Skip to content

NVIDIA A100

The Ampere datacenter GPU the H100 replaced. No FP8, but it still runs most training and inference jobs.

The lowest price you can rent right now is $1.35/GPU/hr, at Hyperstack in CANADA-1. That price comes and goes. The $2.40/GPU/hr one at Vultr was in stock 96% of the last 30 days.

Cheapest A100 in stock right now

As of Nov 16, 22:30 UTC

$1.35/GPU/hr

$0.25 (16%) belowthe 7-day median of $1.60

  • Low $1.35
  • Median $1.60
  • High $2.19
Line chart with 168 data points. Use the left and right arrow keys to move between data points, Page Up and Page Down to jump by a tenth of the data, Home and End to jump to the first and last, Enter or Space to activate the current point, and Escape to dismiss the readout.

$1.35 at Hyperstack was in stock 36% of the last 7 days

Rent it at Hyperstack n3-A100x1 in CANADA-1
  • Ampere
  • 40 GB HBM2 or 80 GB HBM2e
  • 7 providers

A100 prices start at $1.35/GPU/hr. Plan on $2.40/GPU/hr to get one without waiting.

Hyperstack's $1.35/GPU/hr instance was in stock 70% of the last 30 days. Vultr's $2.40/GPU/hr instance was in stock 96% of that time. Spot starts at $1.08/GPU/hr.

How we pick these prices · What availability means

Provider by provider

Each provider's lowest on-demand A100 price, and how often that instance was in stock over the last 30 days.

ProviderLowest priceInstanceRegionIn stock, last 30 daysLowest in stock at last scan
Hyperstack$1.35/GPU/hrn3-A100x1CANADA-170%$1.35/GPU/hr
Lambda$1.99/GPU/hr1x A100 (40 GB PCIe)us-south-213%$1.99/GPU/hr
RunPod$2.09/GPU/hrA100 PCIeCA-MTL-3Only reports in stockSold out
Vultr$2.40/GPU/hrvcg-a100-12c-120g-80vramewr96%$2.40/GPU/hr
AWS$2.74/GPU/hrp4d.24xlargeus-east-2Not publishedNot published
Oracle Cloud$3.05/GPU/hrBM.GPU4.8globalNot publishedNot published
Azure$3.40/GPU/hrStandard_ND96asr_v4eastusNot publishedNot published
Google Cloud$3.48/GPU/hra2-megagpu-16gus-central1Not publishedNot published

AWS, Oracle Cloud, Azure, and Google Cloud list it from $2.74/GPU/hr, 2.0× the lowest price here. That's the published on-demand rate, not a scan error: reserved capacity costs less, and spot starts at $0.55/GPU/hr at AWS. Neither reports stock.

How often each provider had one in stock

Line chart with 719 data points. Use the left and right arrow keys to move between data points, Page Up and Page Down to jump by a tenth of the data, Home and End to jump to the first and last, Enter or Space to activate the current point, and Escape to dismiss the readout.

What the NVIDIA A100 is

The Ampere datacenter GPU that the H100 replaced. It has no FP8 support, so it trails the H100 by a wide margin on transformer work, but it still runs most training and inference jobs well. Providers sell 40 GB and 80 GB versions.

Architecture
Ampere
Memory
40 GB HBM2 or 80 GB HBM2e
Memory bandwidth
2 TB/s
BF16 compute
312 TFLOPS, dense
Lowest precision
BF16
For rent here as
PCIe, PCIe NVLink, and SXM

Other GPUs to consider

Head to head:A100 vs H100A100 vs L40SA100 vs RTX A6000A100 vs RTX PRO 6000

238 NVIDIA A100 instances

Availability
n3-A100x1Hyperstack · CANADA-1
$1.35/GPU/hr
—
available
n3-A100x2Hyperstack · CANADA-1
$1.35/GPU/hr
$2.70/hr total
available
8x A100 (40 GB SXM4)Lambda · asia-northeast-2
$1.99/GPU/hr
$15.92/hr total
available
1x A100 (40 GB SXM4)Lambda · asia-south-1
$1.99/GPU/hr
—
available
8x A100 (40 GB SXM4)Lambda · europe-central-1
$1.99/GPU/hr
$15.92/hr total
available
8x A100 (40 GB SXM4)Lambda · me-west-1
$1.99/GPU/hr
$15.92/hr total
available
1x A100 (40 GB SXM4)Lambda · us-east-1
$1.99/GPU/hr
—
available
1x A100 (40 GB SXM4)Lambda · us-west-2
$1.99/GPU/hr
—
available
vcg-a100-12c-120g-80vramVultr · ewr
$2.40/GPU/hr
$1,750/mo
available
p4d.24xlargeAWS · us-east-2
$2.74/GPU/hr
$21.96/hr total
not reported
p4d.24xlargeAWS · us-west-2
$2.74/GPU/hr
$21.96/hr total
not reported
p4d.24xlargeAWS · us-east-1
$2.74/GPU/hr
$21.96/hr total
not reported
p4d.24xlargeAWS · eu-north-1
$2.96/GPU/hr
$23.72/hr total
not reported
p4d.24xlargeAWS · eu-west-1
$2.96/GPU/hr
$23.72/hr total
not reported
BM.GPU4.8Oracle Cloud · global
$3.05/GPU/hr
$24.40/hr total
not reported
p4d.24xlargeAWS · ca-central-1
$3.16/GPU/hr
$25.25/hr total
not reported
p4d.24xlargeAWS · ap-south-1
$3.29/GPU/hr
$26.34/hr total
not reported
Standard_ND96asr_v4Azure · eastus
$3.40/GPU/hr
$27.20/hr total
not reported
Standard_ND96asr_v4Azure · westus2
$3.40/GPU/hr
$27.20/hr total
not reported
p4d.24xlargeAWS · eu-central-1
$3.43/GPU/hr
$27.43/hr total
not reported
p4de.24xlargeAWS · us-east-1
$3.43/GPU/hr
$27.45/hr total
not reported
p4de.24xlargeAWS · us-west-2
$3.43/GPU/hr
$27.45/hr total
not reported
a2-megagpu-16gGoogle Cloud · us-central1
$3.48/GPU/hr
$55.74/hr total
not reported
a2-megagpu-16gGoogle Cloud · europe-west4
$3.54/GPU/hr
$56.63/hr total
not reported
p4d.24xlargeAWS · ap-southeast-2
$3.57/GPU/hr
$28.54/hr total
not reported
p4d.24xlargeAWS · eu-west-2
$3.57/GPU/hr
$28.54/hr total
not reported
Standard_NC24ads_A100_v4Azure · eastus
$3.67/GPU/hr
—
not reported
Standard_NC48ads_A100_v4Azure · eastus
$3.67/GPU/hr
$7.35/hr total
not reported
Standard_NC96ads_A100_v4Azure · eastus
$3.67/GPU/hr
$14.69/hr total
not reported
Standard_NC24ads_A100_v4Azure · eastus2
$3.67/GPU/hr
—
not reported
Standard_NC48ads_A100_v4Azure · eastus2
$3.67/GPU/hr
$7.35/hr total
not reported
Standard_NC96ads_A100_v4Azure · eastus2
$3.67/GPU/hr
$14.69/hr total
not reported
Standard_NC24ads_A100_v4Azure · westus2
$3.67/GPU/hr
—
not reported
Standard_NC48ads_A100_v4Azure · westus2
$3.67/GPU/hr
$7.35/hr total
not reported
Standard_NC96ads_A100_v4Azure · westus2
$3.67/GPU/hr
$14.69/hr total
not reported
Standard_NC24ads_A100_v4Azure · westus3
$3.67/GPU/hr
—
not reported
Standard_NC48ads_A100_v4Azure · westus3
$3.67/GPU/hr
$7.35/hr total
not reported
Standard_NC96ads_A100_v4Azure · westus3
$3.67/GPU/hr
$14.69/hr total
not reported
a2-highgpu-1gGoogle Cloud · us-central1
$3.67/GPU/hr
—
not reported
a2-highgpu-2gGoogle Cloud · us-central1
$3.67/GPU/hr
$7.35/hr total
not reported
a2-highgpu-4gGoogle Cloud · us-central1
$3.67/GPU/hr
$14.69/hr total
not reported
a2-highgpu-8gGoogle Cloud · us-central1
$3.67/GPU/hr
$29.39/hr total
not reported
a2-highgpu-1gGoogle Cloud · us-east1
$3.67/GPU/hr
—
not reported
a2-highgpu-2gGoogle Cloud · us-east1
$3.67/GPU/hr
$7.35/hr total
not reported
a2-highgpu-4gGoogle Cloud · us-east1
$3.67/GPU/hr
$14.69/hr total
not reported
a2-highgpu-8gGoogle Cloud · us-east1
$3.67/GPU/hr
$29.39/hr total
not reported
a2-highgpu-1gGoogle Cloud · us-west1
$3.67/GPU/hr
—
not reported
a2-highgpu-2gGoogle Cloud · us-west1
$3.67/GPU/hr
$7.35/hr total
not reported
a2-highgpu-4gGoogle Cloud · us-west1
$3.67/GPU/hr
$14.69/hr total
not reported
a2-highgpu-8gGoogle Cloud · us-west1
$3.67/GPU/hr
$29.39/hr total
not reported
Page 1 of 5

Agents can read the Markdown version, which lists every current instance.