
6× H200 AI training server
Gigabyte G493-ZB1-AAP1 · AI training
- Cores
- 128
- GB memory
- 1,536
- GB GPU memory
- 846
As specified, ex VAT£200,932
Start from this buildPrice a real configuration in minutes and take it away as a PDF quote, built to your exact requirements and delivered ready for deployment.
Already have a config? Get in touch or call +44 20 7183 3135

6× H200 NVL GPU server
Indicative price, GBP ex VAT£207,028



Configure and price in minutes
If it is listed, it works together
Full burn-in before it leaves us
Built, tested and shipped from the UK

Part of Options Direct, named in the LDC Top 50 2026 and featured in The Times.







Start from a proven configuration and adjust every component to match your exact requirements. All specs are fully configurable.

Gigabyte G493-ZB1-AAP1 · AI training
As specified, ex VAT£200,932
Start from this buildThe same rows for every build, so the differences are easy to read. Each one opens in the configurator.

Gigabyte G493-ZB1-AAP1 · AI training
As specified, ex VAT£200,932
Start from this build
Supermicro GPU A+ Server AS -5126GS-TNRT2 · GPU / AI
As specified, ex VAT£95,936
Start from this build
ASUS RS521A-E12-RS24U · Storage
As specified, ex VAT£33,963
Start from this buildServerBuilds hardware is deployed across AI labs, media companies, financial institutions and cloud providers. Pick your workload to start from its typical platform.






Typical platform for AI trainingGPU servers
Distributed multi-GPU training with NVLink interconnect, RDMA-capable networking and NVMe-oF shared storage. Configured for PyTorch, JAX and DeepSpeed.
Start from the Gigabyte G493-ZB1-AAP1or customise every part
Low-latency serving infrastructure for production ML models. A mix of GPU and high-memory CPU nodes with vLLM, TensorRT and Triton Inference Server support.
Start from the Supermicro GPU A+ Server AS -5126GS-TNRT2or customise every part
High core-count platforms with large RAM pools for VMware vSphere, Proxmox VE and Hyper-V. Multi-node clusters with shared NVMe storage and live migration.
High-memory configurations for PostgreSQL, Oracle, MySQL and in-memory databases like Redis and SAP HANA. NVMe-backed storage with enterprise SAS redundancy.
Dense JBOD expansion and all-flash arrays with Ceph, MinIO or ZFS. Scale capacity independently from compute with 10GbE to 100GbE back-end networking.
GPU-accelerated render nodes for Blender, V-Ray, Redshift and DaVinci Resolve pipelines. NVENC/NVDEC hardware encoding for broadcast and post-production.
Start from the Gigabyte G493-ZB1-AAP1or customise every part
Each platform is optimised for a specific class of workload. Select a category to begin configuring. Every option shown is guaranteed compatible.
AI / HPCGPU serversMulti-GPU compute for AI training and inference workloadsConfigure
AMD EPYCAMD EPYC serversHigh core count compute nodes for virtualisation and HPCConfigure
NVMe storageNVMe storage serversHigh-throughput NVMe storage clusters and databasesConfigure
General computeRackmount servers1U to 4U dense Xeon and EPYC nodes for apps, CI/CD and edgeConfigure
Memory optimisedHigh memory serversTerabyte-scale DDR5 for in-memory databases, HANA and analyticsConfigure
VirtualisationVirtualisation hostsDual-socket density for vSphere, Proxmox and Hyper-V clustersConfigureCan’t find what you’re looking for? Get in touch or call us on +44 20 7183 3135
A straightforward process from first click to racked server, with expert support at every stage.
Use our online configurator to select your chassis, processor, memory, storage and networking. All options shown are guaranteed compatible.
Our engineers assemble your server at our UK headquarters, run a full burn-in test and verify every component is performing as expected.
From configuration to deployment, our expert team is available around the clock to answer questions and help you get the most out of your hardware.
Infrastructure teams choose ServerBuilds for mission-critical deployments because we obsess over component quality, compatibility and build precision.
ISO/IEC 27001:2022 certifiedInformation Security Management, independently audited.
Every build specifies ECC-registered DDR5, enterprise U.2 NVMe drives and redundant Titanium-rated PSUs. Components sourced from Samsung, Micron, Seagate and Broadcom. No consumer-grade substitutions, ever.
Assembled and burn-in tested at our UK facility. We ship to over 60 countries with full customs documentation, white-glove data centre delivery and colo rack installation coordination.
Select CPU, GPU count, RAM capacity and speed, NVMe layout, RAID card, networking and chassis independently. BIOS optimisation and firmware configuration included as standard on every system.
Our team includes former HPC cluster engineers and data centre architects. Every bill of materials is validated for component compatibility, thermal headroom and sustained performance under load before ordering.
Configure every component of your enterprise server, from GPU and CPU to memory, storage and networking, and get hardware that actually fits your workload. Delivered pre-built and tested.
Enterprise pricing available. Net-30 terms for qualified organisations.