Layer 1 · head to head
Converted bitcoin mine vs On-site natural gas turbines
Two ways to skip the queue. One is a retrofit; the other is a turbine slot that may be 2031.
| Converted bitcoin mine | On-site natural gas turbines | |
|---|---|---|
| Lead time | 12-24 months to convert, against 4-5 years for a greenfield grid connection | 12-24 months, the fastest firm power available |
| Capital cost | Site and power already exist; the spend is cooling, networking and density retrofit | Moderate; turbine supply chains are now the constraint |
| Carbon | Inherits the site's existing supply, which is frequently gas or stranded hydro | High. This is the trade being made. |
| Maturity | proven | proven |
Converted bitcoin mine
The catch: You are buying a building designed for machines nobody minded losing. Mining tolerates outages, poor cooling and low reliability because a lost hash is worth nothing; a lost training run is worth a great deal. The retrofit is real work, and the sites worth converting were largely converted first — quality falls as you move down the list. Deployment still runs 60-70% faster than greenfield, which is the whole argument.
Decides: If your model is training in 2027 rather than 2031, this is one of very few honest answers. It is a power-date decision wearing a real-estate transaction.
On-site natural gas turbines
The catch: It works, it is fast, and it puts your sustainability commitments in direct conflict with your capacity plan. Turbine lead times have themselves stretched as everyone reached the same conclusion.
Decides: The honest bridge. Most 2026-2028 AI capacity that actually energises on schedule will be gas, whatever the announcements say.