A prolonged crunch in Nvidia GPU capacity at Amazon Web Services is handing specialist cloud providers such as Together AI, Runpod and Nebius an opening to win over the next wave of AI startups , according to a report published this week by The Information .
GPU Cloud Wars · AI Infrastructure
AWS Chip Crunch Opens the Door for Specialist GPU Clouds
A prolonged shortage of Nvidia H100/H200 capacity at AWS — plus a fresh price hike — is pushing cost-sensitive AI startups toward challengers like Together AI, Runpod and Nebius, competing on price and instant availability.
~20%
AWS Nvidia compute price rise in mid-2026 — its second increase
10×+
Spread in H100 hourly rates across providers
20–60%
Cheaper than AWS on some alternative clouds
1M+
Nvidia GPUs AWS plans to add in 2026
H100 SXM — on-demand price per GPU-hour
Column height scales to price · lower is cheaper
~$2.69
RunpodSecure Cloud
10×+
high
AWStop of range
Why the window is open
Nvidia H100/H200 demand keeps outrunning supply into 2026 — so hyperscalers ration capacity to their own workloads and large enterprises.
Tight supply
Rising cost
Instant access
✓ Why startups switch
Lower cost, immediate access
Fewer out-of-stock moments
Newest GPUs (H200, B200) available
Per-second billing, cluster + SLA options
△ The trade-offs
Spot / shared tiers carry interruption risk
Cheaper tiers can be less reliable
AWS keeps edge on enterprise stability & SLAs
Challengers can over-provision very small users
The stakes: If AWS can't offer new startups immediate, affordable capacity, more of the next generation of AI companies may make their first infrastructure decisions — and build lasting habits — on rival platforms.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…