Meta Scales AI Infrastructure with Broadcom Custom Chip Expansion

Meta and Broadcom have announced a sweeping extension of their partnership to design custom in-house AI accelerators, securing the collaboration through 2029. The agreement centers on Meta's Training and Inference Accelerators (MTIA), with an initial commitment to deploy one gigawatt of power capacity for these chips, with plans to scale to multiple gigawatts over time. This move is part of a broader industry trend where hyperscalers develop application-specific integrated circuits (ASICs) to bypass the high costs and supply constraints associated with Nvidia and AMD GPUs. While peers like Google and Amazon offer their custom silicon via cloud platforms, Meta utilizes its MTIA chips exclusively for internal operations to build its computing foundation. Alongside the hardware deal, Meta announced significant changes to its corporate governance. Broadcom CEO Hock Tan, who joined the board in 2024, has decided not to stand for reelection. Additionally, Tracey Travis will depart the board after serving since 2020. Market reaction was positive for the chip designer, with Broadcom shares rising 3% in extended trading. The stock has gained approximately 10% so far in 2026, significantly outperforming the S&P 500's 2% gain. Meta's stock remained flat following the announcement.