Meta Platforms will begin manufacturing its custom AI chip, known as Iris, in September [1].
This move represents a strategic shift toward hardware independence. By developing its own silicon, Meta seeks to reduce its reliance on external chip suppliers during a period of strained global supply chains [2].
The company intends to use the Iris chip to double its current AI computing capacity [3]. According to internal memos and reports, Meta is targeting a total AI compute capacity of 14 gigawatts by next year [1].
Developing in-house chips allows the company to optimize hardware specifically for its own large-scale AI models. This vertical integration is a growing trend among AI giants striving to maintain a competitive edge in processing power, and efficiency [2].
The production timeline places the start of manufacturing in September 2026 [1]. This acceleration is designed to meet the increasing demands of Meta's generative AI initiatives and infrastructure needs.
By reaching the 14 gigawatt target, Meta aims to significantly scale its ability to train and deploy complex models [1]. The transition to custom silicon is expected to provide more control over the performance, and cost of its data center operations [2].
“Meta will begin manufacturing its custom AI chip, known as Iris, in September.”
Meta's transition to custom silicon mirrors a broader industry trend where hyperscalers move away from off-the-shelf hardware to avoid supply bottlenecks and reduce costs. By targeting 14 gigawatts of capacity, Meta is positioning itself to handle the massive computational loads required for next-generation AI, potentially reducing its long-term dependency on dominant chip vendors like Nvidia.
