Zettlab Is a NAS That Understands What's Inside Your Files

Most cloud storage services know your files exist. Zettlab knows what's inside them.

Type "sunset at the beach from last summer" and the device finds the exact photo — not by filename, but by understanding the image itself. Ask it to summarize a contract. Search a video by what someone said in it. All of this happens on a box sitting on your desk, with no internet connection, no monthly fee, and nothing leaving your home.

That's the core pitch of Zettlab — the first NAS (Network Attached Storage) device built around a dedicated Neural Processing Unit for running AI entirely on-device. It raised $1.48 million from 1,820 backers on Kickstarter, won the CES 2026 Best of Innovation Award in Computer Peripherals & Accessories, and won a Red Dot Design Award in 2025. It's a meaningful arrival in a market where intelligence usually means uploading your data to someone else's server.

Why NAS Has Always Been Smart at Storage But Dumb About Content

Traditional NAS devices are good at one thing: holding files. They organize by folder and filename, which means finding something means already knowing what it's called and where you put it. That works until you have years of unorganized photos, recordings, and documents — at which point search becomes a guessing game.

Cloud services solved the search problem by analyzing your files on their own servers. Google Photos recognizes faces. Dropbox indexes documents. But doing that means your files pass through third-party infrastructure, contribute to training datasets, and stay accessible to platforms with evolving privacy policies. For personal files, client NDAs, creative work under embargo, or simply people who'd rather their data stayed theirs — that trade-off stopped being acceptable.

Zettlab built the AI layer directly into the hardware to close that gap without the cloud.

How ZettAI Actually Works

Every Zettlab device ships with a dedicated NPU that handles AI inference locally. The base models — D4 (four bays) and D6 (six bays) — run on a Rockchip RK3588 ARM processor with a 6 TOPS NPU and 16GB LPDDR4x RAM. The Ultra models — D6 Ultra and D8 Ultra — step up to an Intel Core Ultra 5 125H with a 34 TOPS NPU, 32GB DDR5 RAM expandable to 96GB, and dual 10GbE networking ports.

The AI layer, called ZettAI, indexes every file as it arrives. Photos get tagged by subject, location, and scene — automatically, locally, within seconds of upload. Audio and video files get transcribed with speaker identification and time-linked navigation. Documents get analyzed for content so natural language queries can retrieve them by meaning, not just filename.

A dedicated AI Chat interface lets users type plain commands — "show me all photos from Italy," "summarize this PDF," "find every recording where someone mentions the budget" — and the NPU handles the inference on-device. An AI Knowledge QA feature lets users ask questions across their entire library and get answers with cited sources. Nothing goes anywhere. The NPU processes everything within the device.

NASCompares, one of the most trusted voices in the NAS community, described the result directly: "The fundamental difference in AI for Zettlab compared to other NAS devices is its deep integration of local big models and natural language processing technology, enabling smarter and more efficient data management and interaction."

What Four Models Actually Offer

Zettlab D4 starts at $429.99 — a four-bay unit for home users, photographers, and small media libraries. Quiet, compact, and capable of up to 100TB raw storage with the 6 TOPS NPU handling everyday AI tasks. SD card and microSD slots on the front make direct camera backup straightforward.

Zettlab D6 adds two more bays on the same ARM platform — better for growing media libraries where the extra capacity matters more than raw processing power.

Zettlab D6 Ultra is where the platform changes character. The Intel Core Ultra 5 125H brings 34 TOPS of AI processing, dual 10GbE networking for 4K and 8K editing directly on the NAS, dual USB4 ports, and M.2 PCIe 4.0 slots for cache acceleration. For a small creative team sharing large project files, this is the central hub those workflows need.

Zettlab D8 Ultra adds two more bays and scales capacity up to 200TB. Same Intel Core Ultra platform. Same 34 TOPS NPU. Premium push/pull drive trays make hot environments and dense setups more manageable.

All four run ZettOS, built on Debian Linux, with Docker support, container orchestration, a native app store, and an external GPU expansion port for users who want to push AI processing further than the onboard NPU allows.

What Reviewers Found When They Actually Used It

Neowin's January 2026 review of the D4 called the hardware "highly recommended" and noted the AI search functionality worked well in real-world use. They flagged that ZettOS is still in alpha — missing features common on more established platforms like Synology or QNAP — while describing the build quality as solid and the drive installation system as genuinely easier than competitors.

TechRadar's February 2026 review of the D6 Ultra was more measured. Their reviewer described ZettOS as "reasonably sophisticated for an early-stage platform" but raised questions about whether the on-device AI processing is fast enough for very large datasets on this class of hardware. Compared with the NPU TOPS figures in Nvidia's high-end GPUs, the 34 TOPS available here is still modest — enough for running local AI models, but not without limits.

The honest picture is a device with genuinely working AI capabilities, hardware that holds up under scrutiny, and software that's still catching up to the ambition. For early adopters who care about privacy and can tolerate an OS that's still maturing, reviewers found the trade-off worthwhile. For users expecting Synology-level software polish on day one, it's worth waiting.