Meta Platforms plans to begin production of its internally designed AI chip, codenamed Iris, in September [1].
This move represents a strategic shift toward hardware independence. By developing its own silicon, Meta seeks to lower its dependence on external graphics processing unit (GPU) suppliers, most notably Nvidia, while scaling its infrastructure to meet the demands of generative AI.
According to an internal memo, the company intends to double its current computing capacity [2]. This expansion is part of a broader push to increase total AI compute power to 14 gigawatts by 2027 [1].
The Iris chip is designed to optimize the workloads associated with Meta's large-scale AI models. By controlling the hardware layer, the company can better align chip architecture with its specific software requirements, a move that could potentially lower operational costs and power consumption.
While the specific manufacturing location for the Iris chips has not been disclosed, the effort is part of a wider industry trend where big tech firms design custom silicon to gain a competitive edge in processing speed and energy efficiency [1], [2].
The timeline for the September production start suggests that Meta is moving quickly to integrate these chips into its data centers. This acceleration comes as the global demand for AI compute continues to outpace the availability of high-end commercial chips [3].
“Meta plans to begin production of its in-house AI chip, codenamed Iris, in September”
Meta's transition to custom silicon signals a move toward vertical integration to hedge against supply chain vulnerabilities and the high cost of third-party GPUs. By targeting 14 GW of power, the company is positioning itself to train and deploy larger, more complex AI models without being throttled by the production schedules or pricing of external vendors.
