AWS vs Nebius GPU prices per hour

AWS and Nebius both list 5 GPU models with a public on-demand price in our data (H100, H200, B200, L40S, B300). AWS is cheaper on 0 of 5, Nebius is cheaper on 5. Biggest gap: B300, $8.30 per GPU-hour (47%) less at Nebius. Lowest price per GPU-hour at each provider, checked .

GPUs both list
5
Public on-demand price
AWS cheaper
0
of 5 shared GPUs
Nebius cheaper
5
of 5 shared GPUs
Last checked
Nov 9, 2026
Date read

Same GPU, two clouds

AWS and Nebius on-demand price for each GPU they both list

US dollars per GPU-hour from public pages. Lowest high or medium confidence on-demand row at each provider. Memory-matched: A100 means the 80 GB model, and H100 NVL 94 GB or B300 268 GB listings are left out. Node prices are divided per GPU. The market median is across every provider we track.

AWS and Nebius lowest public on-demand price per GPU-hour, with sources
GPUAWSNebiusDifferenceCheaperMarket median
H100$6.8880GB, 1xSource$4.50HGXSource$2.38 35% of the higher priceNebius$3.99 17 providers
H200$7.91SXM 141GB, 8x, 8-GPU nodeSource$5.40HGXSource$2.51 32% of the higher priceNebius$5.35 12 providers
B200$14.24SXM 8x, 8-GPU nodeSource$8.50HGXSource$5.74 40% of the higher priceNebius$7.99 10 providers
L40S$1.8648GB, 1xSource$1.55Intel CPUSource$0.31 17% of the higher priceNebius$1.70 10 providers
B300$17.80SXM 8x, 8-GPU nodeSource$9.50HGXSource$8.30 47% of the higher priceNebius$9.25 8 providers
How to read this. Price only. Same GPU name can mean a different variant, node size, region or bundle of CPU, memory and storage. Open both source links before you choose.

Only one of them

GPUs only one provider lists on demand

Only AWS

Only Nebius

None in our data.

FAQ

AWS vs Nebius questions

Is AWS cheaper than Nebius?

Not on most of them. Nebius is cheaper on 5 of 5 shared GPU models (H100, H200, B200, L40S, B300), comparing the lowest high or medium confidence public on-demand price per GPU-hour at each: H100 $6.88 at AWS against $4.50 at Nebius; H200 $7.91 at AWS against $5.40 at Nebius; B200 $14.24 at AWS against $8.50 at Nebius; L40S $1.86 at AWS against $1.55 at Nebius; B300 $17.80 at AWS against $9.50 at Nebius. Checked Nov 9, 2026.

Which is cheaper for the H100, AWS or Nebius?

Nebius. Its lowest public on-demand H100 price is $4.50 per GPU-hour against $6.88, 35% less. The market median across every provider we track is $3.99.

Where do these AWS and Nebius prices come from?

From each provider's own public pricing page or official price file, read without logging in. For each GPU we take the lowest high or medium confidence on-demand row per provider. Every price on this page links to its source, and the data was last checked Nov 9, 2026.

Is the hourly price the whole story?

No. Billing increments, storage, network egress, region, minimum node size and whether capacity is actually free all change the real cost. Some listings here are whole nodes: AWS H200 is sold as a 8-GPU node, AWS B200 is sold as a 8-GPU node, AWS B300 is sold as a 8-GPU node, so the hourly bill is that many GPUs. Open the source page for the exact plan before you buy.

Choosing between AWS and Nebius?

Tell us the GPU, count, term, region and budget. We reply with the public prices that match across every provider we track and the checks to run before you pay. Free, and we do not sell compute.