AMD shows how to run AI agents on a PC with 128 GB of memory

AMD has published a practical guide to running the open-source AI agent OpenClaw on consumer Windows PCs, aimed at configurations with 128 GB of RAM. The guide describes two ready-made AMD hardware builds, RyzenClaw and RadeonClaw, on which users can deploy the agent locally without relying on cloud infrastructure.

What are the RyzenClaw and RadeonClaw configurations?

judging by the names, both configurations are built around AMD's current processor and graphics card lines, Ryzen and Radeon respectively. They offer two alternative routes to running an agent locally: one centered on CPU compute and the platform's unified memory, the other on a discrete graphics card as the main accelerator for model inference. Offering two builds lets AMD make the case that running modern AI agents does not necessarily require dedicated server infrastructure with Nvidia accelerators. The job can also be done on top-end consumer hardware assembled from mass-market components.

The fact that the manufacturer recommends this much RAM for full-fledged agent workloads reflects a broader trend: local AI agents are becoming a hardware requirement of a new tier, closer in load to professional workstations than to an ordinary home PC. For buyers, it is one more factor when choosing a computer. Alongside gaming performance or the amount of video memory, the question of how much memory you need to run a full AI agent locally, rather than a cut-down demo version, comes up more and more often.

• Tool: OpenClaw, an open-source AI agent.
• Operating system: Windows.
• AMD configurations: RyzenClaw and RadeonClaw.
• Memory target: around 128 GB of RAM.
• Publication format: official AMD guide.

Why does an AI agent need so much memory?

Running an AI agent locally differs from running a simple local chatbot. Besides the language model itself, the agent keeps in memory its reasoning history, the results of tool calls, intermediate files and often several parallel task chains at once. The larger the local model and the longer the context window it has to process, the more RAM is needed to hold it all without constantly hitting the disk, which slows things down sharply. The 128 GB target suggests that AMD is aiming at users who want to run fairly large models locally, not lightweight versions trimmed down for minimal configurations.

What does this mean for local AI agents on PCs?

A guide like this from a major CPU and GPU maker is part of a broader move toward local, private AI agents that run on the user's own hardware without sending data to the cloud. For enthusiasts and small teams, this cuts both the cost of subscriptions and API calls and the risks of sending sensitive data to third-party servers. It also benefits AMD, which is positioning its consumer Ryzen and Radeon platforms as a real alternative to the Nvidia-plus-cloud combination for people who want to keep AI agents at home or in the office on their own equipment.

Beyond saving on subscriptions, running agents locally removes the latency question. A cloud API call always adds network delay and depends on a stable internet connection, whereas a locally deployed agent responds as fast as the user's own hardware allows. For developers and enthusiasts, it also means they can experiment with model settings and agent parameters without the restrictions cloud providers usually impose.

It is also telling that the guide comes from a component maker rather than the agent's developer. For AMD, it is a way to show that its processors and graphics cards can handle not only games but also full agentic AI workloads, which until recently were associated almost exclusively with professional accelerators and cloud services. That stance matters all the more amid growing user interest in running AI tools offline, without a subscription and without the risk that a cloud service will one day become paid, change its terms of use or restrict access to some features.