Skip to content
Market report · Live prices

GPU cloud prices

Every GPU machine in every region we track, from our own price index. Each price is the median over the last 24 hours across the region’s zones.

Updated 7 October 2026, 03:54 UTC. AWS: 41 machine types in 23 regions; Google Cloud: 23 machine types in 32 regions.

896 machines priced on spot. Clear filters

MachineRegionGPUsvCPU / RAMSpot / hrSpot / GPU-hrSpot per GB / hrSpot discountSpot vs last weekDaily swing
g5.48xlarge
AWS
Paris (eu-west-3)8× A10G 24 GB192 / 768 GB$20.676$2.584$0.10770%0%0.0%
a2-highgpu-1g
Google Cloud
us-west31× A100 40 GB12 / 85 GB$2.588$2.588$0.0647—0%0.0%
a2-highgpu-2g
Google Cloud
us-west32× A100 40 GB24 / 170 GB$5.177$2.588$0.0647—0%0.0%
a2-highgpu-4g
Google Cloud
us-west34× A100 40 GB48 / 340 GB$10.354$2.588$0.0647—0%0.0%
a2-highgpu-8g
Google Cloud
us-west38× A100 40 GB96 / 680 GB$20.708$2.588$0.0647—0%0.0%
g6e.8xlarge
AWS
Mumbai (ap-south-1)1× L40S 48 GB32 / 256 GB$2.600$2.600$0.054252%+65%16.6%
p5.48xlarge
AWS
N. California (us-west-1)8× H100 80 GB192 / 2048 GB$21.134$2.642$0.033069%+2%2.7%
g4-standard-192
Google Cloud
asia-southeast24× RTX PRO 6000 96 GB192 / 720 GB$10.580$2.645$0.0276—−5%2.5%
g4-standard-384
Google Cloud
asia-southeast28× RTX PRO 6000 96 GB384 / 1440 GB$21.161$2.645$0.0276—−5%2.5%
g4-standard-48
Google Cloud
asia-southeast21× RTX PRO 6000 96 GB48 / 180 GB$2.645$2.645$0.0276—−5%2.5%
g4-standard-96
Google Cloud
asia-southeast22× RTX PRO 6000 96 GB96 / 360 GB$5.290$2.645$0.0276—−5%2.5%
g5.48xlarge
AWS
Sydney (ap-southeast-2)8× A10G 24 GB192 / 768 GB$21.178$2.647$0.11030%0%0.0%
g5.24xlarge
AWS
Tokyo (ap-northeast-1)4× A10G 24 GB96 / 384 GB$10.630$2.658$0.110710%0%0.0%
g5.48xlarge
AWS
Tokyo (ap-northeast-1)8× A10G 24 GB192 / 768 GB$21.260$2.658$0.110710%0%0.0%
g4dn.16xlarge
AWS
Tokyo (ap-northeast-1)1× T4 16 GB64 / 256 GB$2.658$2.658$0.166155%0%0.1%
g6e.48xlarge
AWS
Tokyo (ap-northeast-1)8× L40S 48 GB192 / 1536 GB$21.328$2.666$0.055551%0%0.3%
g6e.8xlarge
AWS
Oregon (us-west-2)1× L40S 48 GB32 / 256 GB$2.677$2.677$0.055841%−9%6.0%
p5.48xlarge
AWS
London (eu-west-2)8× H100 80 GB192 / 2048 GB$21.492$2.687$0.033670%+10%3.9%
g6e.8xlarge
AWS
Seoul (ap-northeast-2)1× L40S 48 GB32 / 256 GB$2.687$2.687$0.056052%+3%2.1%
g4-standard-192
Google Cloud
asia-south24× RTX PRO 6000 96 GB192 / 720 GB$10.760$2.690$0.0280—−3%1.7%
g4-standard-384
Google Cloud
asia-south28× RTX PRO 6000 96 GB384 / 1440 GB$21.520$2.690$0.0280—−3%1.7%
g4-standard-48
Google Cloud
asia-south21× RTX PRO 6000 96 GB48 / 180 GB$2.690$2.690$0.0280—−3%1.7%
g4-standard-96
Google Cloud
asia-south22× RTX PRO 6000 96 GB96 / 360 GB$5.380$2.690$0.0280—−3%1.7%
g5.16xlarge
AWS
Montréal (ca-central-1)1× A10G 24 GB64 / 256 GB$2.732$2.732$0.113940%+35%9.7%
g6e.12xlarge
AWS
eu-south-24× L40S 48 GB48 / 384 GB$10.936$2.734$0.05701%—57.4%
g4-standard-192
Google Cloud
asia-east14× RTX PRO 6000 96 GB192 / 720 GB$10.943$2.736$0.0285—0%0.0%
g4-standard-384
Google Cloud
asia-east18× RTX PRO 6000 96 GB384 / 1440 GB$21.886$2.736$0.0285—0%0.0%
g4-standard-48
Google Cloud
asia-east11× RTX PRO 6000 96 GB48 / 180 GB$2.736$2.736$0.0285—0%0.0%
g4-standard-96
Google Cloud
asia-east12× RTX PRO 6000 96 GB96 / 360 GB$5.471$2.736$0.0285—0%0.0%
g6e.8xlarge
AWS
Osaka (ap-northeast-3)1× L40S 48 GB32 / 256 GB$2.816$2.816$0.058757%+40%14.1%
p5.48xlarge
AWS
Ohio (us-east-2)8× H100 80 GB192 / 2048 GB$22.559$2.820$0.035259%+11%4.4%
g6e.24xlarge
AWS
Mumbai (ap-south-1)4× L40S 48 GB96 / 768 GB$11.307$2.827$0.058938%−37%24.8%
p5.48xlarge
AWS
Mumbai (ap-south-1)8× H100 80 GB192 / 2048 GB$22.672$2.834$0.035466%+18%5.8%
g5.8xlarge
AWS
Mumbai (ap-south-1)1× A10G 24 GB32 / 128 GB$2.901$2.901$0.12091%+159%41.5%
a2-ultragpu-1g
Google Cloud
europe-west41× A100 80 GB12 / 170 GB$2.912$2.912$0.0364—+6%2.9%
a2-ultragpu-8g
Google Cloud
europe-west48× A100 80 GB96 / 1360 GB$23.295$2.912$0.0364—+6%2.9%
a2-ultragpu-2g
Google Cloud
europe-west42× A100 80 GB24 / 340 GB$5.824$2.912$0.0364—+6%2.9%
a2-ultragpu-4g
Google Cloud
europe-west44× A100 80 GB48 / 680 GB$11.647$2.912$0.0364—+6%2.9%
g5.48xlarge
AWS
São Paulo (sa-east-1)8× A10G 24 GB192 / 768 GB$23.563$2.945$0.122715%−1%0.3%
g6e.48xlarge
AWS
N. Virginia (us-east-1)8× L40S 48 GB192 / 1536 GB$23.780$2.973$0.061921%+55%9.4%
g6e.4xlarge
AWS
Oregon (us-west-2)1× L40S 48 GB16 / 128 GB$3.004$3.004$0.06260%+46%15.7%
a2-ultragpu-1g
Google Cloud
us-central11× A100 80 GB12 / 170 GB$3.017$3.017$0.0377—0%0.0%
a2-ultragpu-2g
Google Cloud
us-central12× A100 80 GB24 / 340 GB$6.033$3.017$0.0377—0%0.0%
a2-ultragpu-4g
Google Cloud
us-central14× A100 80 GB48 / 680 GB$12.066$3.017$0.0377—0%0.0%
a2-ultragpu-8g
Google Cloud
us-central18× A100 80 GB96 / 1360 GB$24.133$3.017$0.0377—0%0.0%
p4de.24xlarge
AWS
Frankfurt (eu-central-1)8× A100 80 GB96 / 1152 GB$24.272$3.034$0.037929%−28%11.1%
g6e.24xlarge
AWS
N. Virginia (us-east-1)4× L40S 48 GB96 / 768 GB$12.190$3.047$0.063519%+4%19.7%
p5.4xlarge
AWS
Ohio (us-east-2)1× H100 80 GB16 / 256 GB$3.087$3.087$0.038655%+26%5.7%
g5.8xlarge
AWS
Paris (eu-west-3)1× A10G 24 GB32 / 128 GB$3.107$3.107$0.12950%0%74.6%
p5.48xlarge
AWS
Tokyo (ap-northeast-1)8× H100 80 GB192 / 2048 GB$24.992$3.124$0.039164%+10%4.0%
g6e.8xlarge
AWS
Tokyo (ap-northeast-1)1× L40S 48 GB32 / 256 GB$3.137$3.137$0.065452%−4%2.6%
g6e.48xlarge
AWS
Oregon (us-west-2)8× L40S 48 GB192 / 1536 GB$25.583$3.198$0.066615%−3%23.1%
g5.16xlarge
AWS
London (eu-west-2)1× A10G 24 GB64 / 256 GB$3.206$3.206$0.133638%−5%50.3%
p5.4xlarge
AWS
Mumbai (ap-south-1)1× H100 80 GB16 / 256 GB$3.277$3.277$0.041060%+32%10.0%
g6e.12xlarge
AWS
Frankfurt (eu-central-1)4× L40S 48 GB48 / 384 GB$13.120$3.280$0.06830%0%0.0%
p4d.24xlarge
AWS
Mumbai (ap-south-1)8× A100 40 GB96 / 1152 GB$26.344$3.293$0.08230%0%0.0%
g6e.16xlarge
AWS
Frankfurt (eu-central-1)1× L40S 48 GB64 / 512 GB$3.329$3.329$0.069365%−2%1.0%
g5.16xlarge
AWS
Tokyo (ap-northeast-1)1× A10G 24 GB64 / 256 GB$3.358$3.358$0.139944%+9%2.4%
p5.48xlarge
AWS
Oregon (us-west-2)8× H100 80 GB192 / 2048 GB$27.148$3.393$0.042451%+31%8.8%
p4d.24xlarge
AWS
Frankfurt (eu-central-1)8× A100 40 GB96 / 1152 GB$27.433$3.429$0.08570%0%0.0%
p5.48xlarge
AWS
Sydney (ap-southeast-2)8× H100 80 GB192 / 2048 GB$27.476$3.434$0.042962%−20%7.6%
p5.48xlarge
AWS
N. Virginia (us-east-1)8× H100 80 GB192 / 2048 GB$27.601$3.450$0.043150%+32%10.9%
g6e.16xlarge
AWS
Oregon (us-west-2)1× L40S 48 GB64 / 512 GB$3.466$3.466$0.072254%+4%3.2%
p5.4xlarge
AWS
N. Virginia (us-east-1)1× H100 80 GB16 / 256 GB$3.538$3.538$0.044249%+35%10.7%
g5.8xlarge
AWS
São Paulo (sa-east-1)1× A10G 24 GB32 / 128 GB$3.539$3.539$0.147515%+232%54.0%
p4d.24xlarge
AWS
London (eu-west-2)8× A100 40 GB96 / 1152 GB$28.545$3.568$0.08920%+19%17.1%
a2-ultragpu-1g
Google Cloud
asia-southeast11× A100 80 GB12 / 170 GB$3.592$3.592$0.0449—+4%1.8%
a2-ultragpu-8g
Google Cloud
asia-southeast18× A100 80 GB96 / 1360 GB$28.739$3.592$0.0449—+4%1.8%
a2-ultragpu-2g
Google Cloud
asia-southeast12× A100 80 GB24 / 340 GB$7.185$3.592$0.0449—+4%1.8%
a2-ultragpu-4g
Google Cloud
asia-southeast14× A100 80 GB48 / 680 GB$14.369$3.592$0.0449—+4%1.8%
p5.4xlarge
AWS
Oregon (us-west-2)1× H100 80 GB16 / 256 GB$3.630$3.630$0.045447%+38%12.9%
g6e.16xlarge
AWS
Ohio (us-east-2)1× L40S 48 GB64 / 512 GB$3.740$3.740$0.077951%−1%0.3%
p5.4xlarge
AWS
Tokyo (ap-northeast-1)1× H100 80 GB16 / 256 GB$3.754$3.754$0.046956%+27%10.1%
g6e.24xlarge
AWS
Oregon (us-west-2)4× L40S 48 GB96 / 768 GB$15.066$3.766$0.07850%0%28.3%
g6.16xlarge
AWS
Montréal (ca-central-1)1× L4 24 GB64 / 256 GB$3.772$3.772$0.15720%0%0.0%
p4d.24xlarge
AWS
São Paulo (sa-east-1)8× A100 40 GB96 / 1152 GB$30.261$3.783$0.094615%−1%0.3%
p5.4xlarge
AWS
Sydney (ap-southeast-2)1× H100 80 GB16 / 256 GB$3.825$3.825$0.047857%+37%10.2%
p5.48xlarge
AWS
Seoul (ap-northeast-2)8× H100 80 GB192 / 2048 GB$31.282$3.910$0.048959%+38%10.5%
g6e.48xlarge
AWS
eu-south-28× L40S 48 GB192 / 1536 GB$31.379$3.922$0.08171%—46.0%
g6e.24xlarge
AWS
eu-south-24× L40S 48 GB96 / 768 GB$15.702$3.925$0.08181%—44.9%
g5.16xlarge
AWS
ap-southeast-31× A10G 24 GB64 / 256 GB$3.998$3.998$0.166630%—56.3%
p4de.24xlarge
AWS
Singapore (ap-southeast-1)8× A100 80 GB96 / 1152 GB$32.937$4.117$0.05150%0%0.0%
g5.16xlarge
AWS
eu-south-21× A10G 24 GB64 / 256 GB$4.269$4.269$0.17791%—0.1%
g6e.16xlarge
AWS
Seoul (ap-northeast-2)1× L40S 48 GB64 / 512 GB$4.271$4.271$0.089054%−1%0.2%
g6e.48xlarge
AWS
Mumbai (ap-south-1)8× L40S 48 GB192 / 1536 GB$36.187$4.523$0.09420%0%0.0%
g6e.48xlarge
AWS
Osaka (ap-northeast-3)8× L40S 48 GB192 / 1536 GB$37.075$4.634$0.096515%+41%37.8%
g6e.24xlarge
AWS
Frankfurt (eu-central-1)4× L40S 48 GB96 / 768 GB$18.838$4.710$0.09810%0%0.0%
g6e.48xlarge
AWS
Frankfurt (eu-central-1)8× L40S 48 GB192 / 1536 GB$37.676$4.710$0.09810%0%0.0%
g6e.16xlarge
AWS
Tokyo (ap-northeast-1)1× L40S 48 GB64 / 512 GB$4.978$4.978$0.103755%0%0.2%
g5.16xlarge
AWS
São Paulo (sa-east-1)1× A10G 24 GB64 / 256 GB$5.930$5.930$0.247115%−1%0.3%
g6e.16xlarge
AWS
Mumbai (ap-south-1)1× L40S 48 GB64 / 512 GB$6.396$6.396$0.133330%+201%35.8%
p5.4xlarge
AWS
ap-southeast-31× H100 80 GB16 / 256 GB$6.676$6.676$0.083430%—1.3%
p5.48xlarge
AWS
ap-southeast-38× H100 80 GB192 / 2048 GB$53.607$6.701$0.083829%—2.0%
p5.4xlarge
AWS
London (eu-west-2)1× H100 80 GB16 / 256 GB$8.944$8.944$0.11180%0%0.0%
p5.48xlarge
AWS
São Paulo (sa-east-1)8× H100 80 GB192 / 2048 GB$78.694$9.837$0.123015%−1%0.3%
p5.4xlarge
AWS
São Paulo (sa-east-1)1× H100 80 GB16 / 256 GB$9.837$9.837$0.123015%−1%0.3%

Reading the table

  • Prices are in US dollars per hour for the whole machine, and per GPU. The machine price includes its CPUs and memory, so a GPU on a large host costs more per GPU-hour.
  • Spot per GB divides the spot price by the machine’s total GPU memory: the fair comparison when what you need is room to fit a model.
  • Spot vs last week compares the last 24 hours with the same 24 hours a week earlier. Daily swing is how much the daily price varied over the last 7 days.
  • A dash means we have no price for it, not that it is free: some clouds publish spot prices only, and some regions sell a machine on one market only.
  • A low price is not a promise of capacity. Spot machines can be scarce in exactly the regions where they are cheapest.

Use the data

The whole table is free to use, licensed CC BY 4.0: chart it, build on it and republish it, with credit to Otium.

Get the next report

GPU prices every month, by GPU and by region. One email a month; we’ll ask you to confirm first.

Free. One-click unsubscribe in every issue.

Let the price come to you.

Otium is in closed alpha. Set a deadline for your batch AI work, and it waits for the lowest-cost compute available before then, so you don’t have to watch these prices yourself. Leave your email and we’ll send an invite when a spot opens.

No credit card, no spam — one email when your invite is ready.

Closed alpha — onboarding is gated while we calibrate. Already invited? Sign in.