Meta and Broadcom have officially extended their strategic alliance to co-develop custom artificial intelligence processors. The new Meta Broadcom AI chip deal, announced on Tuesday, extends the collaborative partnership through 2029. This agreement centers on producing several generations of specialized silicon to power generative artificial intelligence features across popular social media applications like Instagram, WhatsApp, and Threads.
The partnership launches with an initial commitment to deploy more than one gigawatt of computing capacity. To put this massive scale into perspective, one gigawatt generates enough energy to power approximately 750,000 average homes in the United States. This initial phase serves as the foundation for an even larger, sustained multi-gigawatt infrastructure rollout designed to handle the surging demands of advanced computing.
Developing Next-Generation MTIA Processors
At the core of this expanded agreement is the Meta Training and Inference Accelerator initiative. Broadcom will supply critical chip design, packaging, and advanced networking technology to support this proprietary silicon roadmap. The resulting processors will represent a major technical milestone, marking the first artificial intelligence chips manufactured using an advanced 2-nanometer process.
These specialized components are designed primarily for inference tasks, which manage how artificial intelligence models process and react to user inquiries in real time. Because these systems prioritize low-precision processing to maximize efficiency, the supporting infrastructure requires near-zero latency. Meta recently unveiled a roadmap featuring four new chips, noting that the first generation, known as the MTIA 300, is already operational within the company’s ranking and recommendation systems. Three additional versions are expected to be released by 2027.
Advanced Networking to Eliminate Data Bottlenecks
Scaling up artificial intelligence data centers requires massive data transfer capabilities. To support the dense computing clusters required for these workloads, Broadcom is delivering a comprehensive suite of Ethernet networking solutions. This hardware creates a standards-based, low-latency fabric essential for moving data rapidly across tens of thousands of computing nodes.
The deployment incorporates high-radix Ethernet switches, optical connectivity products, PCIe switches, and high-speed SerDes capabilities. Together, these technologies enable seamless scaling across the network, eliminating congestion during intensive processing tasks. This future-proof Ethernet backbone ensures that the specialized processor grids remain constantly utilized and gracefully handle evolving memory requirements, drastically reducing the total cost of ownership over the lifespan of the hardware.
Executive Board Changes and Advisory Shifts
Alongside the hardware expansion, the companies announced significant leadership transitions. Broadcom Chief Executive Officer Hock Tan informed Meta that he will not seek reelection to the social media giant’s board of directors. Tan, who joined the board in 2024, will transition into a specialized advisory role focusing directly on the company’s custom chip strategy.
This executive transition aligns with the deepening technical collaboration between the two technology leaders. In a separate corporate governance update, Meta also announced that Tracey Travis, a board member since 2020, will not stand for re-election at the company’s upcoming annual shareholder meeting.
Strategic Push for Personal Superintelligence
The expanded partnership represents a crucial step in Meta’s broader strategy to build independent hardware capabilities. Major technology companies are increasingly designing custom processors to reduce their reliance on costly, third-party hardware from dominant suppliers like Nvidia. Meta previously announced plans to invest up to $135 billion in artificial intelligence infrastructure this year to remain competitive with industry rivals such as OpenAI and Anthropic.
Meta founder and Chief Executive Officer Mark Zuckerberg emphasized the importance of this infrastructure. He stated that the collaboration will help build the massive computing foundation needed to deliver personal superintelligence to billions of users globally. By securing a reliable pipeline for custom silicon, the company can seamlessly sustain its aggressive data center expansion.
Market Reaction and Financial Impact
The financial markets responded positively to the deepened collaboration. Following the announcement, shares of Broadcom experienced a 3% to 3.5% increase in extended premarket and after-hours trading. The chipmaker has emerged as a primary beneficiary of the generative computing boom, boasting a $1.8 trillion market capitalization and reporting a 25% revenue growth over the past twelve months.
Meanwhile, Meta’s stock remained relatively stable following the news. Broadcom’s leadership reassured investors during a recent earnings call that the processor roadmap is progressing well, with next-generation components already shipping. This deal further embeds Broadcom into the core hardware plans of major cloud hyperscalers, cementing its position as a critical supplier in the rapidly expanding artificial intelligence infrastructure market.
