Straight answers, live numbers
The questions people actually ask, each answered in one sentence with today’s data — then the numbers behind it. Answers recompute on every load and say which day their data is from. Quote them with attribution (how to cite).
How much does an NVIDIA H100 cost to rent per hour?
Live median H100 SXM rental price per GPU-hour across cloud providers, updated hourly.
How much does an NVIDIA B200 cost to rent per hour?
Live median B200 rental price per GPU-hour across cloud providers.
How much does an RTX 4090 cost to rent per hour?
Live median RTX 4090 cloud rental price per GPU-hour, marketplace and cloud tiers.
How much does an NVIDIA A100 cost to rent per hour?
Live median A100 80GB SXM rental price per GPU-hour across cloud providers.
Which cloud GPU provider is cheapest?
It depends on the GPU — the same SKU varies up to ~2× across providers. Today's cheapest per SKU, computed live.
How much cheaper are spot/interruptible GPUs than on-demand?
Live spot-vs-on-demand discount per GPU SKU, computed from today's medians.
What is the Price of Compute Index at today?
The live composite index of GPU rental prices — weighted basket of the 6 most-rented SKUs, base 100.
How much have cloud GPU prices dropped since 2023?
What the recovered 2023→now list-price record actually shows — including prices that went up.
Which LLM API is cheapest per token?
Live blended $/MTok ranking across Anthropic, OpenAI, and Google — verified list prices.
What is a GPU-hour and why is it the unit?
The unit every compute price on this site is normalized to, and why per-GPU normalization matters.