Breaking

Amazon Web Services GPU Shortage Opens Door for Specialist Clouds Together, Runpod and Nebius

July 17, 2026 at 14:03 EDT

Infra & Chips
Funding & M&A

A prolonged crunch in NVIDIA GPU capacity at Amazon Web Services is pushing a new generation of AI startups toward specialist cloud providers such as Together AI, Runpod and Nebius, which are undercutting the hyperscaler on price and provisioning speed.

The GPU Supply Supercycle

AWS GPU Crunch Sends AI Startups to Specialist Clouds

A prolonged shortage of NVIDIA accelerators at AWS is pushing a new generation of AI builders toward GPU-native providers — Together AI, Runpod and Nebius — that undercut the hyperscaler on price and provision capacity faster.

30–60%

Specialist pricing cited below AWS rates

57%

Nebius cited cheaper on H100 capacity

5 GW

NVIDIA systems Nebius commits to by 2030

Cited H100 rates — cost per hour

Column height is proportional to the hourly rate. Lower is cheaper.

$5.49

Together AI

HGX H100 · on-demand

$4.55

Together AI

H100 · reserved

$3.29

Runpod

H100 SXM · Secure

$2.69

Runpod

H100 SXM · Community

Why demand is redistributing

Tight NVIDIA supply

H100 · H200 · B200 scarce at AWS

→

Startups blocked

Cannot reserve capacity blocks

→

Specialist clouds win

Cheaper, faster provisioning

The draw

Rates 30–60% below AWS
Fast, lean GPU provisioning
Per-second billing, 30+ GPU types
Early access to latest NVIDIA silicon

The friction

Harder cost allocation across vendors
Differing service-level agreements
Higher operational overhead
Management complexity off the hyperscaler

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.