# MTIA
Latest news and articles about MTIA
Total: 3 articles found
Meta Accelerates Own AI Silicon Push with Four New MTIA Chips, Betting on in‑House Efficiency
Meta announced four new AI chips under its MTIA programme, with MTIA 300 already in production and three further models slated through 2027. The chips aim to accelerate both training and inference for generative features and ranking systems, reflecting a broader industry move toward custom silicon to cut costs and control performance.
Meta Accelerates Push for Custom AI Chips to Power Generative Models and Wean Off Nvidia
Meta revealed plans to roll out four in‑house MTIA chips through 2027, with MTIA 300 already in production and later chips slated for inference-heavy generative AI workloads. The move signals a deliberate strategy to diversify suppliers, lower operating costs, and pair continued purchases of Nvidia/AMD hardware with bespoke silicon aimed at Meta’s unique demands.
Meta Retreats on Ambitious In‑House AI Chip — Turns to AMD, Nvidia and Google to Fill the Gap
Meta has paused work on a high‑end internal AI training chip, Olympus, after design and stability issues, opting instead for a simpler internal design and large purchases from AMD, Nvidia and Google. The move underscores the difficulty of competing with Nvidia’s performance and software ecosystem and signals a pragmatic industry shift toward external partnerships to secure AI compute capacity.