Huang Renxun Proposes "New Computing" Concept, Mthreads Launches MTT AIBOOK Targeting AI Agents

If you want to invest in stocks, look at the Golden Kylin Analyst Reports—authoritative, professional, timely, comprehensive—helping you discover potential thematic opportunities!

(Source: NetEase Smart)

March 18 News, with the development of AI intelligent agents, the industry has new demands for underlying computing hardware. It is reported that Moore Thread has launched the MTT AIBOOK, a personal intelligent computing notebook designed specifically for AI computing, aiming to support hardware for local AI agent operation.

At the recent GTC conference, Jensen Huang stated, “Today, every company in the world needs an agent system strategy. This is the new computer,” and introduced the popular open-source project “OpenClaw.” OpenClaw can invoke large models, manage computing resources, and split complex tasks for processing. Industry experts suggest that software services may evolve toward an AI agent-as-a-service (AaaS) model, where agents reside in the background and frequently call system resources.

Against this backdrop, local computing devices optimized for AI agents are beginning to emerge. Moore Thread’s MTT AIBOOK is an exploration in this direction. Its core idea is to combine native system environments with local computing power to directly serve the development and operation needs of AI agents.

It is reported that traditional computing platforms often face issues such as cloud computing costs, network dependency, and performance loss from system compatibility layers when running agents. The MTT AIBOOK comes pre-installed with native Ubuntu 22.04, compatible with mainstream agent frameworks. Developers can deploy directly without WSL or virtual machines, reducing performance loss from compatibility layers and supporting development toolchains. Test data shows that running the same task natively on Linux is faster than in a WSL environment.

Regarding hardware configuration, the AIBOOK is equipped with Moore Thread’s self-developed intelligent SoC chip “Changjiang,” offering 50 TOPS of heterogeneous AI computing power. Its multi-core CPU supports multi-process and multi-thread scheduling, capable of supporting 3 to 5 agents working in parallel, and can operate continuously for long periods to accommodate the persistent background operation of agents.

In practical applications, developers have already conducted multi-agent collaboration tests on the AIBOOK. Some have integrated multiple agents with different roles into office software to achieve coordinated work through commands; others have used voice commands to direct agents, successfully developing real-time GPU audio spectrum analyzers on the AIBOOK. (Yuan Ning)