Single-socket platform
NX-G12 SP3
A flexible starting point for AI, rendering and GPU compute.
- AMD EPYC · SP3
- DDR4 ECC memory
- Up to 12 compatible GPUs*
Platform from $4,574 · GPUs and drives not included
GPU servers · Managed AI infrastructure
Configure a server for your workload, or launch a GPU rental operation with NodeX handling the build, setup and agreed day-to-day technical services.
Built in Bulgaria. Configured for you.
01 / The hardware
Three server platforms. Your choice of GPU, memory, storage and networking. We confirm compatibility, availability and final pricing in your quotation.
Single-socket platform
A flexible starting point for AI, rendering and GPU compute.
Platform from $4,574 · GPUs and drives not included
Dual-socket platform
Pair GPU capacity with a dual-processor DDR4 platform.
Platform from $5,857 · GPUs and drives not included
Dual-socket DDR5 platform
A DDR5 foundation for your next compute deployment.
Platform from $11,042 · GPUs and drives not included
*Supported GPU count depends on card dimensions, power, cooling and the complete configuration. Illustrations show the platform; your quotation defines the supplied build.
Custom high-memory GPUs
Explore custom RTX 3090 48GB and RTX 4090 48GB cards with dual-slot blower cooling. Modified configurations, with specifications and warranty confirmed for your order.

02 / Your deployment
Start with the part you need. We shape the proposal around your infrastructure, experience and goals.
For your own workloads
Choose the hardware for AI inference, development, rendering or compute. Add software setup and installation where needed.
For a GPU rental business
Own GPU servers and offer computing capacity through platforms such as Vast.ai. Agreed monitoring and technical services support a more passive income model with less daily involvement.
03 / From idea to operation
Define your budget, workloads, location and operating costs.
Approve the specification, then assemble, configure and test.
Arrange colocation or installation on your premises and platform setup.
Run with an agreed scope for monitoring, maintenance and support.
Rental returns depend on demand, utilisation, platform fees and operating costs. Income, profit and a fixed payback period are not guaranteed. Hardware and service costs are set out in your proposal.
04 / GPU guides
Before you choose a configuration, compare the GPUs we build with: published specifications, indicative EU prices and how much memory your models need. All guides
Server, Workstation and Max-Q editions compared, full specifications, EU prices from $11,900 and how it compares with the RTX 5090.
Read the guide141GB HBM3e specifications, NVL vs SXM, EU prices for one to eight GPUs and what an H200 NVL server needs.
Read the guideThe arithmetic for model weights and context, a sizing table from 8B to 670B parameters and the hardware for each tier.
Read the guideSpecifications, what fits in 48GB of VRAM, prices compared with the RTX 3090 48GB, RTX 5090 and RTX PRO 6000, and what to check before buying.
Read the guideBased in Bulgaria
NodeX is operated by Ен Ди Зет ЕООД / N D Z Ltd., based in Bansko, Bulgaria. We bring hardware selection, assembly and deployment planning into one conversation.
Your written quotation defines the equipment, delivery, warranty and services before you commit.
The practical details
Configure a platform and send an enquiry. We review compatibility, availability and requirements, then confirm the complete price, taxes, delivery and lead time in a written quotation.
Yes. The project can include hardware, software, marketplace onboarding, colocation or installation at your premises, plus agreed ongoing services. Rental income is variable and is not guaranteed.
Either route can be included. Power, cooling, connectivity, access and installation work are assessed for the proposed location before deployment.
Your quotation identifies hardware cover and any component exceptions. Monitoring, maintenance, response targets and on-site work are agreed separately for your selected service package.
Your next deployment starts here
A workload. A budget. A business idea. Start there.