Meta is partnering with Broadcom for the development of multiple generations of the hyperscaler’s Meta Training and Inference Accelerator (MTIA) chips.

The agreement, the financial terms of which have not been shared, includes a commitment to deploy more than one-gigawatt (GW) of the chips in its first phase, with a “multi-gigawatt rollout” ultimately planned. Timelines and locations for deployment have also not been detailed.

In a blog post announcing the partnership, Meta said that it will work with Broadcom across chip design, advanced packaging, and networking to help build out the “massive computing foundation we need to deliver real-time AI experiences to billions of people.”

Based on Broadcom’s custom AI accelerator (XPU) platform, the co-development will span multiple silicon generations and integrate with the chip company's Ethernet technologies to support high-bandwidth networking across Meta’s AI compute clusters.

“Meta is partnering with Broadcom across chip design, packaging, and networking to build out the massive computing foundation we need to deliver personal super-intelligence to billions of people,” Meta CEO Mark Zuckerberg noted. “As we roll out more than 1GW of our custom silicon to start and then multiple gigawatts over time, this partnership will give us greater performance and efficiency for everything we’re building.”

Broadcom CEO Hock Tan added: “We are pleased to expand our strategic collaboration with Meta as they pioneer the next frontier of artificial intelligence. This initial MTIA deployment is just the beginning of a sustained, multigeneration roadmap to serve the trajectory of massive growth over the next few years that highlights Broadcom’s unmatched leadership in AI networking and the power of our foundational XPU custom accelerator platform.”

The agreement will also see Tan leave Meta’s board of directors and move into an advisory role at the company, where he’ll provide guidance on Meta’s custom silicon roadmap and future infrastructure investments.

The announcement comes a month after Meta unveiled the next four generations of its MTIA chip, dubbed the MTIA 300, 400, 450, and 500. Designed to support generative AI workloads, the company said the chips have either already been deployed or are scheduled for deployment in the next 18 months.

Each new chip will offer improvements in compute, memory bandwidth, and efficiency, Meta said at the time.

Separately, Broadcom last week announced it had entered into a long-term agreement with Google to develop and supply future generations of its tensor processing units (TPUs), in addition to signing a supply assurance agreement to supply “networking and other components” to be used in Google’s next-generation AI racks until 2031.

That deal followed an agreement last October where Broadcom announced it was working with generative AI giant OpenAI to co-develop custom AI hardware.

At the time, the two companies said they would work together on AI accelerators and Ethernet solutions, with deployment targeted to start in the second half of 2026, resulting in 10GW of chips coming online by 2029. However, since the announcement was first made, no updates on the project have been provided by either company.