What's happening
Meta Platforms announced on April 14, 2026, an expanded multi-year partnership with Broadcom to co-develop custom AI processors through 2029. The initial commitment exceeds 1 gigawatt of computing capacity for Training and Inference Accelerators, representing the first phase of a multi-gigawatt rollout. The 1 GW capacity is sufficient to power roughly 750,000 U.S. homes on average. Meta CEO Mark Zuckerberg stated the company is partnering with Broadcom across chip design, packaging, and networking to build out the massive computing foundation needed to deliver personal superintelligence to billions of people. The partnership builds on Meta's existing MTIA chip development, with four new versions unveiled in March 2026. The first MTIA 300 is already in production powering ranking systems, with three additional versions scheduled through 2027.
Why it matters for markets
The deal strengthens Meta's position in the AI hardware arms race as the company plans to spend up to $135 billion on AI in 2026, a commitment made in January. The multi-gigawatt partnership provides Meta with greater independence from external AI hardware suppliers while securing a significant revenue stream for Broadcom. Broadcom CEO Hock Tan emphasized that Meta's custom accelerator roadmap will scale to multiple gigawatts in 2027 and beyond, contradicting recent analyst reports questioning the program's viability. The partnership complements Meta's broader AI infrastructure strategy, which includes commitments to up to 6 GW of AMD GPUs and millions of Nvidia chips. Markets responded positively to the announcement, with Broadcom shares rising 3-3.5% in extended trading and gaining 2.94% during regular hours to close at $396.72, contributing to the company's 10% year-to-date gain versus the S&P 500's 2% advance.
Sectors and assets to watch
The custom AI chip partnership highlights the growing importance of application-specific integrated circuits in the semiconductor sector. Broadcom, with its $1.88 trillion market cap and expertise in custom ASICs for hyperscalers, stands to benefit from the multi-year revenue commitment as demand for specialized AI processors intensifies. Meta's $1.70 trillion market cap and $200.97 billion in revenue position it as a major customer driving custom chip development. The deal also affects the broader AI infrastructure landscape, where companies like Advanced Micro Devices and Nvidia compete for data center accelerator market share alongside custom silicon providers.
What to watch next
Investors should monitor the timeline for Meta's three additional MTIA chip versions scheduled through 2027 and any updates on the multi-gigawatt scaling beyond the initial 1 GW commitment. Broadcom's transition of CEO Hock Tan from Meta's board to an advisory role on custom chip strategy following his decision not to stand for reelection will be worth tracking for partnership developments. The competitive response from other AI chip suppliers and Meta's progress toward its $135 billion AI spending target for 2026 will provide further insight into the custom silicon market's evolution.