's XPU platform in design, packaging and networking.
1. Deal Extension and Capacity Commitment
Meta and Broadcom agreed to expand their partnership through 2029, anchoring the deal with an initial commitment to deploy over 1 gigawatt of AI compute capacity. This capacity commitment is described as the opening installment in a projected multi-gigawatt buildout across future MTIA accelerator generations.
2. MTIA Chip Specifications and Deployment
The MTIA chips will be the first AI silicon built on a 2nm process node. Broadcom's XPU platform will span the design, packaging and networking layers to develop custom AI accelerators tailored for Meta's training and inference workloads.
3. Broadcom's Role and Executive Change
As part of the agreement, Broadcom CEO Hock Tan will step down from Meta's board to serve in an advisory capacity focused on Meta's silicon roadmap. This shift aims to streamline collaboration on chip design and strategic planning for next-generation AI hardware.
4. Broader AI Infrastructure Strategy
The Broadcom deal complements Meta's broader hardware strategy, which includes multiyear agreements for up to 6 gigawatts of AMD GPUs and millions of Nvidia and Arm-based chips. Meta is planning up to $135 billion in AI infrastructure spend for 2026 and building 31 data centers to support its expanding compute needs.