Retiring GPUs or compute systems?
We buy decommissioned GPU servers, enterprise accelerators, and surplus hardware. Share your inventory for a purchase evaluation.
Complete AI servers
RTX PRO, H200 HGX / NVL & custom systems
GPUs & accelerators
Compare editions, memory, power & cooling
Enterprise supply
Volume orders, recurring purchases & contracts
The whole system. Down to the detail.
Explore the actual hardware platform, its architecture, and the configuration choices that matter to your deployment.
Supermicro SYS-521GE-TNRT · 5U
RTX PRO 6000 AI Server
A configurable foundation for AI inference, private AI, visualization, and mixed enterprise workloads.
HGX reference: SYS-821GE-TNHR · 8U
NVIDIA H200 Systems
High-bandwidth HBM3e platforms for demanding AI training, inference, and scientific computing.
Your requirements. Your configuration.
Custom AI & HPC Servers
Right-size compute, memory, networking, and cooling for a single system or a phased rack deployment.
Three editions. Three different jobs.
RTX PRO 6000 Blackwell comes in Server, Workstation, and Max-Q Workstation editions. Each has 96 GB of memory. The system around it changes.
Server Edition
Passive cooling for qualified rack servers. The accelerator in our flagship 8-GPU configuration.
Technical details
Workstation Edition
Double flow-through cooling for high-performance desktop workstations.
Technical details
Max-Q Workstation Edition
Active cooling and a lower power envelope for dense multi-GPU workstations.
Technical detailsProduct imagery: NVIDIA. Max-Q here refers to the desktop Workstation Edition.
See the platform.
Understand the build.
Trace the GPU expansion area, host processors, ECC memory, drive bays, fan system, and power modules through the manufacturer’s actual chassis views.
- Named motherboard and supported CPU family
- Default bays, optional expansion, and proposed storage
- PCIe topology, power limits, and rack requirements
Compute built around your work.
AI & large language models
Inference, fine-tuning, training, and private enterprise AI.
02 / WORKLOADHPC & research
Scientific computing, engineering simulation, and data analytics.
03 / WORKLOADRendering & visualization
RTX rendering, digital twins, and demanding 3D workflows.
04 / WORKLOADGPU cloud infrastructure
Compute nodes, network fabrics, and data-center deployments.
A supply relationship
built to go further.
Bring procurement and engineering to the same table. Plan repeat orders, consistent configurations, phased releases, replacement hardware, and your long-term capacity roadmap.
Discuss enterprise supply