Meta's Self-Developed AI Chip 'Iris' to Begin Mass Production in September to Cut Computing Costs

CoinFeed reported on July 9, citing Reuters, that Meta plans to mass-produce its self-developed data center AI chip 'Iris' starting from September, as part of its Meta Training and Inference Accelerators 4th-gen project, to enhance the AI capabilities of platforms like Facebook and Instagram and reduce reliance on external GPUs from Nvidia, AMD, etc. An internal memo shows that Iris completed testing in just 6 weeks with no major flaws; Meta plans to deploy 7 gigawatts of computing power this year, increasing to 14 gigawatts by 2027, and AI infrastructure spending in 2024 could reach up to $145 billion. To support the expansion, the company has signed long-term supply agreements with Samsung Electronics, Sandisk, and Sumitomo Electric to address price hikes and shortages of memory and AI chips.