Meta has announced the deployment of its newest custom AI silicon across its global data centers, aiming to significantly reduce operational costs and energy overhead. The move is part of Meta's long-running effort to reduce its dependence on external AI chip suppliers as it scales large language model training and inference.
In a separate development, Meta launched an official Model Context Protocol (MCP) server for WhatsApp Business, allowing AI agents to autonomously set up, configure and manage business messaging pipelines. The MCP server is a significant step toward making Meta's platforms programmable by AI agents, enabling businesses to integrate automated customer communication workflows.
Meta CEO Mark Zuckerberg also openly criticized calls from rival labs for an AI development slowdown, arguing that companies must continue advancing core capability alongside safety engineering. The company's Superintelligence Labs, led by Alexandr Wang, recently reported its first internal breakthrough models, codenamed Project Avocado for text and Project Mango for visual intelligence, with a public release targeted for the first half of next year.
