Explainers
How AI infrastructure works
What is a neocloud?
A neocloud is a cloud provider built around one product: large, tightly networked pools of the latest AI accelerators, rentable with minimal friction.
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).
What is an AI factory?
The phrase AI factory — popularised by NVIDIA and now standard industry vocabulary — is more than marketing. It describes a datacentre engineered as a production line: the raw materials are electricity and data, the machinery is racked accelerated computing, and the output is tokens — the units of text, code, images and actions that AI models consume and produce.
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.
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.
The numbers, read live
H100 price per hour in 2026: what a GPU-hour costs now
The median advertised on-demand price of an H100 GPU-hour was $3.52 in the week of 12 Oct 2026, across 41 providers — up 18% on the same week a year earlier ($2.99). The B200 moved +39% over the same year, to $7.95.
Hyperscaler capex in 2026, quarter by quarter from the filings
Microsoft, Alphabet, Amazon and Meta paid $165.1bn for property and equipment in Q2 2026, up 87% on the same quarter a year earlier ($88.2bn). These are cash figures from the companies' 10-Q and 10-K filings, not guidance.
AI datacentre deals: the last 30 days of transactions
Between 14 Sep 2026 and 14 Oct 2026 we logged 47 transactions across AI infrastructure: 34 deals, 5 listings and 8 sets of talks. 12 disclosed a value in US dollars, together $15.1bn, of which $5.7bn was the repricing of existing facilities.
The full Primer is free with Compute Open: get the AI Infrastructure Primer 2026 →
Free · every trading day · 08:00 ET