Arm unveils AGI CPU chip optimized for agentic AI, Meta as launch partner

Arm, the semiconductor company whose CPU architecture powers nearly every smartphone in the world, made a significant push into the data center market on March 28, 2026, unveiling a new CPU chip specifically optimized for agent AI and artificial general intelligence workloads.

The chip, aimed at large-scale AI data centers, is designed to handle the specific demands of modern AI agent systems. While traditional GPUs are optimized for parallel matrix multiplication during training, Arm's new chip is built for the continuous inference and orchestration tasks that agentic AI requires in production.

Meta is confirmed as Arm's first strategic launch partner for the chip. This is not a coincidence. Meta has been under significant pressure in recent months to reduce its dependency on Nvidia, which dominates the AI accelerator market with over 80 percent market share.

The launch comes at a moment when agentic AI is becoming the dominant paradigm across the industry. Companies like Microsoft, Google, and Anthropic are all building infrastructure for AI agents that operate continuously, make autonomous decisions, and coordinate with other agents. Such systems place fundamentally different demands on underlying hardware than the large training runs that have historically defined the market.

For CIOs planning infrastructure for agent-based workflows, this is an important signal: the competitive landscape in AI hardware is shifting, and Arm may become a credible alternative to Nvidia's dominance in the data center.