Layer 1 · Where the power comes from
On-site natural gas turbines
Anyone who needs hundreds of megawatts before the grid can deliver them, and will accept the emissions.
The shape of it
- Lead time
- 12-24 months, the fastest firm power available
- Capital cost
- Moderate; turbine supply chains are now the constraint
- Carbon
- High. This is the trade being made.
- Maturity
- proven
Verified 2026-09-06.
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.
What it actually decides
The honest bridge. Most 2026-2028 AI capacity that actually energises on schedule will be gas, whatever the announcements say.
What this is powering
A GB200 NVL72 rack draws 120–132 kW and a GB300 pushes 135–200 kW, against a 2026 average rack of about 27 kW. The silicon decision and the power decision are the same decision, made eighteen months apart.
Layer 2 — Silicon · GB200 NVL72 · The 150 W inference option
Weighed against
- Nuclear PPA (existing plants) vs On-site natural gas turbines — Clean and slow against fast and dirty. The defining 2026 trade, and most operators are quietly choosing gas.
- Solar + battery storage vs On-site natural gas turbines — Intermittent and clean against firm and fast. Almost always paired rather than chosen.
- Fuel cells vs On-site natural gas turbines — The premium bridge against the default bridge.
Others in where the power comes from
- Grid interconnection — 4-5 years typical, approaching 5 on average
- Nuclear PPA (existing plants) — 1-3 years to contract; the plant already exists
- Small modular reactors (SMR) — 2030s for meaningful capacity
- Behind-the-meter generation — Depends on the generation, but it skips the queue
- Solar + battery storage — 18-36 months
- Wind PPA — 24-48 months