A pretty compelling argument for investment based on steadily expanding demand. Should demand cool because of poor financial returns downstream then this all becomes a pumpkin.
The main builds need access to steady base power. That is not so easy until we build major new power plants.
I do think that this wiill be a lot slower than promoters think. Way more like the economics of remote mine development. Those take years because you have to build the road and rail connection and the hydro dam as well and the support town. Nothing is off the shelf.
Economics of a Megawatt of AI Data Center
May 2, 2026 by Brian Wang
Granular Unit Economics per MW
https://www.nextbigfuture.com/2026/05/economics-of-a-megawatt-of-ai-data-center.html#more-209946
Cruseo CEO Lochmiller broke down the inputs/outputs with high precision (April 2025 contract pricing and the talk)
Upfront capex is ~$59M per MW of developed capacity.
Roughly half (~$30M) is IT (GPUs, CPUs, networking, storage, etc.).
The rest covers data center build, power plant/generation, and other infrastructure.
Annual revenue (pure infrastructure lease/IaaS layer) is ~$15M per MW.
Higher-value managed services/cloud layer can add another ~$15M+ per MW in revenue.
Annual opex is Only ~$1–1.1M per MW (power + other).
Net result is every MW developed can yield $1M+ in annual net revenue as pure infrastructure, with strong margins at the right layers. Payback is attractive (4 years on infra alone)
No comments:
Post a Comment