Key facts
- Meta plans to begin production of its custom AI chip, 'Iris', in September.
- The chip is designed to enhance AI powering Meta's applications.
- Meta aims to double its computing capacity to 14 gigawatts by next year.
- The company is collaborating with Broadcom on chip design and TSMC on manufacturing.
- Meta expects to spend between $125 billion and $145 billion on AI infrastructure this year.
Meta Platforms is set to begin production of its custom-designed AI chip, codenamed 'Iris', in September, aiming to significantly boost its computing capacity. The company plans to double its compute power to 14 gigawatts by next year, a move intended to reduce reliance on external GPU providers like Nvidia and AMD.
The 'Iris' chip is part of Meta's ongoing Meta Training and Inference Accelerator (MTIA) program, which focuses on developing in-house silicon for AI workloads. The chip has reportedly passed its testing phase in approximately six weeks. Meta is collaborating with Broadcom on the chip's design and utilizing TSMC for its manufacturing. Additional components are being sourced from Samsung for RAM, Sandisk for storage, and Sumitomo Electric for fiber optic equipment.
Meta has been investing heavily in AI infrastructure, with expected capital expenditures between $125 billion and $145 billion this year. The company plans to deploy 7 gigawatts of compute capacity this year and double that amount next year. This strategic push into custom chip development aims to lower costs and increase autonomy in powering its AI initiatives, including ranking and recommendation algorithms for its social media applications.
Meta is not alone in this pursuit, as other major tech companies like OpenAI, Amazon, and Google are also developing or considering their own AI chips to manage escalating computational demands and control costs.