Meta Accelerates AI Chip Production to Cut Costs Amid Global GPU Shortage

Meta is beginning production of AI-specific chips in September, with the goal of reducing dependency on costly GPUs due to a worldwide shortage of components.
The chips, developed under the Meta Training and Inference Accelerator (MTIA) program, are being designed with Broadcom and manufactured by Taiwan Semiconductor Manufacturing Company (TSMC).
Meta aims to deploy these chips for various AI operations such as model training for ranking and recommendation algorithms, even while continuing to invest significantly with chipmakers like Nvidia and AMD.
The initiative is part of Meta's larger strategy, which includes anticipated capital expenditures of $125 billion to $145 billion in 2026 primarily for expanding AI capabilities and data center infrastructure.
Industry-wide, other tech companies like OpenAI, Anthropic, Amazon, and Google are similarly developing, or considering developing, their own chips to mitigate reliance on leading GPU suppliers like Nvidia.
Meta has also secured deals with ARM, AMD, and Amazon for additional compute capacity, reflecting a significant effort to boost its infrastructure for AI workloads across the globe.