Boosting efficiency, automating processes, accelerating innovation – companies, agencies, and developers aiming to stay competitive are increasingly turning to their own AI solutions.
Whether it's training or deploying your own AI models, conducting extensive data analyses, or running advanced AI workloads like image recognition, natural language processing, or simulations – these projects place the highest demands on infrastructure.
Our high-performance GPU servers provide exactly the computing power you need – scalable, reliable, and fast enough to keep up with the speed of your ideas.
349,- €
per month1
plus one time setup fee: 349,- € 1
590,- €
per month1
plus one time setup fee: 590,- € 1
Upon request
NVIDIA Tesla graphics cards are specifically designed for professional use in data centers, delivering massive computational power for AI, machine learning, and scientific computing.
Compared to consumer GPUs, they are optimized for continuous operation, support ECC memory for error correction, and offer high parallel processing capabilities.
In GPU servers, they ensure maximum stability, efficiency, and performance – even when handling complex, compute-intensive tasks.
Our GPU servers with AMD processors deliver powerful multi-core performance combined with high energy efficiency— ideal for compute-intensive AI applications, data processing, or development environments. Even in single-CPU configurations, the modern “Zen” architecture ensures reliable performance, especially for tasks that benefit from multi-threading.
Intel processors stand out for their strong single-core performance, proven stability, and broad compatibility with popular AI frameworks and applications. In single-CPU setups, they deliver consistent performance — especially for data-intensive processes, inference workloads, and applications with high demands on software stability and I/O throughput.
Important projects require well-considered decisions. We’re happy to support you in selecting the optimal server for your needs and provide comprehensive guidance throughout the process.
Our AI servers offer a clear and fair pricing model: all costs — including GPU power consumption — are already included in the fixed price. This ensures optimal cost predictability and protects you from unexpected additional charges. You get high-performance GPU servers based on NVIDIA technology at transparent conditions with full planning reliability.
At Keyweb, service doesn’t end with the delivery of your server — it truly begins there. You can reach a qualified technician at any time — 24/7, every day of the year, including nights, weekends, and holidays. Whether via our ticket system or optionally by phone, you’ll receive fast, direct assistance. Clear Service Level Agreements (SLAs) contractually guarantee a transparent and high-quality support standard.
To effectively protect your servers against the growing number of cyberattacks, we rely on advanced DDoS protection that goes far beyond the standard solutions offered by many hosting providers. Our data centers are equipped with powerful security technology from Voxility — enabling early detection and reliable mitigation of DDoS attacks outside the infrastructure, before they can even reach our servers.
Our GPU servers are located in high-security data centers in Germany — offering multiple advantages: low latency across Europe, maximum availability, and above all, full GDPR compliance. Your sensitive data remains within a strictly regulated legal framework. This means you benefit not only from top-tier technical performance, but also from the highest level of data security and legal transparency.
GPU servers deliver immense computing power — and naturally come with high energy demands. That’s why responsible resource management is essential. Our data centers in Germany are powered entirely by certified green electricity. In addition, we optimize all processes for energy efficiency and long-term sustainability.
For over 25 years, we’ve been developing and operating powerful hosting solutions — reliable, secure, and built to the highest technical standards. This extensive experience is reflected in every one of our products — from infrastructure to support. As seasoned hosting professionals, we know what matters — today and in the future.
With Ollama, you can run state-of-the-art AI models like Llama 3, Mistral, or Gemma directly on your own hardware – completely independent of the cloud.
This is the perfect solution for businesses that prioritize data privacy, speed, and full control over their AI applications.
Thanks to its simple installation and built-in model management, you can easily download, launch, and integrate AI models into your own applications.
Combined with Keyweb’s high-performance GPU servers, you benefit from maximum computing power for training, inference, and real-time applications – whether for chatbots, data analysis, or creative workflows.
GPU servers are essentially servers equipped with high-performance graphics processors (GPUs). These GPUs can be used for a wide range of compute-intensive tasks — from graphic rendering and scientific calculations to artificial intelligence.
AI servers are a specialized type of GPU server specifically optimized for AI applications. This means they are not only equipped with GPUs, but also often include dedicated software, frameworks, and infrastructure designed to efficiently train and run AI models. Additionally, they are tailored to the typical workloads and requirements of AI processes.
If you have specific requirements for an AI server, please don’t hesitate to contact us.
TOPS (Tera Operations Per Second) indicate how many trillion operations an AI server can perform per second. The higher the TOPS value, the faster and more efficiently the server can train or infer complex AI models. Therefore, TOPS are a key metric for evaluating the performance of GPU servers in the field of artificial intelligence.
vRAM (Video RAM) is the memory of a graphics card — similar to a server’s main RAM, but specifically designed for graphics and compute-intensive tasks. In GPU servers, vRAM plays a crucial role: it temporarily stores the data processed during AI computations, machine learning, or complex image and video processing. The more vRAM a GPU has, the larger the models and datasets it can load and handle simultaneously — without performance loss. Sufficient vRAM is therefore essential for the efficient use of AI and deep learning applications.
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
|
Keymachine® Server Brand Hardware
|
|||
|
Processor
|
Intel® Core™ Ultra 9 285K 24 x max 5,7 GHz, Single CPU |
AMD EPYC™ 9254 48 x max 4,15 GHz, Single CPU |
16 to 256 Cores Processor |
|
Processor cores
|
24 x max 5,7 GHz | 48 x max 4,15 GHz | |
|
RAM
|
256 GB DDR5 | 512 GB DDR5 | up to 6 TB DDR5 RAM |
|
Graphics card
|
1x Nvidia RTX 5070 TI 16 GB VRAM |
1x Nvidia Tesla L4 24 GB VRAM |
up to 8 x Nvidia RTX Pro 6000 Blackwell 96 GB VRAM each |
|
Hard disks
|
2 x 1 TB NVMe SSD | 2 x 960 GB NVMe SSD | Hard drives from 256 GB to 125 TB |
|
Alternative hard disks
|
1 x 2 TB NVMe SSD (+16,- €) 1 x 4 TB NVMe SSD (+29,- €) |
2 x 1,92 TB NVMe SSD (+29,- €) 2 x 3,84 TB NVMe SSD (+49,- €) | |
|
Encrypted server management
|
- | KVM with IPMI 2.0 included |
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
| upon request | upon request | upon request | |
| upon request | upon request | upon request | |
|
Integration of own ISO images
|
|||
| upon request | |||
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
|
24/7 Monitoring
|
|||
|
Managed Firewall
|
|||
|
Individual firewall
|
|||
|
Server-side virus protection
|
|||
|
Free reboot service
|
|||
|
Statistics – log file evaluation
|
|||
| optional | |||
|
Voxility DDoS Protection
|
|||
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
| optional with 30% discount | |||
|
Managed Backup Service
|
optional | ||
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
| TÜV-certified server location Germany | |||
|
Network
|
network availability min. 99% | ||
|
Backbone
|
redundant backbone | ||
|
Power supply
|
UPS supported power supply | ||
|
Air conditioning
|
fail-safe cooling infrastructure | ||
|
Fire and burglary protection
|
fire and burglary protection | ||
|
Security
|
24/7 video surveillance | ||
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
|
24/7 - ticket system
|
|||
|
24/7 - telephone availability
|
|||
|
24/7/365 emergency technical support
|
|||
|
Email service
|
|||
|
Live chat
|
|||
|
Extensive FAQ database
|
|||
| Managed Service included, Managed Pro Service optionally bookable | |||
|
Keymachine® 5 year full service guarantee
|
|||
|
Free exchange of defective hardware
|
within max. 60 minutes | ||
|
KI-Machine S
|
KI-Machine M
|
KI-Machine Custom
|
|
|---|---|---|---|
|
Minimum Contract Period
|
1 Month | ||
|
Notice period
|
14 days to the end of the contract | ||
|
Payment method
|
Debit, credit card, transfer, PayPal | ||
|
Billing period
|
monthly, quarterly, semi-annually, annually, 24 months | ||
| up to 20% discount on your dedicated Server | |||
|
Keymachine® Server Brand Hardware
|
|
|---|---|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Processor
|
| KI-Machine S |
Intel® Core™ Ultra 9 285K 24 x max 5,7 GHz, Single CPU |
| KI-Machine M |
AMD EPYC™ 9254 48 x max 4,15 GHz, Single CPU |
| KI-Machine Custom |
16 to 256 Cores Processor |
Processor cores
|
| KI-Machine S | 24 x max 5,7 GHz |
| KI-Machine M | 48 x max 4,15 GHz |
| KI-Machine Custom |
RAM
|
| KI-Machine S | 256 GB DDR5 |
| KI-Machine M | 512 GB DDR5 |
| KI-Machine Custom | up to 6 TB DDR5 RAM |
Graphics card
|
| KI-Machine S |
1x Nvidia RTX 5070 TI 16 GB VRAM |
| KI-Machine M |
1x Nvidia Tesla L4 24 GB VRAM |
| KI-Machine Custom |
up to 8 x Nvidia RTX Pro 6000 Blackwell 96 GB VRAM each |
Hard disks
|
| KI-Machine S | 2 x 1 TB NVMe SSD |
| KI-Machine M | 2 x 960 GB NVMe SSD |
| KI-Machine Custom | Hard drives from 256 GB to 125 TB |
Alternative hard disks
|
| KI-Machine S |
1 x 2 TB NVMe SSD (+16,- €) 1 x 4 TB NVMe SSD (+29,- €) |
| KI-Machine M |
2 x 1,92 TB NVMe SSD (+29,- €) 2 x 3,84 TB NVMe SSD (+49,- €) |
| KI-Machine Custom |
Encrypted server management
|
| KI-Machine S | - |
| KI-Machine M | KVM with IPMI 2.0 included |
| KI-Machine Custom |
|
Connection
|
|
|---|---|
| All products | 1 Gbit Switch Port |
IP addresses
|
| KI-Machine S | 1 IPv4 and 1 IPv6 |
| KI-Machine M | 1 IPv4 and 1 IPv6 |
| KI-Machine Custom | 1 IPv4 and 1 IPv6 |
Extra IP addresses
|
| All products | additional IP addresses if required |
Own name server
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Traffic included
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom | |
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom | |
| KI-Machine S | upon request |
| KI-Machine M | upon request |
| KI-Machine Custom | upon request |
| KI-Machine S | upon request |
| KI-Machine M | upon request |
| KI-Machine Custom | upon request |
Integration of own ISO images
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom | |
| All products | upon request |
|
Access via SSH
|
|
|---|---|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Full administration and root rights
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
|
Installation of any software
|
|
|---|---|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
|
Free access to the domain registration tool
|
|
|---|---|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
|
24/7 Monitoring
|
|
|---|---|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Managed Firewall
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Individual firewall
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Server-side virus protection
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Free reboot service
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Statistics – log file evaluation
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom | |
| All products | optional |
Voxility DDoS Protection
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom | |
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
| All products | optional with 30% discount |
Managed Backup Service
|
|---|---|
| All products | optional |
| All products | TÜV-certified server location Germany |
Network
|
|---|---|
| All products | network availability min. 99% |
Backbone
|
| All products | redundant backbone |
Power supply
|
| All products | UPS supported power supply |
Air conditioning
|
| All products | fail-safe cooling infrastructure |
Fire and burglary protection
|
| All products | fire and burglary protection |
Security
|
| All products | 24/7 video surveillance |
|
24/7 - ticket system
|
|
|---|---|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
24/7 - telephone availability
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
24/7/365 emergency technical support
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Email service
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Live chat
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Extensive FAQ database
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom | |
| All products | Managed Service included, Managed Pro Service optionally bookable |
Keymachine® 5 year full service guarantee
|
| KI-Machine S | |
| KI-Machine M | |
| KI-Machine Custom |
Free exchange of defective hardware
|
| All products | within max. 60 minutes |
|
Minimum Contract Period
|
|
|---|---|
| All products | 1 Month |
Notice period
|
| All products | 14 days to the end of the contract |
Payment method
|
| All products | Debit, credit card, transfer, PayPal |
Billing period
|
| All products | monthly, quarterly, semi-annually, annually, 24 months |
| All products | up to 20% discount on your dedicated Server |
1The gross prices stated apply to consumers and include 19% VAT for Germany, other EU countries may differ. Charges are payable in advance for the respective billing period.
