The cloud GPU market, as one dataset.
Current prices and history from 91 cloud providers, in one API. Track the market and benchmark any provider.
Using an AI coding agent? Point it at our setup guide.
- Providers
- 91
- Hyperscalers to GPU clouds
- Price points
- 4,332
- 110 GPU models
- Fastest refresh
- 15 min
- Per provider, down to once a day
- Price history
- 2024 since
- Weekly, daily since Jul 2026
See the whole market in one call.
Search GPU offers across every provider we list, in USD per GPU-hour. The most expensive H100 costs 6.8x the cheapest.
// Every H100 on-demand listing
GET /api/gpu-offerings?gpu_model=nvidia-h100
&billing_type=ON_DEMAND&sort=price_per_gpu_hour
See where prices are going.
Per GPU model and per provider: weekly snapshots since mid-2024, and a daily series since July 2026. Each point is the market's min, median and max for that day.
// Daily H100 market-wide point
GET /api/gpu-price-history?gpu_model=nvidia-h100
&billing_type=ON_DEMAND&interval=daily
Benchmark any provider against the market.
One provider's series against the market's, week by week or day by day. We check each provider on its own schedule, from every 15 minutes to once a day.
// One provider's H100 against the market, daily
GET /api/gpu-price-history?gpu_model=nvidia-h100
&provider=gcore&billing_type=ON_DEMAND&interval=daily
34% below the market median
Source compute at the right price.
Need 8x A100 on spot under $15 an hour? Filter by spec, billing model and provider country.
// 8x A100, spot, under $15/hr, US providers
GET /api/gpu-offerings?gpu_model=nvidia-a100
&min_gpu_count=8&billing_type=SPOT
&max_price=15&provider_country=US
How it works
-
01
Get API credentials
Sign up with a work email and get a trial key right away. Subscribe when you are ready.
-
02
Authenticate
Plain REST with a Bearer token. No SDK required; any HTTP client works.
-
03
Build
Page through offerings, pull history, and put market data in your platform.
Built for people who watch this market
Market visibility
See the global GPU market as it stands and as it moved, per model or per provider.
Competitive benchmarking
Put your rates beside every other provider's, in one unit, and watch them move week by week.
Investment and market research
A price series since 2024 for a market that moves daily. Chart it, model it, cite it.
Cost-aware sourcing
Pick the lowest priced provider for each workload, filtered by spec, billing model and country.
Predictable pricing
One plan, full access, no per-seat pricing. Pay monthly or save with a year up front.
Every plan includes
- Requests
- 3,000 a day
- Coverage
- All 91 providers, 110 GPU models
- Filters
- Filtering and metadata on every endpoint
- Refresh
- Every 15 minutes to daily1
- History
- Weekly and daily price history2
- Support
- Email support
Need higher quotas? Contact us for an enterprise plan.
Free trial
100 requests on the full API. No card needed, just a work email.
[1] A small number of providers are updated less frequently. [2] Weekly snapshots since mid-2024, daily since July 2026, with coverage widening over time.
Plan prices exclude applicable taxes.
Common questions
Are these listed prices or transacted prices?
Listed. Each figure is the price the provider published when we checked: on-demand, spot or reserved, converted to USD per GPU-hour. Nothing in the API is a transaction, a quote or a negotiated rate. A spot figure is the published spot price at the check, not a fill.
How is the data sourced?
We read each provider's public pricing pages and APIs and normalize the figures. Always verify current rates with the provider before provisioning.
How often do you update pricing?
It depends on the provider. Most offerings are re-checked every 15 minutes or hourly, the rest once a day. Providers whose prices move through the day are on the fastest tier. A provider with a stable price list is checked daily. A small number of providers whose prices rarely change are maintained by hand. Each offering carries the time of its last check.
How far back does the history go?
Weekly points start in July 2024, daily points in July 2026. Both exist per GPU model and per provider. Filter the history endpoint by provider to get that provider's own min, median and max per week or day. A provider's series starts the week we listed it. Each market-wide point says how many providers it covers.
Can we use the data in internal models and reports?
Yes. Anything you build from the data, an index, a benchmark, a model, a report or a dashboard, is a Derived Product, it belongs to you, and you may commercialize it. Among the use cases we do not allow: redistributing, mirroring, sublicensing or reselling our data, or running a competing pricing API or directory on it. See section 10 of the Terms.
What happens to the data if we cancel?
You keep it. Copies you received stay with you for records, reproducibility, audit, compliance and disputes. Derived Products you built are unaffected. Your key stops working at the end of the billing period. See section 10 of the Terms.
What formats can we export?
JSON. Every endpoint pages, and a history response is one row per week or day. For a flat file, the open GPU prices dataset publishes the weekly market-wide series per GPU model as CSV and JSON, free under CC BY 4.0. Daily points and per-provider rows are API only.
Do you offer a trial?
Yes. Sign up with a work email and get 100 requests on the full API, no card needed. Once they are used, the API answers 402 until you subscribe.
Can I set it up with an AI coding agent?
Yes. Paste this into Claude Code, Cursor, Codex or any coding agent: "Read https://getdeploying.com/help/api/agent-setup.md and follow it to set me up with the GetDeploying API." It reads our setup guide, asks you to add your key to the project and writes the integration.
What happens if I hit the rate limit?
The API will return a 429 status code. If your application requires higher throughput, please contact us for an Enterprise plan.
How does billing work?
All plans are billed at the start of each billing cycle. Monthly plans renew every month, while annual plans renew once per year at a discounted rate.
How do I cancel my subscription?
You can cancel anytime by emailing [email protected] - no questions asked. Your subscription remains active until the end of the current billing period. Please note no partial refunds are issued for unused time.
Get started today
The cloud GPU market from 91 providers, in one schema, with price history since 2024.
Questions? [email protected]