AI server: keep your data in your own hands
Run AI agents, automations, open-source language models and vector databases on your own server. High-RAM, All-Flash NVMe cloud servers in Istanbul process your customer data without sending it to third-party services.
- 32 GB – 128 GB RAM options
- All-Flash NVMe storage
- Ollama, n8n, Open WebUI, Qdrant
- Data in Turkey, Istanbul Tier III
- Fixed monthly price, no token fees
- Free DDoS protection
- Istanbul Tier IIIMARS data center
- Free DDoS protectionLayer 3-4, 2.8 Tb/s
- Weekly backupsFree, automatic
- 24/7 supportPhone · ticket · live chat
Recommended plans for AI workloads
Language models are memory hungry. RAM needs grow with model size; we recommend 16 GB for small models and 32 GB or more for medium models.
- CPU2 vCPU
- RAM8 GB DDR4
- Disk150 GB NVMe SSD
- Setupn8n pre-installed
- CPU4 cores · Xeon Platinum 8173M
- RAM16 GB DDR4
- Disk200 GB All-Flash NVMe
- Operating systemWindows Server or Linux
- CPU10 cores · Xeon Platinum 8173M
- RAM32 GB DDR4
- Disk500 GB All-Flash NVMe
- Operating systemWindows Server or Linux
- CPU16 cores · Xeon Platinum 8173M
- RAM64 GB DDR4
- Disk1000 GB All-Flash NVMe
- Operating systemWindows Server or Linux
*The large figure is the monthly amount when billed yearly (12 months); below it are the yearly total and the monthly billing price. Prices in Turkish lira (₺) are updated automatically from our order system; monthly, 3, 6 and 12-month billing is available. Our servers have no graphics card (GPU); models run on the CPU and RAM.
What is an AI server for?
Using services such as ChatGPT, Gemini or Claude through an API gets you started quickly, but you pay for every request and the data you send is processed on the provider’s infrastructure. On your own AI server you run open-source models, automations and your knowledge base under your own control.
How to build AI infrastructure at Sunucun
- Automation and agents: on our n8n server, automate email, CRM, WhatsApp and Telegram processes with AI Agent nodes, using either external APIs or a local model.
- Local language model: run Llama, Qwen, Gemma, Mistral or DeepSeek models on your own server with Ollama.
- Company knowledge base (RAG): load your documents into a vector database such as Qdrant, pgvector or Chroma so the model answers from your own documents.
- Chat interface: give your team an internal, ChatGPT-like interface with Open WebUI.
Does AI work without a GPU?
Yes, within limits. Quantized models with 3B–8B parameters answer at a usable speed on strong processors with enough RAM, which is often enough for agents, classification, summarization and internal assistants. Very large models and many simultaneous users need GPU infrastructure; in that case we recommend a hybrid setup that combines an external API with your local server.
How much RAM does each model need?
Approximate starting points for 4-bit quantized models; longer context and more simultaneous requests increase the requirement.
| Model size | Example models | Recommended RAM | Recommended plan |
|---|---|---|---|
| 1B–4B | Llama 3.2 3B, Gemma 3 4B, small Qwen versions | 8–16 GB | CLD-7 / CLD-9 |
| 7B–8B | Llama 3.1 8B, Qwen 7B, Mistral 7B | 16–24 GB | CLD-9 / CLD-10 |
| 12B–14B | Gemma 3 12B, Qwen 14B, DeepSeek-R1 14B (distill) | 32 GB | CLD-12 |
| 27B–32B | Gemma 3 27B, Qwen 32B | 64 GB and up | CLD-15 and up |
Large models respond slowly on CPU; we recommend testing with your own scenario before production use.
What can you do with an AI server?
Internal assistant
A chat assistant for your team with Open WebUI, where data never leaves your server.
Documents and knowledge base
A RAG system that answers from contracts, procedures and product documents.
AI agents
n8n agents that classify email, open CRM records and draft replies to messages.
Customer service
First-response and routing bots for WhatsApp and web chat.
Text processing
Run summarization, translation, tagging and data extraction in bulk.
Development and testing
Try models in your own environment before adding AI features to your application.
Set up your AI server in four steps
- Choose a planPick a cloud plan with enough RAM for the model size you will use.
- Install LinuxUbuntu LTS is recommended; server details arrive in your panel.
- Install the toolsInstall Ollama, Open WebUI, n8n and a vector database with Docker.
- Secure itProtect interfaces with passwords and HTTPS and close unused ports.
Common questions
What is an AI server?
A server with enough CPU, RAM and fast disk to run AI workloads such as language models, AI agents, automations and vector databases.
Do your servers have GPUs?
No. Our servers have no graphics card (GPU); models run on CPU and RAM. That is enough for small and medium quantized models.
Which models can I run?
With Ollama you can run open models such as Llama, Qwen, Gemma, Mistral, Phi and the distilled versions of DeepSeek. Check each model’s license before use.
Why my own server instead of the ChatGPT API?
On your own server data is not sent to a third-party service, you pay no per-request fees and you choose the model. If you need very large models, you can combine an API with your local server.
Can I build an AI agent with n8n?
Yes. Our n8n server comes pre-installed; its AI Agent nodes can use external APIs or an Ollama server on the same network.
Do you help with setup?
We install the server and operating system. For application setup we guide you and can provide consulting when needed.
More server solutions
Take control of your AI infrastructure
Tell us which model and scenario you plan to use and we will suggest a plan. Our team is with you 24/7 by phone, live chat and WhatsApp.
