What is an HGX vs a DGX system?

DGX (Data Centre GPU) is NVIDIA's own branded complete system — GPU board, NVSwitch, CPU, memory, storage, networking, BMC, and chassis are all assembled and validated by NVIDIA. HGX is the OEM platform: NVIDIA supplies the GPU baseboard with NVSwitch and GPU modules to partners like Supermicro, Dell EMC, HPE, and Lenovo who build their own complete server chassis around it. HGX systems offer more flexibility in chassis choice, networking options, and storage configurations; DGX systems are reference-validated and come with NVIDIA Base Command software.

How much power does a DGX H100 need?

The DGX H100 draws up to 10.2 kW. It requires two C19/C20 connections from separate 30A, 240V circuits in North America, or a 3-phase 32A feed in Europe. A standard 20A, 120V residential circuit is completely unsuitable. Plan for dedicated PDUs or direct panel feeds, and verify your UPS capacity before ordering.

Does an AI server need liquid cooling?

Entry and mid-range AI servers (2–8 GPU PCIe configurations) can use high-airflow air cooling, though high-velocity fan noise and exhaust heat at 10 kW+ make rear-door heat exchangers strongly advisable. DGX H100 and HGX H100/H200 in SXM form factor are designed for air cooling at the system level but generate significant heat at the rack level — liquid cooling becomes practically necessary above 20–30 kW per rack. The GB200 NVL72 at 120 kW per rack requires direct liquid cooling with no air-only option.

Can I run a 175B model on a DGX H100?

At FP16, a 175B model requires approximately 350 GB of VRAM — more than the 640 GB in a DGX H100 can hold after accounting for inference KV cache overhead. In practice, INT8 quantisation halves this requirement to ~175 GB, fitting comfortably. The DGX H200 (1.1 TB HBM3e) can run 175B models in FP16 with headroom. A GB200 NVL72 (13.5 TB) can run multiple instances of 175B models simultaneously.

What networking is required between multiple AI servers?

For training across multiple nodes, high-speed RDMA fabric is essential. NVIDIA InfiniBand HDR (200 Gb/s) or NDR (400 Gb/s) is the standard for DGX/HGX cluster networking. Alternatively, RoCEv2 (RDMA over Converged Ethernet) with 400 GbE ConnectX-7 NICs is a lower-cost option for inference clusters. All-to-all collective operations (AllReduce) during training are extremely sensitive to inter-node bandwidth and latency — commodity 25/100 GbE switching introduces unacceptable stalls in large-scale training.

Do you offer installation or integration services?

Yes. For large AI server orders we can arrange factory configuration, burn-in testing, and logistics to your data centre or colocation facility. Contact our sales team to discuss your deployment timeline and facility requirements. We work with colocation providers across Europe, the UAE, and North America.

How do I contact MillionMiner for a server quote?

Our team is available 24/7 via WhatsApp (+49 176 777 888 33), email and phone. AI server purchases typically involve custom configurations and volume pricing — we will match your workload requirements to the right platform and provide a detailed quote including logistics.

AI servers and GPU workstations

Every ai server here ships assembled, burned in and workload-ready: tell us the job, LLM training, inference, fine-tuning or a local stack, and the configuration matches it. The ai server price sits on each listing, and where it reads contact for price, the quote comes back inside a day.…

47 products

Filters

Availability

In stock 8 Out of stock 39

Brand

NVIDIA 16 Supermicro 11 Gigabyte 7 ASUS 4 Lenovo 3 Quanta 3 Dell 1

On request

Supermicro

Product Power Price

Supermicro SuperServer SYS-741GE-TNRT 4-GPU H100 AI Server On request — On request

Supermicro AS-4124GO-NART+ 8x A100 HGX AI Server On request — On request

NVIDIA DGX A100 Deep Learning System (8x A100, 640GB) On request — On request

Exeton Quasar 640X 8x A100 AI Server (640GB NVLink) On request — On request

Lenovo NVIDIA HGX H200 141GB 700W 8-GPU Server On request — On request

Supermicro HGX H800 SuperServer SYS-821GE-TNHR On request — On request

GIGABYTE G893-ZD1-AAX5 HGX B200 8-GPU Server On request — On request

Gigabyte G492-ZD0 HGX A100 8-GPU Server, Dual AMD EPYC 7742, 4U (Used) On request — On request

Supermicro GPU A+ Server AS-8126GS-NB3RT NVIDIA HGX B300 NVL8 On request — On request

ASUS XA NB3I-E12 NVIDIA HGX B300 NVL8 GPU Server On request — On request

Supermicro SYS-111C-NR-G1 1U Rackmount X13 UP Gold Series CloudDC SuperServer In stock — $5,700.00

Supermicro SYS-521C-NR-G1 2U Rackmount X13 UP Gold Series CloudDC SuperServer In stock — $5,700.00

Product	Price
Supermicro SuperServer SYS-741GE-TNRT 4-GPU H100 AI Server On request	On request
Supermicro AS-4124GO-NART+ 8x A100 HGX AI Server On request	On request
NVIDIA DGX A100 Deep Learning System (8x A100, 640GB) On request	On request
Exeton Quasar 640X 8x A100 AI Server (640GB NVLink) On request	On request
Lenovo NVIDIA HGX H200 141GB 700W 8-GPU Server On request	On request
Supermicro HGX H800 SuperServer SYS-821GE-TNHR On request	On request
GIGABYTE G893-ZD1-AAX5 HGX B200 8-GPU Server On request	On request
Gigabyte G492-ZD0 HGX A100 8-GPU Server, Dual AMD EPYC 7742, 4U (Used) On request	On request
Supermicro GPU A+ Server AS-8126GS-NB3RT NVIDIA HGX B300 NVL8 On request	On request
ASUS XA NB3I-E12 NVIDIA HGX B300 NVL8 GPU Server On request	On request
Supermicro SYS-111C-NR-G1 1U Rackmount X13 UP Gold Series CloudDC SuperServer In stock	$5,700.00
Supermicro SYS-521C-NR-G1 2U Rackmount X13 UP Gold Series CloudDC SuperServer In stock	$5,700.00

Showing 12 of 47

Factory-Grade Systems

Configured, tested & burn-in verified

Freight Logistics

Insured pallet & air freight worldwide

Crypto Accepted

Pay with BTC, ETH, USDT & more

Expert Support

AI infrastructure architects on hand

AI Server Systems

Multi-GPU AI Servers: From 2-GPU Workstations to 8-GPU HGX Nodes

An AI server is not simply a PC with a GPU added. Serious AI infrastructure requires purpose-built platforms: high-bandwidth NVSwitch fabrics for GPU-to-GPU communication, high-capacity DDR5 memory, enterprise-grade CPUs with sufficient PCIe lanes to feed every GPU, NVMe storage for fast checkpoint I/O, and power delivery systems rated for 10+ kW per rack unit. We supply complete AI server and AI data center platforms from 2-GPU tower systems for research labs to NVIDIA HGX H100/H200 8-GPU nodes and the new GB200 NVL72 rack-scale architecture, with data center GPU server racks for scale ready for immediate deployment.

DGX H100 GPU Count

8 × H100 SXM5

640 GB total HBM2e via NVSwitch

NVL72 GPU Count

72 GPUs

GB200 in a single rack — 1 logical unit

DGX H100 AI Perf.

32 PFLOPS

FP8 sparse — combined 8 × H100

HGX H200 Memory

1.1 TB HBM3e

8 × H200 SXM5 — 38.4 TB/s aggregate BW

AI Server Platform Comparison

Key Platforms We Supply

NVIDIA DGX H100

8 × H100 SXM5, 640 GB

Complete AI supercomputer in a box. NVSwitch full-mesh at 900 GB/s per GPU. Dual Xeon or AMD Epyc, 2 TB DDR5. Reference platform for enterprise AI.

NVIDIA DGX H200

8 × H200 SXM5, 1.1 TB

Successor to DGX H100. 4.8 TB/s HBM3e per GPU. Preferred for large model inference at scale where memory capacity is critical.

NVIDIA HGX H100/H200

OEM platform (8-GPU)

The GPU board + NVSwitch complex shipped to OEM partners (Supermicro, Dell, HPE, Lenovo). Enables identical GPU fabric in custom server chassis.

NVIDIA GB200 NVL72

72 × GB200 GPUs

Single rack-scale system. 36 Grace CPU + 72 Blackwell GPU modules. 1.44 ExaFLOPS FP4. NVLink 5 fabric — entire rack operates as one GPU.

4-GPU Tower / Rack (RTX PRO 6000)

Up to 4 × 96 GB

Workstation-class server for research labs and AI startups. Dual-socket Xeon W9 or Threadripper Pro, 4 × RTX PRO 6000 BW via NVLink Bridge pairs.

2-GPU Inference Server (A100 / H100 PCIe)

2 × 80 GB HBM2e

Cost-efficient inference nodes. Standard 2U chassis, dual PCIe 5.0 slots. Fits standard rack colocation with no liquid cooling required.

1U Edge Inference (L4 / A2)

4–8 × 24 GB PCIe

High-density, low-power inference. Single-slot L4 GPUs at 72 W each. Up to 8 GPUs in 1U at under 600 W total — ideal for co-lo edge PoPs.

The Architecture Behind AI Servers

Why AI Servers Are Fundamentally Different from Standard Servers

The fundamental distinction between a general-purpose server and an AI server is the GPU fabric. In a standard server, GPUs communicate via PCIe — a bus architecture rated at roughly 32–64 GB/s bidirectional per slot. In an NVSwitch-based AI server, every GPU in the system connects to every other GPU simultaneously through the NVSwitch ASIC at 900 GB/s per GPU (H100) or 1.8 TB/s per GPU (H200 / Blackwell). This is not a bottleneck improvement — it is a qualitative change that enables model parallelism across GPUs as if they were a single larger device.

The CPU and memory subsystem must match GPU bandwidth demands. The DGX H100 pairs its 8-GPU NVSwitch fabric with dual Intel Xeon Platinum CPUs and 2 TB of DDR5 ECC RAM — enough system memory to stage training datasets, run preprocessing pipelines, and manage checkpoint I/O without becoming the bottleneck. PCIe 5.0 lanes are carefully distributed so that every GPU has a direct x16 connection to the CPU fabric.

Power delivery is an engineering challenge in its own right. A fully loaded DGX H100 draws approximately 10.2 kW — requiring 2 × 30A, 240V circuits in North America or 3-phase 32A in Europe. Cooling is equally demanding: NVIDIA's reference design uses a rear-door heat exchanger or direct liquid cooling for the GPU module in DGX configurations. Planning your facility's power and cooling before ordering is not optional — it is the first step.

Rack-Scale AI

NVIDIA GB200 NVL72: 72 GPUs as One Logical Accelerator

The GB200 NVL72 is NVIDIA's most ambitious system architecture to date. A single rack houses 36 Grace CPU modules and 72 Blackwell GB200 GPU dies, all connected over a fifth-generation NVLink fabric at 1.8 TB/s per GPU — enabling the entire rack to behave as a single logical accelerator. The aggregate FP4 sparse compute across all 72 GPUs exceeds 1.44 ExaFLOPS and total HBM3e memory reaches 13.5 TB.

Each GB200 module pairs a Grace Arm CPU with two Blackwell GPU dies via NVLink-C2C at 900 GB/s — eliminating the PCIe interface between CPU and GPU entirely. The CPU serves as a high-bandwidth memory controller and orchestration engine, not a bottleneck. This design makes the NVL72 uniquely capable for trillion-parameter model inference at low latency: the entire model fits within the rack's memory, all-reduce operations happen inside NVLink fabric, and no inter-node network is required for single-model serving.

From a facility standpoint, the NVL72 draws up to 120 kW and requires direct liquid cooling — a dedicated 3-phase power feed and facility liquid cooling infrastructure are mandatory. The economics are compelling at hyperscale: one NVL72 rack replaces what previously required entire rows of DGX H100 nodes for equivalent inference throughput.

GB200 NVL72 Specifications

Full Rack Spec Sheet

GPU Dies 72 × Blackwell GB200

CPU Modules 36 × Grace Arm (72 cores each)

GPU Memory 13.5 TB HBM3e total

GPU Memory BW 345.6 TB/s aggregate

AI Performance 1.44 ExaFLOPS FP4 sparse

GPU Interconnect NVLink 5.0 — 1.8 TB/s per GPU

CPU-GPU Link NVLink-C2C — 900 GB/s bidirectional

System Memory Up to 13.5 TB LPDDR5X

Power Draw Up to 120 kW per rack

Cooling Direct Liquid Cooling (mandatory)

Networking 8 × 400 GbE / InfiniBand per rack

Buyer's Guide

Which AI Server Is Right for Your Use Case?

From lab-scale experimentation to full hyperscale deployment — match your workload to the right platform tier.

Research / Lab

2–4 GPU Tower or 2U Rack

GPUs A100 80GB / RTX PRO 6000

GPU Memory 160–384 GB GPU VRAM

Power 2–4 kW total draw

Model fine-tuning, RAG development, local inference, research reproducibility. Fits under a desk or in a standard rack. No facility upgrades typically required.

Startup / SME

HGX H100 OEM 8-GPU

GPUs 8 × H100 SXM5 80GB

GPU Memory 640 GB HBM2e

Power ~10 kW draw

Full fine-tuning of 13–70B models. Production inference for AI-native products. Needs 3-phase power or dual 30A feed. Air cooling viable with rear-door HX.

Enterprise

DGX H200 / HGX H200

GPUs 8 × H200 SXM5 141GB

GPU Memory 1.1 TB HBM3e

Power ~11 kW draw

Full-parameter training of 70–400B models. Low-latency inference for very large models. Liquid cooling recommended. The current industry standard for serious AI workloads.

Hyperscale

GB200 NVL72

GPUs 72 × Blackwell GB200

GPU Memory 13.5 TB HBM3e

Power Up to 120 kW

Trillion-parameter model pre-training and large-batch inference. Full liquid cooling and dedicated 3-phase power mandatory. Rack-as-a-single-GPU architecture via NVLink 5.

FAQ

Frequently Asked Questions

Ready to Deploy Your AI Infrastructure?

Browse our full range of AI server platforms above, from research-grade 2-GPU nodes to DGX H200 and GB200 NVL72 rack-scale systems. Our team will help you plan power, cooling, networking, and logistics end-to-end.

Browse AI Servers Contact Sales