Current H100 cloud pricing
The CT-H100 median, today’s per-provider on-demand rates, and how the spot and reserved tiers relate to the list price.
GPU Pricing
Current GPU Cloud Pricing
Sourced GPU cloud rates from public provider price pages — current H100, H200, and B200 on-demand bands, each row source-linked and dated.
Every rate on this page comes from a public provider price page that has been editorially checked and carries its own source link and price date. Start with the on-demand bands below, open a chip page for the per-provider read, or go straight to the directory for every reviewed row.
$4.29 / H100-hrCT-H100 Index
Median sourced on-demand H100 rate across 7 providers, as of Jun 11, 2026
$2.40–$12.29 (7 providers)H100 on-demand band
Lowest to highest sourced public list price, as of Jun 11, 2026
15Reviewed provider rows
Source-linked price-page rows in the directory, generated Jun 11, 2026
On-demand bands
GPU price bands by chip
Each band is the lowest-to-highest sourced public on-demand list price for that chip. A band is only as wide as its evidence: chips sold mainly through reservations or allocation show no band rather than an invented one.
| Chip | On-demand band ($/GPU-hr) | Where to read more |
|---|---|---|
| H100 | $2.40–$12.29 (7 providers) | Current H100 pricing |
| H200 | $3.50–$10 (5 providers) | Current H200 pricing |
| B200 | $5.89–$8.60 (3 providers) | Current B200 pricing |
Public on-demand list prices from provider price pages — not negotiated quotes, reserved terms, or spot rates. Each row carries its own source link and price date. How we source, label, and date evidence: methodology.
Go deeper
The per-chip read
Each chip page is the per-provider read behind its band: sourced on-demand rates, lowest list price first, every row linked to the public price page it was observed on.
H100 cloud comparison
The same sourced rows arranged for a buying decision — provider type, region, availability, and cost-estimate prefills.
Current H200 cloud pricing
Sourced per-provider H200 on-demand rates, with the H100-equivalent column for reading the memory premium.
Current B200 cloud pricing
Sourced per-provider B200 on-demand rates — the public list-price slice of a market that still sells mostly reserved.
The data behind this page
Inspect the evidence
This hub is a reading layer. The numbers live on four surfaces, each with its own methodology trail and machine-readable export where available.
ComputeTape H100 Index (CT-H100)
The daily, citable median of verified on-demand H100 list prices — with history chart, CSV/JSON, and an embeddable card.
Provider Directory
Every reviewed provider row: public price-page links, H100e-normalized rates, availability, and price dates.
Capacity Board (CT-CAP)
GPU availability signals by provider and chip — what is actually orderable, not just what is priced.
Market Signals
The capacity-led market page: availability shifts, the directional H100e market-rate estimate, and the daily read.
Compute College
Understand the numbers
The lessons explain what moves these rates — the pricing unit, the contract tiers, and the hardware generations behind the bands.
What is a GPU-hour?
The unit every rate on this page is quoted in.
H100 price per hour, explained
What determines the rate, and why two H100 quotes rarely describe the same product.
On-demand vs reserved vs spot
The three contract tiers behind every price band, and when each one is the right buy.
H100 vs H200 vs B200
How the generations compare on performance per dollar, memory, and availability.