Skip to content
Compute

GPU Server

NVIDIA GPUs for AI training and inference, with data kept in Türkiye.

Servers with NVIDIA L4, L40S and H100 GPUs. Run large language models on your own infrastructure in line with Turkish data-protection law. Ready CUDA, PyTorch and vLLM images.

New 99.90% SLA Per-second billing
Deploy now Get a quote

Quick facts

What is GPU Server?

Servers with NVIDIA L4, L40S and H100 GPUs. Run large language models on your own infrastructure in line with Turkish data-protection law. Ready CUDA, PyTorch and vLLM images.

NVIDIA GPUs for AI training and inference, with data kept in Türkiye.

GPU
NVIDIA L4 / L40S / H100
Regions
Istanbul-1, Ankara-1
Billing
Hourly, monthly cap
Data residency
Türkiye only
Starting price
US$1,095.77/mo

Pricing

Plans and pricing

All prices are in Turkish lira and fixed for 12 months. Click a price to open the wizard with that plan preselected.

Price display
Commitment
incl. VAT
Plan GPUGPU memoryvCPUMemory Hourly Monthly Action
G-L4 1× NVIDIA L424 GB8 vCPU64 GB US$1.5011/hr US$1,095.77 /mo
G-H100 1× NVIDIA H10080 GB24 vCPU240 GB US$6.1702/hr US$4,504.23 /mo
G-H100×8 8× NVIDIA H100 NVLink640 GB192 vCPU1,9 TB US$46.2666/hr US$33,775 /mo

Calculator

Configure and see the cost instantly

Pick a plan, quantity, region and add-ons to see the monthly total and its hourly equivalent.

Plan
Add-ons

Key features

  • NVIDIA L4, L40S and H100 options
  • Ready CUDA / PyTorch / vLLM images
  • Hourly rental with a monthly cap
  • Data and models stay in Türkiye
  • NVLink clusters up to 8 GPUs

Who is it for?

In-house AI

Run open-source language models on your own infrastructure so customer data never leaves Türkiye.

Training and fine-tuning

Train or fine-tune models on your own data; rent by the hour and switch off when done.

Vision and video

Accelerated inference for quality-control cameras, document OCR and video analytics.

Automation

From the command line

You can do the same from the panel, the REST API or our Terraform provider.

bash
$ cekirdek sunucu olustur --ad llm-01 --plan g-l40s --bolge ist-1 --imaj vllm-cuda12

Location

Available regions

By default your data is stored only in our data centres in Türkiye.

Infrastructure and data centres

Frequently asked questions

Can I run my own language model?

Yes. With the vLLM or Ollama image you can serve open models (Llama, Mistral, Qwen, etc.) behind an OpenAI-compatible endpoint in minutes. Data never leaves Türkiye.

Are there commitment discounts?

12- and 36-month commitments get up to 15% and 30% off respectively; talk to sales for a quote.

Do GPU Server prices include VAT?

Prices are shown including 20% VAT by default; use the switch at the top of the page to see them excluding VAT. Prices are fixed in TRY for 12 months and identical in every region.

Where is my data stored?

By default your data is stored only in our data centres in Türkiye (Istanbul-1, Ankara-1) and is never transferred abroad without your explicit instruction.

Often used together

All products
Free: 12 months of S-1

Cloud Server

NVMe servers ready in 55 seconds, billed per second.

  • Per-second billing with a monthly cap
  • Built-in DDoS protection and firewall
  • Snapshots and automated backups
starting fromUS$6.17/mo · incl. VAT
Free: First 25 GB

Object Storage (S3-compatible)

S3-compatible, unlimited storage kept as three copies in Türkiye.

  • S3 API compatible
  • Immutable (WORM) buckets
  • Versioning and lifecycle
starting fromUS$4.20/mo · incl. VAT
Free: Control plane

Managed Kubernetes

We run the control plane, you run your apps. Clusters ready in 4 minutes.

  • Free standard control plane
  • HA control plane (99.95% SLA)
  • Node autoscaling
starting fromUS$33.72/mo · incl. VAT

Shall we talk before you decide?

Our solution architects listen to your needs and work out the right plan and cost with you. Migration support is free.

Cookie & content preferences

Necessary Always on

Required for the site and features you request (theme, currency, VAT display, price estimate, these choices). Cannot be turned off.

External content

The infrastructure map loads from OpenFreeMap / OpenStreetMap servers, which receive your IP address.

Analytics

Visit statistics. Not used at the moment; would only run with your consent.

Marketing

Advertising and retargeting. Not used at the moment.