Broadcom Expands AI Custom Silicon Footprint with Meta and Alphabet Extensions

Broadcom (AVGO) continues to solidify its dominance in the custom AI accelerator market, announcing a major extension of its partnership with Meta Platforms. The agreement focuses on the development of custom AI chips through 2029, with an initial commitment of one gigawatt of capacity. To facilitate this strategic roadmap, Broadcom CEO Hock Tan will transition from Meta's board to an advisory role. This expansion comes as hyperscalers increasingly seek alternatives to expensive general-purpose GPUs to optimize for inference and energy efficiency. By leveraging application-specific integrated circuits (ASICs), companies like Meta can achieve higher performance and lower operational costs. The partnership has already yielded four generations of Meta Training and Inference Accelerator (MTIA) chips, with the MTIA 300 currently in use and the 400, 450, and 500 series designed for comprehensive AI workloads. Broadcom's growth trajectory extends beyond Meta. The company recently extended its TPU development partnership with Alphabet and secured a deal to supply Anthropic with 3.5 gigawatts of chips starting in 2027. Broadcom currently holds $21 billion in TPU orders from Anthropic for the current year alone. These wins align with Broadcom's ambitious target of delivering $100 billion in custom AI chips by fiscal 2027. This represents a massive projected increase from the company's fiscal 2025 AI revenue of approximately $20 billion. Furthermore, the surge in custom silicon demand is expected to drive growth in Broadcom's data center networking segment, particularly its Tomahawk Ethernet switching solutions, as AI clusters continue to scale.