Model provider cost

Host DeepSeek-R1 on AWS

DeepSeek-R1 needs 8x 80GB+ GPUs, and AWS's current cheapest qualifying row is A100 SXM4 at $24.70/hr.

8x 80GB+ GPUs 671B params Frontier, cluster-only MIT
Cheapest on this provider
$24.70/hr
A100 SXM4
Monthly estimate
$18,033/mo
730 hours at the current median
VRAM baseline
80GB
8x 80GB+ GPUs
Qualifying rows
3
Updated Jun 21, 2026

AWS rows that can host DeepSeek-R1

The cheapest tracked way to host DeepSeek-R1 on AWS is A100 SXM4 at $24.70/hr. The overall tracked market floor is $8.01/hr on Vast.ai, so AWS is $16.69/hr above the current floor.

GPU VRAM Per GPU Estimated hourly Estimated monthly Updated
A100 SXM4 80GB $3.09/hr $24.70/hr $18,033/mo Jun 21, 2026
H100 SXM 80GB $6.88/hr $55.04/hr $40,179/mo Jun 21, 2026
H200 141GB $7.91/hr $63.30/hr $46,206/mo Jun 21, 2026

Why this setup does or does not fit

VRAM floor

8x 80GB+ GPUs

8x 80GB GPUs or larger, plus fast interconnects. Long prompts, batching, and KV cache can require extra headroom.

Model quality

Frontier, cluster-only

Potentially exceptional reasoning quality, but this is an engineering program more than a plug-in model choice.

Operational note

DeepSeek 671B

This is no longer a single-box deployment problem; networking, scheduling, and batching dominate.

DeepSeek-R1 on AWS FAQ

Can I host DeepSeek-R1 on AWS?

The cheapest tracked way to host DeepSeek-R1 on AWS is A100 SXM4 at $24.70/hr.

What GPU memory does DeepSeek-R1 need?

Our baseline for DeepSeek-R1 is 8x 80GB+ GPUs. The practical recommendation is 8x 80GB GPUs or larger, plus fast interconnects.

Is AWS the cheapest provider for DeepSeek-R1?

The overall tracked market floor is $8.01/hr on Vast.ai, so AWS is $16.69/hr above the current floor.

How fresh is this AWS DeepSeek-R1 cost page?

This page recalculates from the latest tracked on-demand rows. The freshest qualifying AWS row shown here is from Jun 21, 2026.

Compare this setup