Meta Expands Broadcom Partnership to Boost AI Chip Development
Meta has announced an expanded partnership with Broadcom to develop multiple generations of custom artificial intelligence processors, as the company accelerates efforts to strengthen its computing infrastructure for AI-driven features across its platforms.
The agreement, revealed on April 14, extends the collaboration through to 2029. It includes an initial commitment exceeding one gigawatt of computing capacity, which is roughly equivalent to the energy required to power around 750,000 average homes in the United States.
Strategic Shift Towards Custom AI Hardware
As part of the revised arrangement, Broadcom’s chief executive, Hock Tan, will step down from Meta’s board. However, he will continue to contribute in an advisory capacity, focusing on the company’s custom chip strategy.
This move reflects a broader trend within the technology sector. Major firms are increasingly designing in-house processors to reduce reliance on external suppliers, particularly as demand for artificial intelligence capabilities rises sharply. Consequently, companies aim to gain greater control over performance, cost, and scalability.
Rising Demand Fuels Industry-Wide Chip Innovation
The rapid expansion of AI applications has significantly increased the need for advanced computing power. Therefore, technology leaders such as Meta, Google, and Amazon are investing heavily in proprietary chip development to support their growing ecosystems.
Broadcom has emerged as a key beneficiary of this shift. The company collaborates closely with clients to design tailored processors while also providing essential infrastructure software. Following the announcement, Broadcom’s shares rose in extended trading, whereas Meta’s stock remained largely unchanged.
Meta’s chief executive, Mark Zuckerberg, stated that the partnership would help establish the large-scale computing foundation required to deliver advanced AI capabilities to a global user base.
Multi-Phase Rollout of AI Infrastructure
Meta recently outlined plans for four new chips under its Meta Training and Inference Accelerator programme. The initial processor, known as MTIA 300, is already being used to support ranking and recommendation systems.
Moreover, the company confirmed that three additional chip generations are scheduled for release by 2027. These future processors will focus primarily on inference tasks, enabling AI systems to respond more efficiently to user queries.
The partnership with Broadcom marks the first phase of what Meta describes as a sustained, multi-gigawatt expansion of its AI infrastructure. In addition, Broadcom’s Ethernet networking technology will play a central role in connecting Meta’s rapidly growing clusters of AI systems.
Board Changes Announced Alongside Deal
Separately, Meta disclosed that Tracey Travis, who has served on the company’s board since 2020, will not seek re-election at its upcoming annual shareholder meeting.
This leadership update comes alongside Meta’s continued investment in artificial intelligence and infrastructure, underscoring a period of strategic transformation for the company.
With inputs from Reuters
