OWC NAS for local AI
Updated
Is OWC good for local AI?
No, not on what OWC publishes. The Jellyfish Mobile, Nomad and Tower are video storage appliances managed through Jellyfish Manager, and OWC documents no Docker, container, GPU or LLM support, so all three rate as Traditional NAS. That is absence of documentation rather than a stated no, and the Mobile is discontinued, replaced by the Jellyfish Studio.
Which OWC systems do we track?
3 systems, 0 with a verified UK price. Sorted by AINAS Score.
| System | AI class | Score | Max RAM | Bays | Network | GPU expansion | Price |
|---|---|---|---|---|---|---|---|
| Jellyfish Mobile | Traditional NASDiscontinued | 42 | 256GB | 8 | 2 x 10GbE | None | No verified price |
| Jellyfish Nomad | Traditional NAS | 34 | 128GB | 0 | 6 x 10GbE | None | No verified price |
| Jellyfish Tower | Traditional NAS | 28 | 512GB | 0 | 1GbE | None | No verified price |
What AI features do OWC systems support?
Local LLM
0 of 3
0 documented by OWC.
Ollama
0 of 3
0 documented by OWC.
Open WebUI
0 of 3
0 documented by OWC.
GPU expansion
0 of 3
0 documented by OWC.
AI photos
0 of 3
0 documented by OWC.
Surveillance AI
0 of 3
0 documented by OWC.
Classes in this range: 3 Traditional NAS.
What does OWC do for local AI?
Across the Jellyfish line, OWC offers no local AI route that we can find. No NPU or TOPS figure appears for the Nomad, Tower or Mobile, and no GPU path, integrated or expansion, or accelerator is documented on any of them. That leaves every model rated Traditional NAS, and the deciding factor is software rather than silicon: OWC documents no Docker or container support, so there is no sanctioned way to host a model. This is absence of documentation, not a stated refusal.
Memory is the one figure that looks tempting, and the Jellyfish Tower sets the ceiling at 512GB of DDR4, with 256GB fitted, on dual Xeon Scalable processors. The Jellyfish Mobile reaches 256GB from 128GB fitted, and the Jellyfish Nomad is listed at 128GB, with ECC DDR4 reported. All of that serves shared editing storage, and with nowhere to run models it brings no AI benefit. Sourcing for these numbers is weak, so confirm them with OWC before relying on them.
Which OWC range fits which buyer?
Jellyfish Nomad is for an editing team that needs a compact all-flash unit with swappable storage. OWC's September 2023 release gives six direct attached 10GbE ports and two user-swappable NVMe shuttles, in 16TB, 32TB and 64TB versions. No Thunderbolt port is described. A dual 25GbE card was previewed at IBC 2026, but we do not count it in the Nomad's networking.
Jellyfish Tower is the choice for a post-production group whose footage needs room, with versions from 120TB to 480TB. OWC quotes capacity and gives no bay count. Network details vary by source and order: a comparison page lists two 1GbE ports and four optional cards, while a product page says up to 16 onboard ports at 1GbE or 10GbE, so settle the port layout with OWC when ordering.
Jellyfish Mobile is discontinued, so it suits only a team already running one and planning around it. OWC's September 2023 release says the Jellyfish Studio replaces it, and we found no separate end of life notice. It came in 48TB and 96TB versions with 128GB of memory, and we record it as an eight-bay unit. Its manual shows two built-in 10GbE ports, and we record no Thunderbolt port.
Where OWC falls short
For AI, every capability we check is a blank: OWC documents no support for local LLMs, RAG, Ollama or Open WebUI, and there is no AI photo, search, video or surveillance feature. These are closed appliances, managed through Jellyfish Manager, and that closed design is the real barrier to adding any of it yourself.
Disclosure is thin as well. OWC says the range uses Intel Xeon Scalable processors, but it publishes neither the exact models nor the core counts, which makes it hard to judge CPU performance even in theory. It does not name the operating system beneath Jellyfish Manager, and virtual machine support is equally undocumented. We found no warranty term for any of the three on the pages we saw, so raise that with OWC too. Fairly judged, these gaps matter far more to an AI buyer than to a video team, since the range is built around shared editing storage.
What OWC documents officially
OWC's public material covers video storage and is silent on local AI: the Which Jellyfish comparison page compares the models on memory and ports, the Jellyfish v3 introduction on OWC's blog is our source for the Tower's memory ceiling, the September 2023 press release on the all-SSD Nomad and Studio announces the Nomad, the guide to accessing Jellyfish Manager explains the management interface, and the Jellyfish Mobile support manual covers that unit's hardware. Several searches of OWC's site turned up no Docker, container, GPU or language model material, and OWC publishes no statement of what any Jellyfish can or cannot run locally.
How reliable are these OWC details?
Treat them with care. They come from short excerpts of OWC's own pages, manuals and press releases rather than a single complete specification, which is why the discrepancies above remain open. Ask OWC to confirm anything an order depends on.
What should you buy instead for AI?
Anyone wanting video storage and local AI in one box should look at a Thunderbolt NAS such as QNAP's TVS-h874T, where QNAP publishes Ollama and Open WebUI guides.
Why do OWC's scores look higher than its verdicts?
The stored local LLM scores for the OWC Jellyfish Mobile, OWC Jellyfish Tower and OWC Jellyfish Nomad are 46, 42 and 41, higher than several AI Capable units, because the scores weigh memory and processor class on paper. Every AI capability behind them is no and Unverified, so each gets a poor verdict for Local LLM, Private RAG and Creative AI, and the model-fit section does not apply to any of them, however much memory they hold. Read those numbers as a description of hardware no model can reach, not of a working model host. The Mobile is also discontinued, so its figures matter mainly to teams that already own one.
OWC comparisons
Guides for OWC owners
See all AI NAS guides.