Meta Accelerates Custom AI Chip Deployment to Reduce Energy Consumption

Meta Platforms (META) stock rose by as much as 2% on Tuesday morning after reports surfaced that the company is accelerating its deployment of custom AI chips. The proprietary processor, code-named MTIA 450 or Arke, represents Meta's third-generation custom silicon initiative developed with Broadcom and TSMC.

The Arke chip is designed to reduce energy consumption and infrastructure expenses while decreasing the company's dependence on Nvidia hardware for executing AI computational workloads.

Meta has allocated resources exceeding one gigawatt's worth of proprietary chips over a 12-month timeframe, which would be financially impractical if relying on commercially available alternatives. Song emphasized that operating at such magnitude, absorbing a 30% cost elevation proves financially impractical, underscoring the significance of the internally developed solution.

The company is also working on a subsequent fourth-generation chip, named Astrid, which will complete its design phase within approximately 30 days and be deployed in late 2027.