Contabo Launches GPU VPS with NVIDIA RTX PRO 6000 and 96 GB VRAM

Contabo has launched a new GPU VPS offering aimed at AI workloads, scientific computing, and GPU-accelerated creative applications. The service combines a dedicated NVIDIA GPU with the company's existing VPS infrastructure and monthly pricing model.

The new GPU VPS is built around an NVIDIA RTX PRO 6000 with 96 GB of GDDR7 VRAM, based on NVIDIA's Blackwell architecture. According to Contabo, the configuration is designed to provide dedicated GPU capacity without requiring customers to commit to an entire physical server.

Hardware configuration

The GPU VPS includes:

• NVIDIA RTX PRO 6000 — 96 GB GDDR7 VRAM
• 18 vCPUs
• 96 GB system RAM
• 900 GB NVMe storage
• Ubuntu 24.04
• CUDA pre-installed
• Unlimited traffic, subject to a fair-use policy
• Access to VPS add-ons and Contabo's 1-Click App library

The service is currently available in Europe and US Central.

Pricing

Contabo lists the GPU VPS at €999 / $1,199 / £969 per month.

Longer commitments receive additional discounts:

• 12-month term: 15% discount
• 24-month term: 20% discount

This puts the service into a different pricing category from conventional CPU VPS products, but potentially below the cost of acquiring and maintaining comparable high-end GPU hardware outright.

Designed for AI, research and rendering

Contabo positions the GPU VPS primarily around three use cases.

For AI and machine learning, the 96 GB of VRAM can accommodate larger models on a single GPU and can be used for inference, fine-tuning and LoRA training.

For scientific and engineering workloads, the company highlights computational fluid dynamics, molecular dynamics, finite element analysis and data science applications where GPU acceleration can significantly change processing times.

The third target area is video, 3D and visual effects, where a dedicated GPU can be used for rendering, motion design and compositing without requiring users to rely entirely on their local workstation.

Why 96 GB of VRAM matters

The most notable specification is the GPU's 96 GB of VRAM. Large VRAM capacity can be particularly important for AI workloads because models and associated data must fit within available GPU memory.

Contabo specifically highlights the ability to run a 70-billion-parameter model on a single GPU rather than distributing it across multiple instances. Actual requirements, however, depend on model architecture, quantization, batch size and workload configuration.

A different approach to GPU infrastructure

The launch reflects a broader shift toward making dedicated GPU resources available through monthly infrastructure products rather than exclusively through hyperscale cloud platforms.

For customers running predictable workloads, fixed monthly pricing can provide a simpler cost structure than usage-based GPU billing. At the same time, the €999/month starting price means the service is primarily aimed at professional, research and business workloads rather than occasional GPU experimentation.

Contabo has also published documentation for the new product as part of its open-source documentation project, with the GPU VPS documentation available alongside its other infrastructure documentation.

DohoHub will track the new GPU VPS alongside other GPU, VPS and cloud infrastructure offerings, including pricing, specifications and availability as the market evolves.

Source: Contabo