1 gigawatt of AI data center capacity costs about $50-60 billion.

Judged as of 2026-09-04, allowing for the claim's hedge 'about'. TRUE: credible estimates current at that date (NVIDIA statements, analyst reports or reported deal values per GW) put the total cost of 1 GW of AI data-centre capacity, including compute, at roughly $45–65B. MOSTLY TRUE: credible estimates cluster close by, roughly $40–45B or $65–75B. MIXED: estimates vary widely and only some support the figure. MOSTLY FALSE: most credible estimates are well outside, for example around $30B or $90B or more, though the figure reflects a real but narrow measure. FALSE: credible estimates are far from the figure, for example under $25B or over $100B.

Set before the research began.

1. E1 The finances of AI: Data & research Epoch AI · 8 January 2026

   “A typical one-gigawatt AI data center requires $38 billion in up-front capital expenditure (CapEx) and $0.9 billion in annual operating expenses (OpEx).”

2. E2 What a Gigawatt Costs Data Gravity (datagravity.dev newsletter) · 27 July 2026

   “A gigawatt of AI compute costs about $37.6 billion.”

The two independent estimates, about $38B in up-front CapEx [E1] and about $37.6B [E2], are below the $45–65B range for true. They are also below the roughly $40–45B mostly-true band and above the roughly $30B mostly-false example, so they fall between bands the criteria define. With only these two estimates, the criteria do not settle whether the claim is mostly true or mostly false.