B200 SXM from $1.69 per GPU-hour
Blackwell doubles the memory of an H100 and moves data almost twice as fast, which is what decides throughput once a model no longer fits comfortably. It is the machine you take when the model is large, the context is long, and the run is worth the hourly rate.
Billed per started minute. 1440 GB of GPU memory on a 8× node. Persistent storage $0.08 per GB-month. No egress fees.
- GPU memory
- 180GBHBM3e · MIG capable
- Bandwidth
- 7,700GB/sSets token throughput on inference
- Node sizes
- 1× · 2× · 4× · 8×Same price per GPU at every size
- Reclaim rate
- 5–10%Spot instances reclaimed over 30 days
What you get with each B200 node.
| Node | GPU memory | vCPU | RAM | Local NVMe | Spot | On-demand | Action |
|---|---|---|---|---|---|---|---|
| 1× B200 | 180 GB | 28 | 320 GB | 4 TB | $1.69/h | $2.49/h | Configure |
| 2× B200 | 360 GB | 56 | 640 GB | 8 TB | $3.38/h | $4.98/h | Configure |
| 4× B200 | 720 GB | 112 | 1,280 GB | 16 TB | $6.76/h | $9.96/h | Configure |
| 8× B200 | 1440 GB | 224 | 2,560 GB | 32 TB | $13.52/h | $19.92/h | Configure |
GPU-to-GPU link: NVLink 5, 1.8 TB/s GPU-to-GPU; PCIe Gen5 128 GB/s. Local NVMe is scratch space wiped when the instance ends; keep anything you need on a persistent disk.
16 B200 GPUs free right now.
Straight from the capacity pool the console books against. Prices are identical in every region — pick the one closest to your data.
A region with zero free GPUs still accepts reserved capacity requests — we hold hardware for a term rather than sell what is already taken.
What fits in 180 GB — and what it costs per hour.
Weights need about 2 GB per billion parameters at FP16 / BF16, 1 GB per billion parameters at FP8, 0.5 GB per billion parameters at INT4. We keep 20% of the memory free for activations, the KV cache and the CUDA context, then take the smallest node that still fits.
| Open model | Parameters | FP16 / BF16node · spot price | FP8node · spot price | INT4node · spot price |
|---|---|---|---|---|
| Mistral 7BAssistants, RAG, classification | 7.2B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Llama 3.1 8BAssistants, agents, fine-tuning | 8B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Gemma 2 27BHigher-quality assistants | 27B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Qwen2.5 32BCode and reasoning | 32B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Mixtral 8x7BMixture of experts, high throughput | 46.7B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Llama 3.3 70BThe common production baseline | 70B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Qwen2.5 72BMultilingual, long context | 72B | 1× $1.69/h | 1× $1.69/h | 1× $1.69/h |
| Mixtral 8x22BMixture of experts, large capacity | 141B | 2× $3.38/h | 1× $1.69/h | 1× $1.69/h |
| Llama 3.1 405BLargest widely used open model | 405B | 8× $13.52/h | 4× $6.76/h | 2× $3.38/h |
Real usage depends on context length, batch size and the serving engine: a long context can add tens of gigabytes of KV cache. Treat the table as the floor, not the ceiling — and if a job is close to the limit, take the next node size or quantise one step further.
What a dollar buys on a B200.
Ranked 21 of 21 data-center models in this catalogue on cost per gigabyte of GPU memory.
GPU memory you get for $1.00 of spot time, per hour.
Bandwidth decides token throughput far more often than raw FLOPS.
Spot price divided by published FP32 throughput.
One GPU on spot, billed per minute, storage excluded.
Against the rest of the market
50 providers · September 2026- Median on-demand elsewhere
- $5.60Our on-demand is 56% below it
- Cheapest on-demand seen
- $3.12Our on-demand $2.49 is 20% below it
- Cheapest spot seen
- $2.40Our spot $1.69 is 30% below it
- 24 hours on one GPU
- $40.56 on spot$59.76 on-demand · $74.88 at the cheapest rate found elsewhere
Index built from published prices for the same GPU across 137 providers (public price index (getdeploying.com)), September 2026. How the guarantee is enforced.
What people run on a B200 SXM.
Training and full fine-tuning
Enough memory for optimiser states and activations on models a smaller card can only run in inference.
Serving large models
A 70B model at FP8 on a single GPU, with room for a long context window.
Multi-GPU jobs over NVLink
GPU-to-GPU traffic stays off the PCIe bus, which is what makes tensor and pipeline parallelism worth it.
Partitioned serving (MIG)
The GPU can be split into isolated instances, each with its own memory and compute slice.
NVIDIA B200 SXM, as published by NVIDIA.
Taken from the vendor datasheet. Where a figure is not published, the row is absent rather than estimated.
- Vendor
- NVIDIA
- Architecture
- Blackwell (2025)
- Segment
- Datacenter
- Memory
- 180 GB HBM3e
- Memory bandwidth
- 7,700 GB/s
- FP32
- 75 TFLOPS
- Board power
- 1000 W
- Form factor
- SXM6
- Interconnect
- NVLink 5, 1.8 TB/s GPU-to-GPU; PCIe Gen5 128 GB/s
- Multi-instance (MIG)
- Supported by the GPU
- 8-bit float (FP8)
- Supported by the architecture
Source: dam-cdn.nvd.orangelogic.com · www.nvidia.com · verified against the vendor document
Software that runs on it
8 templatesPre-built environments, pulled on the node before you land on it. Or bring any OCI image from a public or private registry.
On the spot tier
5–10% reclaimed · 30 d- 2 minutes of notice on the metadata endpoint, a webhook and the console
- The instance is stopped, not deleted — disk and IP stay attached
- Auto-relaunch on the next free B200, or switch the same disk to on-demand
- You save $19.20 a day per GPU against on-demand
If the B200 is not the right fit.
Same memory or more, 36% cheaper on spot
MI300X 192 GB HBM3 · CDNA 3 $0.99/h spotNext step up in memory: 192 GB instead of 180 GB
Or put two of them side by side: B200 vs MI325X · B200 vs MI300X
B200 SXM — the questions that come up
How much does a B200 SXM cost per hour?
Spot is $1.69 per GPU-hour and on-demand is $2.49, both billed per minute — every started minute costs the hourly price divided by 60, so $0.0282 on spot. Reserved capacity is $1.85 per GPU-hour on a monthly commitment. Storage and egress are not included in that rate: persistent disks are $0.08 per GB-month and there are no egress fees.
How many GPUs can I put in one instance?
Node sizes are 1×, 2×, 4×, 8× — up to 8 B200 SXM GPUs in a single virtual machine, with CPU, RAM and local NVMe scaled with the GPU count. The price per GPU is identical at every node size.
Can it run a 70B model?
Yes. At FP8, a 70B model needs about 70 GB for the weights, so it fits on 1× B200 SXM ($1.69 per hour on spot) with roughly 20% of the memory left for activations and the KV cache.
Which regions have it?
Available now in US East (Virginia). Prices are identical in every region; pick the one closest to your data.
How often is a spot B200 reclaimed?
Over the trailing 30 days, 5–10% of spot instances on this model were reclaimed. You get a 2-minute notice on the metadata endpoint, through a webhook and in the console; the instance is then stopped, never deleted, and the persistent disk stays attached.
Anything else about this model? Ask an engineer — the same people run the nodes.
A B200 SXM for $1.69 an hour, running in under 60 seconds.
Pay as you go — no contracts, no minimum commitment. Add credit, launch, stop whenever you want.
Billed per minute from the moment the instance is reachable. Minimum credit $40, no subscription. 16 B200 GPUs available right now.