Meta Extends Custom Chip Partnership with Broadcom Through 2029 in AI Infrastructure Push

Meta has extended its custom chip partnership with Broadcom through 2029, securing a commitment of more than one gigawatt of computing capacity to support AI across its platforms, according to Tech Startups.

The deal gives Meta a longer runway for its in-house silicon strategy, which is designed to reduce its dependence on Nvidia GPUs for AI training and inference workloads. Meta has been developing its own chips, called Meta Training and Inference Accelerators, since 2020.

The Broadcom partnership covers the design and manufacturing of custom AI chips tailored to Meta's specific workloads, including recommendation systems, content ranking, and generative AI features across Facebook, Instagram, and WhatsApp.

ASML, a critical supplier in the semiconductor industry, raised its 2026 outlook this month due to accelerated expansion plans by customers tied to AI demand. TSMC, the leading advanced chip manufacturer, is also spending above historical levels to meet orders for AI systems, servers, smartphones, and PCs.

Google is separately in discussions with Marvell to develop new AI chips, while former Apple chip executive Gerard Williams is leading a startup called Nuvacore focused on building a better CPU for the AI era.

More than half of planned U.S. data center builds have been delayed or canceled due to supply shortages, power constraints, and reliance on Chinese imports for components. Local resistance to AI data centers is also growing, with $18 billion in projects blocked and $46 billion delayed over the past two years due to community pushback.

Maine lawmakers passed legislation this month to freeze approvals for new data centers requiring over 20 megawatts of power until October 2027 to study their impact on the power grid.