Learn
Start here
New to AI infrastructure? Read these five in order. Each of the first four comes from one of the tracks and takes a few minutes; together they cover the building, the economics, the money and our data, before the Primer takes you through the whole sector.
- What is a datacentre? The building, its power and cooling, and what changes when the tenant is an AI cluster.
- GPU-hour economics What a GPU earns and costs, and how long a cluster takes to pay back.
- Hyperscaler capex in 2026 What the largest spenders put into the build-out, from their filings.
- How to read the price board What our daily GPU prices mean, and how much weight each one can bear.
- The AI Infrastructure Primer 2026 The whole sector in one report. The first seven pages are free with Compute Open, and the full Primer comes with any paid plan.
Step 1
What is a datacentre?
What a datacentre is, the parts every one shares, how its capacity is measured and what changes when the tenant is an AI cluster.
Step 2
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).
Step 3
Hyperscaler capex in 2026, quarter by quarter from the filings
Microsoft, Alphabet, Amazon and Meta paid $201.9bn for property and equipment in Q3 2026, up 108% on the same quarter a year earlier ($97.3bn). These are cash figures from the companies' 10-Q and 10-K filings, not guidance.
Step 4
How to read the price board
What each cell on the price board means: the provider classes, the sample size, the published, indicative and census labels, and what each plan sees.
Step 5
AI Infrastructure Primer 2026
The capstone: its opening pages are free with Compute Open, and the whole Primer comes with every paid plan.
Free · every trading day · 08:00 ET