GPU MODELS / A100 PCIE

A100 PCIe cloud GPU pricing

6 listings across 5 providers. Cheapest published rate: $1.150/GPU/hr at Denvr Dataworks. Sorted by per-GPU hourly price. Every row links to the provider's own published price.

WHAT IS THE A100 PCIE

The PCIe A100 brings Ampere tensor cores to standard servers at 250 W. Many providers list it simply as "A100 PCIe" without saying 40 or 80 GB - two very different products at one ambiguous name, and the first thing to ask before renting. Either variant trains well; neither has the SXM part's NVLink fabric.

MAKER
NVIDIA
ARCHITECTURE
Ampere
LAUNCHED
2021
MEMORY
40 or 80 GB HBM2e (providers often list without the variant)
MEMORY BANDWIDTH
1,555-1,935 GB/s depending on variant
TENSOR COMPUTE
312 TFLOPS FP16/BF16 tensor dense (624 sparse)
POWER DRAW
250 W
INTERCONNECT
PCIe 4.0 x16 (optional NVLink bridge)

WHAT IT IS GOOD AT

Single-GPU fine-tuning
The 80 GB variant handles QLoRA on 70B-class models; the 40 GB variant is comfortable to ~13B full fine-tune.
Inference serving
Strong single-card serving for models that fit; PCIe bandwidth is no constraint at inference.
Multi-GPU training
Weaker fit - no NVLink fabric, so all-reduce runs over PCIe unless the provider bridges the cards. Ask.

WHO SHOULD NOT RENT THE A100 PCIE

Multi-GPU training at scale - SXM A100/H100 nodes with NVLink fabric win clearly.
Any rental where the provider will not state the memory variant.
ProviderConfigVRAMLocation$/hr$/GPU/hrSource
Denvr DataworksDenvr DataworksA100 PCIe × 440 GBCanada$4.600$1.150receipt
RunPodRunPodA100 PCIe × 180 GBGlobal$1.190$1.190receipt
VultrVultrA100 PCIe (bare metal) × 480 GBNot published$5.160$1.290receipt
Lambda LabsLambda LabsA100 PCIe × 140 GBUS$1.990$1.990receipt
CrusoeCrusoeA100 PCIe × 180 GBUS / Iceland$2.000$2.000receipt
VultrVultrA100 PCIe × 180 GBNot published$2.397$2.397receipt

THE A100 PCIE MARKET, BY THE NUMBERS

PROVIDERS
5
CONFIGURATIONS
6
MEDIAN $/GPU/HR
$1.99
SPREAD
2.1x

5 providers publish 6 A100 PCIe configurations on the panel today. Published per-GPU hourly rates run from $1.150 (Denvr Dataworks) to $2.40 (Vultr), a 2x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $1.99/GPU/hr. 3 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.

CHEAPEST A100 PCIE BY GEOGRAPHY
canadaDenvr Dataworks$1.150/hrunited statesLambda Labs$1.990/hricelandCrusoe$2.000/hr
A100 PCIE PRICE TREND, DAILY PANEL SNAPSHOTS

Across every A100 PCIe row on the panel, the daily floor held flat at $1.150/GPU/hr over 11 days of tracking. The median continuously-tracked listing held flat at $1.99. Snapshots run daily; the full history is public in the repo.

A100 PCIE PROVIDER NOTES
Denvr Dataworks · 1 config · 1 region
Sets the panel floor for this model.
from $1.150/hr
RunPod · 1 config · 1 region · 1 in stock
Within 15% of the panel floor.
from $1.190/hr
Vultr · 2 configs · 2 in stock
Within 15% of the panel floor.
from $1.290/hr
Lambda Labs · 1 config · 1 region
Published catalog rates.
from $1.990/hr
Crusoe · 1 config · 1 region
Published catalog rates.
from $2.000/hr
BUYING A100 PCIE: WHAT THE PANEL SAYS

At the current floor, one A100 PCIe running around the clock costs about $839 a month (730 hours of on-demand arithmetic). Per-GPU floors by config size: 1x $1.19 · 4x $1.15. Compare the alternatives below before committing - the same budget often buys more than one chip class.

A100 PCIE VS THE ALTERNATIVES

A100 PCIe vs A100 80GB SXM
The SXM part adds NVLink 3.0 at 600 GB/s and higher power limits. Same per-card compute; the fabric is what you pay for.
A100 PCIe vs H100 PCIe
The H100 PCIe roughly doubles dense FP16 throughput at a higher hourly rate and the same interconnect limits. For single-GPU work it is the faster option; neither links well.

Compare live rates: A100 PCIe vs A100 80GB SXM · A100 PCIe vs H100 PCIe

A100 PCIE PRICING FAQ

What is the cheapest A100 PCIe cloud GPU right now?
The cheapest published A100 PCIe rate on the panel is $1.150 per GPU-hour at Denvr Dataworks (A100 PCIe x 4, Canada). Every price links to the provider's own published rate.
How much does a A100 PCIe cost per hour?
Across 5 providers tracked today, published per-GPU hourly rates for A100 PCIe run from $1.15 to $2.40, with a median of $1.99. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent A100 PCIe GPUs?
5 providers publish 6 distinct A100 PCIe configurations on Compute Cafe. 3 of those rows were reported in stock by the provider's live endpoint at last observation.
Where is A100 PCIe cheapest?
By published per-GPU hourly rate, the cheapest geography for A100 PCIe today is canada at $1.15/hr (Denvr Dataworks). The spread across 3 tracked geographies runs to $2.00/hr in iceland.
Are A100 PCIe prices going up or down?
Over the 11 days Compute Cafe has tracked the panel, the A100 PCIe floor is holding flat: $1.150 to $1.150/GPU/hr (+0.0%). The median continuously-tracked listing moved from $1.99 to $1.99 (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a A100 PCIe cost per month?
At the current floor of $1.150/GPU/hr, a A100 PCIe running 24/7 costs about $839 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 an "A100 PCIe" 40 GB or 80 GB?
Either - the market sells both under that name. The panel row links the provider's page; confirm the variant there before comparing prices, because an 80 GB card at a small premium is usually the better buy.
Is the PCIe A100 good for LLM serving?
Yes for single-model serving that fits in its memory. Multi-GPU tensor parallelism runs over PCIe, which is slow - use one big card or an SXM node instead.

See also alternatives to the A100 PCIe · cheapest GPU-hour per model · inference chips · training chips · by location