Straight answers to the important questions.
Energy, data, models and operations: here is what you need to know before bringing AI into your own site.
No black boxes.
If something is still unclear, talk to us directly.
01What is an off-grid AI data center?
It is a local AI compute system powered by solar generation and battery storage. It can operate without relying on the electrical grid.
02Does data leave the site?
No. Documents, prompts and results remain on site. Connectivity is for technical management, not sensitive-data transfer.
03Which models do you use, and under what license?
We select fit-for-purpose open-source models, including DeepSeek, GLM, Qwen, Mistral and gpt-oss, with MIT or Apache 2.0 licenses where stated. The license is checked for every installed release.
04How much power does it need?
It depends on size: from around 0.3 kW continuously for Edge to around 170 kW for D6. The study assesses the site’s load and solar resource.
05Does it work without the internet?
Yes. Inference and documents work over the local network; an air-gapped configuration is available. Updates can be delivered on encrypted media.
06What happens at night or under cloud cover?
Battery storage powers the system. Barrel OS progressively lowers the load and can start a generator or use the grid as backup if configured.
07Is it GDPR and AI Act compliant?
Local architecture makes data governance and audit trails easier. Compliance for each deployment still requires a dedicated assessment, configuration and documentation.
08How much does it cost?
It depends on capacity, autonomy, models, installation and service level. We start with a feasibility study, deducted from the order.
09Who provides support?
Barrel provides testing, commissioning, monitoring, updates and on-site response under the service agreement.
10Can the grid serve as a backup?
Yes. The system can be designed off-grid or with the electrical grid as a backup source.