NVIDIA A100 Systems
NVIDIA A100 Servers
The A100 is a generation behind current parts and that is frequently the point. For teams extending an existing A100 estate, for workloads already tuned to it, and for budgets where the price per usable FLOP matters more than the headline, it remains a rational purchase — and MIG partitioning still makes it one of the better inference platforms for serving many small models on one card.
We supply A100 in the configurations that actually get bought: 8-GPU SXM4 training nodes, 4-GPU PCIe with NVLink bridges, and MIG-partitioned inference servers.
8 configurations below · quotes returned within 48 business hours · purchase orders accepted
NVIDIA A100 Servers we configure and supply
Every system below is quoted to your workload — accelerator count, CPU, memory, storage and fabric are specified together rather than sold as a fixed SKU.
NVIDIA DGX A100 640GB
Eight A100 80GB GPUs with 640GB HBM2e and MIG partitioning across the node.
- MIG partitioning
- Matches an installed base
- Multi-tenant by design
NVIDIA DGX A100 320GB
Eight A100 40GB GPUs — the lower-memory DGX A100 configuration.
- Lower entry point
- MIG partitioning retained
- Matches an installed base
NVIDIA DGX Station A100
Four A100 GPUs, desk-side and refrigerant-cooled — a data-center-class node without the data center.
- No data center required
- NVLink at the desk
- MIG for shared use
Nexus HGX A100 8x SXM4 80GB Training Node
Eight NVSwitch-linked A100 80GB GPUs for full-scale model training.
- Full all-to-all bandwidth
- Burned-in and validated
- Proven training platform
Nexus A100 PCIe 80GB 4-GPU NVLink Server
Four NVLink-bridged A100 80GB cards for fine-tuning without datacenter density.
- 80GB memory per GPU
- Standard-rack friendly
- Paired NVLink bandwidth
Nexus A100 PCIe 40GB 8-GPU MIG Inference Server
Eight A100 40GB GPUs partitioned with MIG for dense multi-tenant inference.
- Up to 56 isolated instances
- High utilization economics
- Predictable tenant isolation
Nexus A100 SXM4 40GB 4-GPU HPC Node
Compact four-way SXM4 A100 node for tightly-coupled HPC and simulation.
- Low-latency collectives
- Dense compute, smaller footprint
- Double-precision strength
Nexus HGX A100 8x SXM4 80GB Liquid-Cooled Node
Liquid-cooled eight-way HGX A100 80GB for sustained high-density training.
- Sustained peak performance
- Higher rack density
- Improved power efficiency
What we need to quote accurately
A configuration quote takes minutes when these are known and days of back-and-forth when they are not. You do not need all of them to start — send what you have.
- The workload: model sizes, training or inference, and expected concurrency
- Node count, and whether this is a single system or a scaling cluster
- Rack power and cooling available per rack, and inlet temperature
- Existing estate — the vendor and management tooling you already run
- Network fabric: InfiniBand, Ethernet, and the speed you are standardised on
- Timeline, and whether the budget is approved or being built
How we quote
- 1
Send the requirement
An email, a bill of materials, or a rough description of the workload. All three work.
- 2
We validate the configuration
Accelerator, chassis, fabric, power and cooling checked against each other before anything is priced.
- 3
Itemised quote within 48 hours
Line-by-line pricing, current availability and lead time, with warranty terms stated.
A100 server — buyer questions
Is it still sensible to buy A100 systems?
It can be. If you are adding capacity to an existing A100 cluster, mixing generations inside one training job is awkward and matching what you run is often worth more than newer silicon. If you are serving many small models, MIG partitioning gives you hardware-isolated instances that later parts handle differently. If you are training large models from scratch, look at H100 or H200 instead — we will say so.
What is MIG and does it matter for our workload?
Multi-Instance GPU splits one A100 into as many as seven hardware-isolated instances, each with its own memory and compute slice. It matters if you are serving many small models or giving separate teams guaranteed capacity on shared hardware. It does not matter if you run one large job across whole GPUs.
Do you supply A100 40GB as well as 80GB?
Yes, in both SXM4 and PCIe form factors. The 40GB parts are usually the better value for inference and HPC work that fits; the 80GB parts matter when the model or the dataset does not.
Compare with other platforms
NVIDIA H100 Systems
NVIDIA H100 Servers
View systemsNVIDIA H200 Systems
NVIDIA H200 Servers
View systemsDell PowerEdge
Dell PowerEdge XE9680 GPU Servers
View systemsBrowse the full range of GPU servers and AI clusters, or the component catalogue for drives, memory, optics and spares.