Semiconductor giant AMD held its developer event "AMD AI Developer Day 2026" in Shanghai, China, on May 19, 2026[1][3]. It was the first AI Developer Day AMD has held outside the United States, with Chair and CEO Dr. Lisa Su delivering the keynote[2][3]. At the venue, mini PC brand MINISFORUM exhibited a NAS and a mini workstation built for AI agents, and the company says its flagship "MS-S1 MAX" earned Dr. Su's signature[1].
AMD's First Overseas Developer Day, Lisa Su Forecasts AI Adoption
AMD AI Developer Day 2026 was the first AI developer event the company has held outside the United States, drawing many Chinese AI engineers and developers[3]. In her keynote, Dr. Lisa Su projected that the number of daily AI users would reach around 5 billion within the next few years, and framed 2026 as a turning point from generative AI toward AGI (artificial general intelligence)[1][2].
MINISFORUM took part in the event as one of AMD's key partners, displaying a NAS and a mini workstation for AI agents[1]. Both are powered by AMD's "Ryzen AI Max+ 395" and are designed to run AI functions at the edge (on the device side) so that multiple agents can handle corporate tasks[1]. MINISFORUM says the mini workstation "MS-S1 MAX," recognized for its performance on AI workloads, received Dr. Su's signature at the venue[1].
AI Agent NAS "N5 MAX" and Mini Workstation "MS-S1 MAX"
Of the two models on display, the AI agent NAS "N5 MAX" features 64GB of quad-channel memory and two 10GbE ports, and can be expanded with up to five HDDs and five SSDs[1]. According to MINISFORUM, when running Alibaba's language model "Qwen3.6-35B," generation speed peaks at over 70 tokens per second, and even with four users running concurrently it maintains over 20 tokens per second per user[1]. On the software side, it runs the company's own "MinisCloud OS," which works with local large language models and the AI agent development tool "OpenClaw" to perform tasks such as semantic image search and automated document and code processing within an in-house environment[1].
The other model, the mini workstation "MS-S1 MAX," can form a combined 512GB high-bandwidth unified memory pool by clustering four units together[1]. MINISFORUM says that when this four-node configuration runs the 671-billion-parameter "DeepSeek" model (4-bit quantized version), it cut volume by more than 50%, power consumption by more than 80%, and total cost by more than 77% compared with a 5U rack-sized GPU server equipped with the NVIDIA RTX 5090[1].
Proposing Full-Scenario Use Across Enterprise, Education and More
MINISFORUM also outlined plans to apply this AI agent infrastructure to practical work across industries[1]. For enterprises, it points to a use case in which a network of cooperating agents automates business workflows and deploys AI across departments[1]. In education, it introduced an example combining a 12-node MS-S1 MAX cluster with six compact AI agent units, allowing universities to deliver AI education using large language models in a local environment[1].
Founded in 2018, MINISFORUM is an edge-computing brand whose lineup includes Agent NAS, AI mini workstations, AI mini PCs, and the "AtomMan" gaming mini PCs[1]. The company says it has more than 4 million users across over 100 countries and regions[1].
Summary
With this announcement, MINISFORUM put forward AI agent hardware at AMD's first AI Developer Day held outside the United States. It reflects the trend toward "local AI" — running large language models on devices at hand rather than relying on the cloud. That said, the performance figures and reduction rates are all based on MINISFORUM's own announcement, and it is worth noting that no third-party verification has been published.
出典:[3] https://www.digitimes.com/news/a20260519PD227/amd-shanghai-ceo-lisa-su-development.html
