Assuming a PUE of about 1.35 (roughly 11.2GW of critical IT) and $14 million per megawatt, the 15.2GW of data centers under construction need around $156.8 billion in annual GPU rental revenue to be worth it.

Not graded on evidence

Outcome: Can’t be checked

An opinion, a recommendation, or too vague to test. Listed, not scored.

Evidence up to 2026-10-28. Checked 2026-10-31.

Why

The post (28 April 2026) assumes 15.2GW under construction, a PUE of about 1.35 and $14 million per megawatt. From those inputs the arithmetic roughly works: 15.2GW divided by 1.35 is about 11.26GW, and 11.2GW times $14 million per MW is $156.8 billion. The core of the claim is that this revenue is 'required to actually make these data centers worth it.' The passage applies $14 million per MW as annual revenue but gives no basis for that figure. It does not define 'worth it' as payback, profit or covering cost. The packet did not find a source for the 15.2GW input. BloombergNEF (e1) counted about 23.1GW of IT capacity under construction worldwide at the end of September 2025, including 15.9GW in the US. Its scope and date differ from his figure, so it neither confirms nor refutes it. Without a defined standard for 'required,' only the arithmetic can be checked, and the claim cannot be graded.

Caveats

• The packet did not find the TD Cowen note behind the $14 million per MW figure or the tracker behind the 15.2GW figure.
• The arithmetic check was done from his stated inputs, not from an outside source.
• The BloombergNEF count has a different scope and date and cannot test his capacity input.

Evidence (1 item)

Claims like this aren’t scored, so there is no evidence to show.