At Computex, Synology is showcasing the next generation of DiskStation Manager, which is set to evolve from a storage operating system into a platform for private AI and enterprise management. Ambitions are high, but most features will only be introduced gradually, and true AI power requires specific Synology GPU hardware.
DSM celebrates its twentieth anniversary this year. According to Synology, more than fourteen million systems worldwide run on the operating system, together accounting for more than 400 exabytes of data. With the next generation, the company no longer wants to build DSM solely for human users, but also for AI agents. Two pillars are central: AI ready and enterprise ready.
For AI processing, Synology offers three methods: a public cloud model, an OpenAI-compatible API, or its own locally running LLM on GPU-equipped NAS hardware. The latter option never leaves the device and is aimed at sensitive data. You cannot choose this local LLM yourself; it is provided by Synology and must run on their specific AI hardware.
The Office Suite will feature an AI assistant, semantic search, and embedding of unstructured data in 91 languages. However, the translation function remains dependent on the chosen model and currently supports a more limited list of languages.
DSM Agent takes over IT tasks
The most important announcement at Computex 2026 is DSM Agent. Version 1.0 will be available in early June as an ‘on-demand AI consultant’ that provides advice on troubleshooting and configuration via natural language, with awareness of the screen the administrator is working on.
With version 2.0, the agent performs tasks itself: investigating and deactivating a suspicious account, or automatically generating and sending a storage report every month. The agent is not only aimed at IT administrators but also at end users, for example, for compiling a quarterly report.
Synology is building its own model on an open-source foundation that it fine-tunes further, and says it uses AI to keep up with model development. Integration with existing AI environments takes place via MCP and CLI, where access, according to Synology, remains shielded by the DSM permissions set by the administrator. How strict that governance is in practice remains to be seen. That the company takes its AI foundations seriously was already evident during the launch of DSM 7.3.
Cluster Manager and management at scale
On the enterprise side, Synology introduces Cluster Manager, which relies on a new containerized storage architecture. Workloads are thus decoupled from the underlying hardware, simplifying migration between devices and Quality of Service settings.
Meanwhile, Active Insight is evolving from a monitoring tool into a platform for centralized management, allowing administrators to deploy and configure multiple NAS devices in batches.
Security, compliance, and the fine print
The updated Log Center centralizes logs and exports them to observability platforms such as Datadog, Elastic, and Grafana. Furthermore, there will be a hardware-based secure element as a root of trust, granular role-based access control, and a FIPS 140-3 certification that is still ‘in progress’. Synology remains largely vague on concrete version numbers and release dates: the roadmap will roll out ‘gradually’. It also remained unclear which version will bring data deduplication and whether the PAS series runs entirely on DSM.
Those who want to run heavier AI workloads will also need new GPU-NAS, ranging from the RackStation 26 series to an AI Station for models exceeding 100 billion parameters.
With this, Synology, which recently celebrated its 25th anniversary, is taking an ambitious step toward enterprise and AI. The direction is clear; the concrete implementation and timing much less so.