How much does a GPU cluster cost?

A GPU cluster's cost is driven by the GPU type, node count, and commitment term. As a 2026 reference, an 8× H100 node runs roughly $15–28/hr on-demand, an 8× H200 node $24–36/hr, and an 8× A100 node $9–18/hr. At cluster scale (16–128+ GPUs) on 3–12 month reserved terms, partner pricing is typically 25–45% below on-demand. Exact pricing depends on availability, region, interconnect, and term length.

Is on-demand or reserved cheaper for a GPU cluster?

Reserved is materially cheaper for sustained workloads. On-demand is flexible but priced for short bursts; a 1-year reserved commitment on the same GPUs is typically 25–45% lower per GPU-hour. The break-even is usually a few months of continuous use — beyond that, reserved capacity almost always wins, and it also guarantees availability of constrained GPUs like H100/H200.

Do I need InfiniBand for a GPU cluster?

For multi-node distributed training, yes — InfiniBand (or an equivalent RDMA fabric) provides the GPU-to-GPU bandwidth and low latency that keep GPUs fed during all-reduce communication. Without it, scaling efficiency falls off sharply past a single node. For embarrassingly-parallel inference or single-node jobs, standard Ethernet can be sufficient.

How many GPUs do I need to train an LLM?

It depends on model size and timeline. Fine-tuning a 7–13B model often fits on a single 8-GPU H100 node. Pre-training or large fine-tunes in the 70B+ range typically use 16–64+ GPUs with InfiniBand to finish in a reasonable window. The practical lever is wall-clock time: more GPUs reduce training time roughly linearly until communication overhead dominates.

Can I rent a GPU cluster month-to-month?

Yes. Reserved cluster capacity is commonly offered on 3, 6, or 12-month terms, and flexible/month-to-month arrangements exist at a higher rate. Shorter terms cost more per GPU-hour but avoid long lock-in; longer terms unlock the deepest discounts.

What's the difference between a node and a cluster?

A node is a single server, almost always 8 GPUs connected by NVLink. A cluster is multiple nodes linked by InfiniBand so they train or serve as one system. Pricing usually scales with GPU count, with a small premium for the cluster interconnect and orchestration.

How fast can I get GPU cluster capacity?

On-demand single nodes are often available immediately. Reserved cluster capacity for constrained GPUs (H100/H200/B-series) depends on inventory — QuantaCloud brokers across multiple data-center partners and returns a custom quote, typically within 24 hours, for 16+ GPU reserved or cluster configurations.

How much does an H100 SXM5 cluster cost per hour?

Lambda Labs lists reserved H100 SXM5 InfiniBand clusters from $5.54 to $6.16 per GPU-hour depending on size: 16–63 GPUs at $6.16/GPU-hr, 64–255 GPUs at $5.85/GPU-hr, 256+ GPUs at $5.54/GPU-hr. That is $44.32–$49.28 per 8-GPU node-hour, on a 1-week minimum term, in Texas, US and Washington DC, US. A 16-GPU cluster runs about $16,558 per week.

How much does a B200 SXM cluster cost per hour?

Lambda Labs lists reserved B200 SXM InfiniBand clusters from $8.87 to $9.86 per GPU-hour depending on size: 16–63 GPUs at $9.86/GPU-hr, 64–255 GPUs at $9.36/GPU-hr, 256+ GPUs at $8.87/GPU-hr. That is $70.96–$78.88 per 8-GPU node-hour, on a 1-week minimum term, in Ohio, US and Texas, US. A 16-GPU cluster runs about $26,504 per week.

ReferenceUpdated September 2026Live reserved cluster rates16–1,024+ GPUs

GPU Cluster Pricing, 2026

Q: What is a GPU cluster?

A GPU cluster is a group of GPU servers (nodes) connected by a high-bandwidth, low-latency network — usually NVLink within a node and InfiniBand between nodes — so they act as one machine for large training or inference jobs. A single node is typically 8 GPUs; a cluster is multiple nodes, from 16 up to thousands of GPUs.

A GPU cluster is more than one GPU node — typically 8 GPUs each, linked inside the node by NVLink — joined by a node-to-node fabric such as InfiniBand, RoCEv2 or high-speed Ethernet so they train and serve as one machine. This page covers what a cluster costs in 2026, how reserved cluster capacity compares to renting single nodes on-demand, and when each makes sense.

As a single-node baseline: an 8× H100 node on-demand runs roughly $15–28/hr, an 8× H200 node $24–36/hr, and an 8× A100 node $9–18/hr. Those are one node with no fabric between nodes. Multi-node reserved clusters on 3–12 month terms typically price 25–45% below on-demand per GPU-hour.

Live today: Lambda Labs reserved H100 SXM5 clusters from $5.54/GPU-hr and Lambda Labs reserved B200 SXM clusters from $8.87/GPU-hr, 16 to 1,536 GPUs on InfiniBand with a 1-week minimum. See the live cluster rates.

What Is a GPU Cluster?

A single node is one server, almost always 8 GPUs connected internally by NVLink. A cluster is two or more nodes connected by a node-to-node fabric — InfiniBand, RoCEv2 over Ethernet, or plain high-speed Ethernet — so dozens or thousands of GPUs cooperate on a single training or inference workload. NVLink never leaves the node; the fabric is what makes it a cluster.

Clusters are how teams pre-train and large-fine-tune models that don't fit on one node, and how high-throughput inference fleets are run with predictable capacity. The defining feature versus on-demand single GPUs is the interconnect: it keeps every GPU fed during the all-reduce communication that dominates distributed training.

Live GPU Cluster Pricing

Reserved multi-node cluster rates published by providers, verified against their pricing pages and refreshed hourly. Per-GPU-hour pricing steps down with cluster size; every listing names its node-to-node fabric and minimum term. Pick a size in the live table to compare it against on-demand nodes.

Lambda Labs · 16–512× H100 SXM5

8-GPU nodes · 80 GB VRAM per GPU · NVLink within each node · InfiniBand between nodes · 1-week minimum

$5.54/GPU/hr

from $44.32/node-hr

Cluster size	Per GPU-hour	Per 8-GPU node-hour	Per week at minimum size
16–63 GPUs	$6.16	$49.28	$16,558
64–255 GPUs	$5.85	$46.80	$62,899
256+ GPUs	$5.54	$44.32	$238,264

Sizes: 16, 32, 64, 128, 256, 512 GPUs·Regions: Texas, US, Washington DC, US·Per node: 208 vCPU, 1,800 GB RAM, 22 TB SSD

Compare in live table Reserve at Lambda Labs

Lambda Labs · 16–1536× B200 SXM

8-GPU nodes · 192 GB VRAM per GPU · NVLink within each node · InfiniBand between nodes · 1-week minimum

$8.87/GPU/hr

from $70.96/node-hr

Cluster size	Per GPU-hour	Per 8-GPU node-hour	Per week at minimum size
16–63 GPUs	$9.86	$78.88	$26,504
64–255 GPUs	$9.36	$74.88	$100,639
256+ GPUs	$8.87	$70.96	$381,481

Sizes: 16, 32, 64, 128, 256, 512, 1024, 1536 GPUs·Regions: Ohio, US, Texas, US·Per node: 208 vCPU, 2,900 GB RAM, 22 TB SSD

Compare in live table Reserve at Lambda Labs

On-Demand Node Baseline

Not cluster prices: what a single 8-GPU node costs by the hour, with no fabric between nodes. This is the number a cluster quote is compared against. Per-GPU ranges are typical 2026 market rates; the cheapest live 8-GPU node comes from our tracker and refreshes hourly.

GPU	VRAM	On-Demand / GPU·hr	Cheapest 8-GPU Node Now	Inside the node	Cluster fabric
H100 SXM 80GB	80 GB HBM3	$1.90–3.50	$19.51/hrCoreWeave · United States · $2.44/GPU	NVLink + NVSwitch	InfiniBand or RoCEv2
H200 SXM 141GB	141 GB HBM3e	$3.00–4.50	$20.64/hrCoreWeave · United States · $2.58/GPU	NVLink + NVSwitch	InfiniBand or RoCEv2
B200 SXM	192 GB HBM3e	$4.00–6.50	$34.87/hrCoreWeave · United States · $4.36/GPU	NVLink 5 + NVSwitch	InfiniBand or RoCEv2
GB300 NVL	288 GB HBM3e	$5.50–9.00	Rack-scale (NVL72)typical range	NVLink 5 across the NVL72 rack	InfiniBand or RoCEv2
A100 SXM 80GB	80 GB HBM2e	$1.10–2.20	$9.20/hrDenvr · Virginia · $1.15/GPU	NVLink + NVSwitch	InfiniBand or RoCEv2
MI300X	192 GB HBM3	$1.90–3.50	$24.64/hrCirrascale · United States · $3.08/GPU	Infinity Fabric	InfiniBand or RoCEv2

Single-node on-demand, not cluster quotes. Per-GPU rates update every 60 seconds on the linked live pricing pages; confirm current cluster pricing before procuring.

Reserved vs On-Demand

On-Demand

Flexible, pay-by-the-hour, available immediately for single nodes. Best for experimentation, short runs, and bursty workloads. Priced at the top of the range, and constrained GPUs (H100/H200/B-series) can be unavailable at peak.

Reserved / Cluster

Two or more nodes on a 1-week to 12-month commitment, 16+ GPUs, typically 25–45% cheaper per GPU-hour, with guaranteed availability and a node-to-node fabric. Best for sustained training and production inference. Break-even versus on-demand is usually a few months of continuous use.

Interconnect & InfiniBand

Within a node, NVLink connects the 8 GPUs at terabytes per second. Between nodes, InfiniBand (commonly 400–800 Gb/s per GPU on modern fabrics) provides the low-latency RDMA bandwidth that makes multi-node training scale near-linearly.

For distributed training the interconnect is not optional — without it, scaling efficiency drops sharply past a single node because GPUs stall waiting on gradient communication. For single-node jobs or embarrassingly-parallel inference, standard Ethernet is often enough.

QuantaCloud

Need GPUs at scale?

Building out an inference fleet or training cluster? QuantaCloud brokers reserved capacity across multiple data center partners. 16+ GPUs, flexible terms, custom quote in 24 hours.

No waitlist24hr quote turnaroundInfiniBand fabric

GPU Cluster FAQ

Citation

GPUPerHour: GPU Cluster Pricing Reference (September 2026)

Source: https://gpuperhour.com/gpu-cluster

6 GPU classes. Market ranges, manually verified.

28 live reserved cluster listings across 2 GPU products, refreshed hourly.

Last updated: September 2026

Per-GPU rates are tracked live across providers and update every 60 seconds. Cluster and reserved pricing varies by availability, region, term, and interconnect — confirm current rates before making procurement decisions.