RTX PRO 6000 and H200 remain our core sourcing focus. DGX, additional system platforms, and AI networking are reviewed per project. Sourcing, availability, exact configuration, installation, licensing, and support must be confirmed in a written proposal. A listing is not a stock or authorized-reseller claim.
DGX B200 and DGX B300.
Complete systems, rather than standalone add-in GPU cards. These manufacturer specifications are planning references, not an offer of inventory.
DGX B200
Eight NVIDIA B200 GPUs with 1,440 GB of total GPU memory and fifth-generation NVLink connectivity. A system platform for AI training, inference, and analytics.
- Integrated GPU, CPU, storage, and cluster-networking configuration
- Confirm facility power, rack space, cooling, and network integration
DGX B300
Eight NVIDIA B300 GPUs, each with 288 GB of GPU memory, for approximately 2.3 TB total. Designed for demanding AI training and inference workloads.
- Review the exact network interfaces and complete system bill of materials
- Installation and support eligibility must be part of the project plan
Total GPU memory is aggregate capacity across the system. Usable model capacity depends on the software, precision, and parallelism strategy.
Prepare a DGX inquiry.
The GPU baseboard and host CPU are manufacturer-defined. NVIDIA documents 2 TB host memory with up to 4 TB for these systems; upgrade eligibility and the exact option SKU are confirmed before quoting.
NVIDIA requires installation by NVIDIA Partner Network personnel or NVIDIA field service engineers; otherwise, the hardware warranty is voided. The installation provider and support entitlement must be verified before purchase. NexTec does not represent this listing as NVIDIA certification or authorization.
Read NVIDIA’s installation requirementsDGX, HGX, and rack-scale systems
DGX identifies NVIDIA’s complete AI system platform. HGX is a multi-GPU platform integrated into systems by server manufacturers. An HGX B200 or B300 inquiry must name the exact OEM server, GPU baseboard, support coverage, and configuration; it is not interchangeable with a DGX order.
GB200 / GB300 NVL72 projects add rack-scale integration and facility-level planning. Treat them as a separate deployment scope, with power, liquid cooling, rack delivery, and commissioning requirements confirmed against the selected manufacturer design.
Discuss an HGX or rack-scale projectFrom a system to a cluster
DGX BasePOD and DGX SuperPOD are broader infrastructure approaches that bring compute, networking, storage, and deployment planning together. They require an architecture and qualified delivery plan, not simply a GPU quantity.
- Define model size, training or inference goals, concurrency, and acceptance criteria
- Confirm storage throughput, fabric topology, optics, and management networks
- Verify delivery, installation, software licensing, support entitlement, and service responsibilities
Bring a deployment brief
Include the preferred model, system or rack count, workload, U.S. deployment location, facility power and cooling, network requirements, and target date. Feasibility and sourcing are reviewed before any commercial commitment.
Explore AI networking