What is the cheapest GPU cloud provider for H100 in 2026?

The cheapest H100 GPU cloud provider in 2026 is Lambda at $1.99/hr on-demand for H100 SXM5. Spot/marketplace pricing on Vast.ai reaches $1.49/hr but is interruptible. CoreWeave charges $2.23/hr on-demand and $1.79/hr reserved (1-year). AWS, Google Cloud, and Azure range from $3.90–$4.10/hr per GPU. GridStackHub tracks H100 pricing daily across 20+ providers.

How much does it cost to rent an H100 GPU per hour?

H100 GPU rental costs range from $1.49/hr (Vast.ai spot) to $4.10/hr (major cloud providers) as of April 2026. Lambda on-demand is $1.99/hr, CoreWeave is $2.23/hr, RunPod is $1.99/hr. AWS p4d instances run approximately $3.90/hr per GPU normalized. Reserved 1-year pricing on CoreWeave is $1.79/hr — roughly 20% below on-demand.

Which GPU cloud provider is cheapest overall in 2026?

For H100 GPUs: Lambda ($1.99/hr), Vast.ai spot ($1.49/hr), CoreWeave ($2.23/hr). For A100 80GB: Vast.ai ($0.42/hr spot), FluidStack ($1.21/hr), Lambda ($1.29/hr). For H200: Lambda ($2.99/hr), Crusoe ($3.15/hr). For B200: Lambda ($5.29/hr), CoreWeave ($5.49/hr). Independent cloud providers (Lambda, CoreWeave, RunPod, Vast.ai, FluidStack) are consistently 40–60% cheaper than AWS, Google Cloud, and Azure for equivalent GPU capacity.

What is the cheapest way to run an LLM in the cloud?

For small models (7B–13B): RTX 4090 or A100 40GB on Vast.ai or RunPod at $0.20–$0.50/hr. For mid-size models (30B–70B): A100 80GB at $0.42/hr on Vast.ai spot, or $1.29/hr on-demand at Lambda. For large models (70B–405B): H100 SXM5 at $1.99/hr on Lambda, or H200 at $2.99/hr. For frontier-size models needing max throughput: B200 at $5.29/hr. GridStackHub tracks pricing for 32 providers to find the cheapest option updated daily.

How much does CoreWeave charge for H100 GPUs?

CoreWeave charges $2.23/hr on-demand for H100 SXM5 as of April 2026. Their reserved 1-year pricing is $1.79/hr — approximately 20% below on-demand. CoreWeave also offers H100 PCIe at lower rates. For 8-GPU H100 SXM5 nodes, CoreWeave prices are generally competitive with Lambda while offering stronger SLA guarantees and dedicated cluster options.

Tracking 32 providers in real-time

The GPU cost intelligence
cloud vendors don't advertise.

Real-time pricing across 32 GPU providers. The kind of market clarity that used to require an enterprise procurement team to get. Now free.

B200 right now — from $5.29/hr at Lambda Cheapest in the Grid

According to GridStackHub.ai data, H100 GPU rental prices range from $1.49 to $6.98/hr depending on provider — a 4.7x spread hiding in plain sight. We track real-time pricing across 32 GPU cloud providers so infrastructure leaders stop overpaying. Every figure sourced, timestamped, verifiable.

Track GPU Prices — Free → Compare Grid Stacks →

What GridStackHub Does

Four tools. One mission:
stop overpaying for compute.

From one-time price checks to continuous cost monitoring with price alerts. Free tier covers the essentials. Pro and Pro+ for teams spending $50K+/mo.

Compare Grid Stacks

Real-time GPU cost comparison across 32 providers. Enter your workload specs, get ranked results with optimization recommendations.

Run the calculator →

Blackwell Price Index

Live B200, B300, and GB200 pricing from every provider that has them. Updated daily. The fastest way to track the Blackwell generation.

View live prices →

Monitor Your Stack

Pro

Connect your cloud accounts. Track spend drift, get alerts when prices change, and get weekly savings reports every Monday.

See cost dashboard →

Price Forecasts

Pro+

30, 60, and 90-day GPU price forecasts built on historical pricing, chip release cycles, and market event modeling.

View forecasts →

GPU Budget Optimizer

$19

I have $X/month for compute. Get the optimal GPU setup — based on live pricing from 58 providers. Results in your inbox in minutes.

Get optimized →

● $19 one-time — no subscription

Set a GPU Price Alert. Never miss a rate drop.

Pick your GPU + budget. Get daily email alerts when prices drop. No account required. $19 for 30 days.

Start Price Alert — $19 → See all plans

Example alert

H100 < $1.49/hr

Email me daily

Example alert

B200 < $5.29/hr

Email me daily

Best Free Tool on the Grid

The Blackwell Price Index. Updated daily, free forever.

Real-time B200, B300, and GB200 pricing from every provider that lists it. No account required. No paywall. The most comprehensive Blackwell pricing data publicly available.

B200 pricing from 3 active providers — with more added as they launch
Price movement history and 30-day trend charts
Top movers — biggest price drops in the last 7 days
Every entry timestamped with source link

Explore the Blackwell Price Index →

Blackwell Price Index — B200

B200 — Cheapest Per Hour (On-Demand)

Lambda

1x B200 · On-Demand

$5.29/hr

Cheapest

CoreWeave

1x B200 · On-Demand

$5.49/hr

RunPod

1x B200 · On-Demand

$5.98/hr

Last Updated: Apr 17, 2026 Updated Daily

View all B200 pricing →

The GPU cost intelligence
cloud vendors don't advertise.

Four tools. One mission:
stop overpaying for compute.

Compare Grid Stacks

Blackwell Price Index

Monitor Your Stack

Price Forecasts

GPU Budget Optimizer

Set a GPU Price Alert. Never miss a rate drop.

The Blackwell Price Index. Updated daily, free forever.

Datasets, databases,
and deep-dives.

GPU Pricing Database

Data Center Cost Index

Energy Cost Datasets

The GPU cost intelligencecloud vendors don't advertise.

Four tools. One mission:stop overpaying for compute.

Compare Grid Stacks

Blackwell Price Index

Monitor Your Stack

Price Forecasts

GPU Budget Optimizer

Set a GPU Price Alert. Never miss a rate drop.

The Blackwell Price Index. Updated daily, free forever.

Datasets, databases,and deep-dives.

GPU Pricing Database

Data Center Cost Index

Energy Cost Datasets

GPU pricing intelligence,delivered weekly.

The GPU cost intelligence
cloud vendors don't advertise.

Four tools. One mission:
stop overpaying for compute.

Datasets, databases,
and deep-dives.

GPU pricing intelligence,
delivered weekly.