Hetzner launches GEX45 GPU server for AI workloads

Hetzner has added the GEX45 GPU server to its dedicated server range, targeting entry-level professional GPU workloads.

The system uses an NVIDIA RTX PRO 4000 Blackwell SFF Edition graphics card with 24 GB of GDDR7 ECC VRAM, a 13th-generation Intel Core i5-13500 processor, 64 GB of DDR4 memory and two 512 GB NVMe solid-state drives.

Hetzner describes the GEX45 as suited to AI tasks including inference and fine-tuning, as well as 3D rendering, CAD and video production. It is positioning the server as a lower-cost entry point into dedicated GPU infrastructure for customers that do not need larger, more expensive systems.

The server costs EUR €214 a month and USD $249, with a one-off setup fee of EUR €209 and USD $249. The package includes an IPv4 address.

Hardware focus

The graphics card includes fifth-generation Tensor Cores with FP4 support and fourth-generation RT Cores. That combination makes the server suitable for workloads in which the GPU handles much of the compute demand.

The CPU has six performance cores and eight efficiency cores, for a total of 20 threads. This places the machine below larger multi-GPU or high-core-count server offerings, while still providing enough general-purpose processing for mixed CPU and GPU workloads.

Storage is based on NVMe flash drives, typically used to reduce loading times and improve data access speeds compared with older storage media. That matters for customers moving data sets for AI jobs, media files for video editing, or models and textures for rendering work.

European footprint

The launch also reflects continued demand for infrastructure hosted within Europe, particularly from customers handling sensitive data or working in regulated sectors. Hetzner says its dedicated servers run in the company's own ISO 27001-certified data centres in Europe.

That keeps customer data under European legal jurisdiction and subject to the General Data Protection Regulation. For some businesses, this can be a deciding factor when choosing between regional hosting providers and larger international cloud groups with more complex cross-border data arrangements.

Hetzner has long built its market position around owned infrastructure and comparatively simple pricing for dedicated servers and virtualised products. Adding a more specialised GPU server suggests it is responding to broader demand for AI-related compute without moving fully into the premium end of the GPU market.

Broader market

Competition in hosted GPU infrastructure has intensified as businesses seek systems for model inference, tuning smaller models and handling graphics-heavy workloads. While many providers have focused on high-end accelerators for large training jobs, there is also a growing market for smaller deployments that need predictable monthly costs and do not require a full-scale AI cluster.

The GEX45 appears aimed at that middle ground. Its specification is likely to appeal to software developers, creative professionals and smaller companies that need dedicated access to a professional GPU for steady workloads rather than burst access through a shared cloud service.

Founded in 1997, Hetzner operates data centre sites in Germany and Finland and has also invested in infrastructure in Singapore and the United States. The company is best known in Europe for its dedicated servers and virtual server products, serving workloads ranging from small web projects to larger enterprise and development use cases.

With the GEX45, Hetzner is extending that portfolio into a part of the market where AI demand, visual computing and data-location concerns increasingly overlap.