Meta and Broadcom deepen chip collaboration; Broadcom CEO leaves Meta board to become advisor.
```
Broadcom and Meta announced on Monday an expanded multi-generation strategic partnership, providing underlying technological support for the large-scale deployment of Meta's custom AI chips. The agreement has been extended to 2029, marking a new stage of deep binding between the two tech giants in the AI infrastructure sector.
According to a joint statement from both parties, the first phase of this collaboration commits to a deployment scale exceeding 1 gigawatt (GW), serving as the starting point for an ongoing multi-gigawatt expansion plan, aimed at providing real-time generative AI features and so-called “personal superintelligence” services to billions of users of WhatsApp, Instagram, Threads, and other applications.
This scale indicates that Meta is dramatically increasing its bet on custom silicon and also highlights Broadcom’s central position in the field of AI custom accelerators.
Notably, Broadcom President and CEO Hock Tan will step down as a member of Meta’s board and take on an advisory role, focusing on providing guidance for Meta’s custom silicon roadmap and infrastructure investment strategy. This move signals that the two companies’ cooperation is shifting from capital-level collaboration to deeper technological and business synergy.
Driven by this news, Broadcom’s stock price rose 3.3% after hours.
Multi-Gigawatt Deployment Plan: Custom Silicon Accelerates Expansion
The core product of this cooperation is Meta's self-developed MTIA (Meta Training and Inference Accelerator) chip. Broadcom will leverage its XPU custom accelerator platform to provide MTIA with full-stack technical support from chip design and advanced packaging to network interconnect.
The statement shows that Broadcom's XPU platform can tightly couple logic units, memory, and high-speed I/O, supporting current deployments while establishing a highly scalable blueprint for multi-generation joint development.
Meta emphasized that MTIA is the core pillar of its broader silicon strategy, optimizing both performance and total cost of ownership (TCO) in large-scale deployment by precisely matching dedicated hardware to specific workloads.
Meta founder and CEO Mark Zuckerberg stated in the announcement: “Meta is working closely with Broadcom in chip design, packaging, and networking to build the massive computing foundation needed to deliver personal superintelligence to billions of people.”
Ethernet Network Architecture: Supporting Large-Scale Cluster Expansion
In addition to custom chips, Broadcom will also provide Meta with its Ethernet networking solutions, covering rack-level scale-up, node-to-node scale-out, and cross-domain scale-across needs.
Specifically, Broadcom’s high-radix Ethernet switches, optical interconnect products, PCIe switches, and high-speed SerDes technology will form a low-latency network infrastructure based on standard protocols. This backbone network is crucial for eliminating network congestion during high-intensity AI workloads at tens of thousands of nodes scale.
Since MTIA is primarily designed for inference and low-precision computational tasks, its supporting interconnect infrastructure must feature near-zero latency.
Broadcom stated that its Ethernet-based rack-level interconnect solution will ensure the continued efficient operation of both current and future MTIA clusters, and significantly reduce TCO throughout Meta’s multi-generation infrastructure lifecycle.
Hock Tan’s Role Adjustment: Partnership Moves Toward Deep Tech Integration
Hock Tan stepping down from Meta's board and transitioning to an advisory role is another market-worthy detail in this announcement.
The joint statement says Hock Tan will provide guidance as an advisor for Meta’s custom silicon roadmap and help shape its infrastructure investment direction.
This arrangement means that as the partnership scales up, the companies’ leadership interaction will shift from a formal board governance structure to a more operational technical advisory mechanism, further enhancing the depth of joint R&D and system-level optimization collaboration.
Hock Tan stated: “This initial deployment of MTIA is just the beginning, heralding a continuous multi-generation roadmap. It also highlights Broadcom’s unparalleled leadership in the AI networking field and the strength of our XPU custom accelerator platform.”
Risk Warning and DisclaimerThe market has risks. Investment requires caution. This article does not constitute individual investment advice and does not take into account the specific investment objectives, financial condition, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment made accordingly is at their own risk. ```