Learn · track 2 of 4

The economics

What a GPU-hour costs and earns: how rental prices form, how contracts turn them into revenue and how long a chip keeps earning.

A GPU earns its money by the hour, and most questions in this sector come back to what that hour is worth and for how long. The same hour has several prices at once. For an H100 on the latest price board, a neocloud's on-demand rate reads

$3.89 per GPU-hourCompute Terminal price board (public rate cards and marketplaces)As of 11 Nov 2026Source
(as of 11 Nov 2026), the hyperscalers' list price
$10.04 per GPU-hourCompute Terminal price board (public rate cards and marketplaces)As of 11 Nov 2026Source
(as of 11 Nov 2026) and the marketplace price for spare capacity
$2.41 per GPU-hourCompute Terminal price board (public rate cards and marketplaces)As of 11 Nov 2026Source
(as of 11 Nov 2026).

This track explains where those prices come from and what they pay for. It begins with GPU-hour economics: what a cluster costs to build and how long it takes to pay back. It then converts between a datacentre lease, priced per kilowatt-month, and a price per GPU-hour. Next come take-or-pay contracts, the reserved capacity that lenders finance, and how on-demand, reserved and spot prices form. It ends with a chip's economic life: how many years it keeps earning, and what that means for the depreciation in a company's accounts.

Sources

  • Compute Terminal price board, from public rate cards and marketplaces; methodology.
  • AI Infrastructure Primer 2026, Compute Terminal.
  1. 1 · 3 min read

    GPU-hour economics: what a GPU earns and costs

    A GB300-class megawatt costs roughly $45–55m all-in to stand up (author estimate, Jul 2026).

  2. 2 · 2 min read

    From $/kW-month to $/GPU-hour

    CleanSpark's 20-year, $6.6bn, 175 MW lease in Georgia (ready late 2027) works out at roughly $1.9m per MW per year, or ~$157 per kW-month (author arithmetic) — an order of magnitude below GPU-capacity contracts, because the tenant buys the hardware.

  3. 3 · 5 min read

    Take-or-pay compute contracts

    The industry's commercial core is the take-or-pay reserved-capacity contract: a customer commits to pay for a defined block of GPUs (or MW) for two to five-plus years, whether or not it uses them.

  4. 4 · 6 min read

    On-demand, reserved and spot: how GPU rental prices form

    The layers of price for the same GPU-hour, from on-demand list prices to spot, reserved terms and multi-year contracts, and what moves each of them.

  5. 5 · 6 min read

    GPU economic life and depreciation

    How long a GPU keeps earning, how our Economic Life Tracker reads that from rental prices, and why it matters for the depreciation in company accounts.

Next track: The capital → · Glossary · Start here

Free · every trading day · 08:00 ET

Get the CTI Compute Cost Index free in your inbox every morning.

By continuing you agree to the Terms of Subscription and the Terms of Use.

Get the CTI Compute Cost Index free in your inbox every morning.

By continuing you agree to the Terms of Subscription and the Terms of Use.