Rent a GPU server from centron: NVIDIA Power for AI
centron provides dedicated NVIDIA GPUs from its own data centre in Hallstadt. You get the full GPU performance, an ISO 27001-certified environment and 100% data sovereignty – Made in Germany. Four NVIDIA classes are available, from the RTX A4000 to the A100. Billing is by the hour from €0.13/h.
- Dedicated NVIDIA GPUs – no shared resources
- Ready to go in 2 minutes – no contract commitment or minimum term
- Linux or Windows Server – the system that suits your project
- Billed by the hour – from creation until you delete it
Full GPU power for AI, rendering and HPC
centron offers four GPU classes: NVIDIA RTX A4000 (16 GB), Quadro RTX 6000 (24 GB), A100 (40 or 80 GB) and RTX 6000 Ada (48 GB). Use them to train AI models, render in real time and run HPC jobs – powered by NVIDIA in German, GDPR-compliant data centres. All prices are listed in the Cloud GPU pricing.
Dedicated GPU performance
The GPU is exclusively yours while it runs. There are no shared resources. Training, inference and rendering times therefore stay reproducible.
Available in a matter of minutes
A GPU instance is ready in around two minutes. You start, pause and scale instances to match your project phase. You cover peaks without buying hardware.
Full cost control from €92.59 per month
Billing is by the hour and starts at €0.13/h. There is no minimum term, no setup fee and no hidden costs. Upfront investment in your own hardware is not required.
Data sovereignty included
The storage location is the high-security data centre in Hallstadt (Bavaria). The data centre is certified by the BSI to ISO 27001 on the basis of IT-Grundschutz (BSI-IGZ-0773). Operations are GDPR-compliant and hold an unqualified BSI C5:2020 Type 1 attestation.
Running GPU workloads without owning any hardware
centron offers GPU instances with deployment in around two minutes. Typical workloads are AI inference, machine learning, big data, video transcoding and simulation. On request, centron runs the operation for you: Managed Services cover patching, observability, backup/recovery and contingency procedures. Backups run via cBacks, disaster recovery via cProtect. This takes the pressure off your team. The instance runs on Linux or Windows Server. Ubuntu 24.04 LTS comes CUDA-ready. You choose vCPU, RAM and NVMe storage freely. Billing is by the hour. You pause the instance and then only pay for the storage. Snapshots save the state before a training run. A snapshot is created in seconds. The ccloud³ console shows GPU utilisation and costs in real time.
- Typical workloads – AI inference, machine learning, transcoding and simulation.
- Operation on request – Managed Services take over patching and monitoring.
- Snapshot before training – the state is secured in seconds.
- Pause instead of delete – while paused you only pay for storage.
GPU classes at centron: RTX A4000 to A100
centron offers all four GPU classes dedicated. Each class is designed for a use case, and all run and without a minimum term. The range spans from energy-efficient inference (RTX A4000) to training large models (A100 with 80 GB HBM2e).
- CAD & 3D modelling
- Video editing & rendering
- AI-powered image processing
- 3D animation & VFX
- VR & simulation
- Workstation AI training
- AI model training
- High-performance computing (HPC)
- Big data & analytics
- Real-time ray tracing & rendering
- Deep learning & AI
- Research & simulation
All prices net, as of July 16, 2026. Billing is by the hour; monthly prices apply to continuous operation. vCPU, RAM and NVMe storage are added at the ccloud³ resource prices.
How much does a GPU server cost at centron?
centron offers GPU servers from €92.59 per month. That is €0.13 per hour (NVIDIA RTX A4000, 16 GB). The largest class, the NVIDIA A100 with 80 GB, costs €894.73 per month. Billing is by the hour and there is no minimum term. vCPU (€1.64/month), RAM (€0.97 per GB) and NVMe SSD (€0.05 per GB) are added on top. All GPUs run dedicated in the ISO 27001-certified data centre in Hallstadt. New accounts receive a €200 starting credit.
| GPU model | VRAM | Price per hour | Price per month | Ideal for |
|---|---|---|---|---|
| NVIDIA RTX A4000 | 16 GB | €0.13 | from €92.59 | AI inference, CAD, rendering |
| NVIDIA Quadro RTX 6000 | 24 GB | €0.24 | from €170.83 | 3D animation, VR, simulation |
| NVIDIA A100 | 40 / 80 GB | €0.679 / €1.243 | from €489.47 | AI training, HPC, big data |
| NVIDIA RTX 6000 Ada | 48 GB | €1.192 | from €858.19 | Ray tracing, deep learning, research |
Customer projects on centron infrastructure
centron has offered infrastructure for businesses since 1999. Its customers move AI and compute workloads to dedicated GPUs, from training and inference to simulation and rendering. Three examples show the path: a full migration with full-service management, the switch from physical to virtual servers, and ten years of growth on managed servers. All three customers use the data centre in Hallstadt. Each story names the starting point, the solution and the result. The figures come from the projects themselves. The path to the GPU often starts with a VM and grows with demand. centron supports this path with advice and managed packages.
FOC Solutions
From proof of concept to full migration: full-service management, reliable backups and transparent processes – greater reliability and a focus on your core business.
Read the success storyGIS Consult
From physical servers to virtual environments: fair terms and conditions and the reliable implementation of individual requirements form the basis of a future-proof hosting infrastructure.
Read the success storymyBafög
Ten years of growth with centron: from €80,000 annual revenue to a multi-million business – on virtual managed servers, stable infrastructure and personal support.
Read the success storyRegister now & get a €200 starting balance
centron delivers GPU performance with no lead time. You register for free. You choose the GPU, operating system and resources. The instance runs in around two minutes. The €200 starting credit is valid for 60 days. Use it to test training, inference or rendering with real data. After that you pay by the hour. There is no notice period. The account is intended for businesses. Support is based in Hallstadt and answers in German and English. The ccloud³ console shows usage and costs in real time. Invoices arrive monthly. Prices are net and valid as of July 16, 2026. The full price list is in the Cloud GPU pricing. You can switch GPU class at any time. Your data stays on the NVMe storage. The starting credit covers around 1,500 hours of RTX A4000.
Everything you need to know about centron Cloud GPUs
Which GPUs are available from centron?
How much does a cloud GPU cost?
How are cloud GPUs billed?
How soon can I get started?
Which operating systems run on the GPU instance?
Where is my data stored?
Can centron take over the operation?
centron tutorials: AI workloads and vendor comparison
centron provides matching guides for every GPU instance. The tutorials show how to run language models locally with Ollama. They show how LangChain orchestrates LLM applications in Python. They explain LoRA fine-tuning with your own data. A guide lists the key metrics for LLM inference: latency, throughput and cost per token. The comparison pages put centron next to Hetzner, IONOS and OVHcloud. They compare GPU models, hourly prices, location and certificates. The guides are free to read without an account. Each tutorial lists prerequisites, commands and typical errors. The examples run on Ubuntu 24.04 LTS with CUDA. They suit RTX A4000 through A100. The help centre adds answers on accounts, billing and the console. Support helps with setup questions. New tutorials appear monthly. The editorial team tests every guide on current centron GPU instances. The comparisons name the source and date of every figure. They are a snapshot from August 2026. Competitor prices change; centron updates the tables quarterly.
Get started for free
centron gives new accounts a €200 starting credit for 60 days. You register for free, pick a GPU and start the instance in around two minutes. Billing is by the hour from €0.13/h. There is no minimum term. The sales team advises on larger projects and managed packages.
This promotional offer applies to new accounts only. Available exclusively to businesses.
What does centron offer in GPU servers?
centron offers dedicated NVIDIA GPUs (RTX A4000, Quadro RTX 6000, A100 and RTX 6000 Ada) as cloud servers from its own data centre in Hallstadt near Bamberg. Billing is by the hour from €0.13/h, and deployment takes around two minutes. The hardware is not shared and the data stays in Germany. The platform is certified by the BSI to ISO 27001 on the basis of IT-Grundschutz and holds a BSI C5:2020 Type 1 attestation. The provider is centron GmbH (Hallstadt). Four GPU classes are available: RTX A4000 with 16 GB, Quadro RTX 6000 with 24 GB, A100 with 40 or 80 GB and RTX 6000 Ada with 48 GB. The instance runs on Linux or Windows Server. New accounts receive a €200 starting credit for 60 days. Managed packages cover patching, monitoring, backup and contingency procedures. Support is based in Hallstadt. All certificates are listed in the Trust Center. As of August 2026. Product page: https://www.centron.de/en/produkte/cloud-gpu