GPU MODELS / V100 SXM2
V100 SXM2 cloud GPU pricing
2 listings across 2 providers. Cheapest published rate: $0.230/GPU/hr at RunPod. Sorted by per-GPU hourly price. Every row links to the provider's own published price.
WHAT IS THE V100 SXM2
The V100 was the first GPU with tensor cores and dominated deep-learning training from 2017 to 2020. Its HBM2 bandwidth is still respectable, but it predates BF16 and sparsity support, and modern Hopper chips out-compute it roughly 8x. Today it survives in cloud catalogs as a budget training and batch-inference option.
TENSOR COMPUTE
125 TFLOPS FP16 tensor (SXM2)
POWER DRAW
300 W (SXM2) / 250 W (PCIe)
INTERCONNECT
NVLink 2.0 300 GB/s (SXM2) or PCIe 3.0
WHAT IT IS GOOD AT
Budget training
Fine for small-to-mid models and coursework. FP16 at 125 TFLOPS trains, just slowly by modern standards.
Batch inference
900 GB/s HBM2 keeps latency-sensitive serving acceptable for models that fit in 32 GB.
Legacy CUDA stacks
Some older HPC codes are tuned for Volta; if your stack assumes it, the V100 is the cheap way to run it.
WHO SHOULD NOT RENT THE V100 SXM2
New training work where an A100 is within reach on price - the compute gap is 2.5x and growing in value.
Workloads needing BF16, FP8 or MIG - Volta has none of them.
| Provider | Config | VRAM | Location | $/hr | $/GPU/hr | Source |
|---|
 RunPod | V100 SXM2 × 1 | 16 GB | Global | $0.230 | $0.230 | receipt |
 CloudRift | V100 SXM2 × 1 | 16 GB | Not published | $0.250 | $0.250 | receipt |
THE V100 SXM2 MARKET, BY THE NUMBERS
2 providers publish 2 V100 SXM2 configurations on the panel today. Published per-GPU hourly rates run from $0.230 (RunPod) to $0.25 (CloudRift), a 1x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $0.25/GPU/hr. 1 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.
V100 SXM2 PRICE TREND, DAILY PANEL SNAPSHOTS
Across every V100 SXM2 row on the panel, the daily floor held flat at $0.230/GPU/hr over 11 days of tracking. Snapshots run daily; the full history is public in the repo.
V100 SXM2 PROVIDER NOTES
RunPod · 1 config · 1 region · 1 in stockSets the panel floor for this model.
from $0.230/hrCloudRift · 1 configWithin 15% of the panel floor.
from $0.250/hrBUYING V100 SXM2: WHAT THE PANEL SAYS
At the current floor, one V100 SXM2 running around the clock costs about $168 a month (730 hours of on-demand arithmetic). Compare the alternatives below before committing - the same budget often buys more than one chip class.
V100 SXM2 VS THE ALTERNATIVES
The A100 triples memory bandwidth, adds BF16 and MIG, and delivers about 2.5x the FP16 tensor throughput. Panel prices often put the A100 within reach of V100 money - check the spread before renting Volta.
A T4 is far cheaper and newer, but has one-fifth the memory bandwidth and no NVLink. For training, the V100 wins; for light serving, the T4 wins on cost.
Compare live rates:
V100 SXM2 PRICING FAQ
What is the cheapest V100 SXM2 cloud GPU right now?
The cheapest published V100 SXM2 rate on the panel is $0.230 per GPU-hour at RunPod (V100 SXM2 x 1, Global). Every price links to the provider's own published rate.
How much does a V100 SXM2 cost per hour?
Across 2 providers tracked today, published per-GPU hourly rates for V100 SXM2 run from $0.23 to $0.25, with a median of $0.25. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent V100 SXM2 GPUs?
2 providers publish 2 distinct V100 SXM2 configurations on Compute Cafe. 1 of those rows were reported in stock by the provider's live endpoint at last observation.
Are V100 SXM2 prices going up or down?
Over the 11 days Compute Cafe has tracked the panel, the V100 SXM2 floor is holding flat: $0.230 to $0.230/GPU/hr (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a V100 SXM2 cost per month?
At the current floor of $0.230/GPU/hr, a V100 SXM2 running 24/7 costs about $168 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
Is the V100 obsolete?
For frontier model training, yes. For fine-tuning mid-size models, batch jobs and older HPC codes it remains serviceable, and it is often the cheapest HBM GPU on the panel. Its 32 GB variant matters more than the 16 GB one today.
See also alternatives to the V100 SXM2 · cheapest GPU-hour per model · inference chips · training chips · by location