NVIDIA Blackwell Systems
NVIDIA B200 & GB200 Blackwell Servers
Blackwell arrives in two shapes that are not interchangeable. HGX B200 is a familiar 8-GPU node that drops into a conventional rack and replaces an H100 or H200 node in your existing operational model. GB200 NVL72 is a rack — 72 GPUs and 36 Grace CPUs as one liquid-cooled NVLink domain, delivered and commissioned as a unit.
The second is not a bigger version of the first. It is a different facility conversation, with different power density, different cooling and a different installation path. We quote both, and we will tell you plainly when your data centre is not ready for the rack-scale option.
10 configurations below · quotes returned within 48 business hours · purchase orders accepted
NVIDIA B200 & GB200 Blackwell 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.
B200 GPU Server
Next-generation Blackwell compute for organizations planning their next AI build-out.
- Latest-generation performance
- Future-proof investment
- Roadmap-aligned planning
NVIDIA DGX B200
Eight Blackwell GPUs and 1,440GB of HBM3e in NVIDIA’s unified training and inference platform.
- One accountable vendor
- Blackwell generation
- A designed scaling path
NVIDIA DGX GB200 NVL72
Seventy-two Blackwell GPUs and thirty-six Grace CPUs as one liquid-cooled NVLink domain.
- One NVLink domain
- Rack-scale density
- Delivered as a unit
Nexus Compute HGX B200 8-GPU 4U Liquid-Cooled Training Node
Eight liquid-cooled Blackwell GPUs in 4U for frontier-scale model training.
- Maximum density per rack
- Lower cooling overhead
- Tested before delivery
Nexus Compute HGX B200 8-GPU 10U Air-Cooled Server
Drop-in Blackwell training and inference with no liquid loop required.
- No liquid retrofit
- Faster time to deploy
- Configured and warranty-backed
Nexus Compute GB200 NVL72 Grace-Blackwell Rack
A liquid-cooled rack as one GPU for trillion-parameter training and inference.
- Rack acts as one GPU
- Real-time giant-model inference
- Delivered as a system
Nexus Compute GB200 NVL2 MGX Single-Node Inference Server
Grace-Blackwell coherent memory in one node for mainstream LLM inference.
- Right-sized Grace-Blackwell
- Large coherent memory
- Integrates into your DC
Nexus Compute HGX B200 8-GPU EPYC Inference Server
EPYC-driven Blackwell density for high-throughput, low-latency model serving.
- High serving throughput
- Efficient cost-per-request
- EPYC I/O headroom
Nexus Compute GB200 NVL72 SuperPOD-Ready Scale Unit
Multi-rack Grace-Blackwell AI factory wired for non-blocking scale-out.
- Scale beyond one rack
- Engineered as one factory
- One accountable supplier
Nexus GB200 NVL72 Blackwell Rack-Scale Cluster
72 Blackwell GPUs as one giant GPU — exascale-class AI in a single rack.
- One rack, one giant GPU
- Real-time at trillion scale
- Roadmap-aligned sourcing
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.
B200 server — buyer questions
What is the difference between HGX B200 and GB200 NVL72?
HGX B200 is an 8-GPU server you rack like any other node. GB200 NVL72 is a full rack of 72 Blackwell GPUs and 36 Grace CPUs wired as a single NVLink domain, liquid-cooled, and treated as one system. NVL72 gives you a far larger coherent memory space for the biggest models; HGX B200 gives you Blackwell performance without rebuilding the room around it.
Does GB200 NVL72 require liquid cooling?
Yes. It is a direct-liquid-cooled rack and needs the facility water, the power density and the floor loading to match. This is the part of a Blackwell project that takes longest and it should be scoped before anything is ordered. We will walk through your site constraints before quoting.
Can we start with B200 nodes and scale to rack-scale later?
That is a common and sensible path — prove the workload on air-cooled or liquid-cooled HGX B200 nodes, then commit to NVL72 once the utilisation and the facility work justify it. We will design the fabric so the first purchase is not stranded.
Compare with other platforms
NVIDIA H200 Systems
NVIDIA H200 Servers
View systemsNVIDIA H100 Systems
NVIDIA H100 Servers
View systemsNVIDIA DGX
NVIDIA DGX Systems
View systemsBrowse the full range of GPU servers and AI clusters, or the component catalogue for drives, memory, optics and spares.