💧 Water — is the cooling medium of AI still there?

Latest observations

Percent of area is the Drought Monitor's own cumulative statistic (D1 or worse). "Not yet measured" means missing, not zero. Machine-readable: /us/api/water_us.json

Observation tables

Question this table answers: what are the values and dates in each row of "USGS - Estimated use of water in the United States (5-year compilation) - national figur…"? (source or retrieval date on each row)

Question this table answers: what are the values and dates in each row of "Water-disclosure status by operator (14 companies, 5 levels: WUE figure / volume / polic…"? (source or retrieval date on each row)

Question this table answers: what are the values and dates in each row of "CONUS area in drought, D1 (moderate) or worse - percent of area (U.S. Drought Monitor, w…"? (source or retrieval date on each row)

Question this table answers: what are the values and dates in each row of "U.S. total (50 states + Puerto Rico) area in drought, D1 or worse - percent of area (wee…"? (source or retrieval date on each row)

Candidates (not yet observed) — added to this chapter once verification completes

• USGS water use - machine-readable re-retrieval (NWIS water-use service) — The national figures are published as a ledger above (Circular 1441); the NWIS retrieval endpoint now redirects to an information page (2026-09-04), so no machine-readable series exists yet.
• Data-center water withdrawal by state — No public periodic statistic exists (checked 2026-09-03); operator disclosures are tracked in the ledger above instead.

Method, scope, and why this chapter is not judged

• Not judged: no criterion says what percent of the country in drought is "bad for AI"; the status stays "observing". We do not infer a threshold from the drought categories.
• Drought series: the U.S. Drought Monitor's weekly area statistics (statisticsType=1, cumulative) for the contiguous U.S., published as-is: the headline value is the percent of area in D1 (Moderate Drought) or worse; the JSON keeps every cumulative column (D0+ … D4) and the "none" share. The U.S.-total area (with Alaska, Hawaii and Puerto Rico) is kept as a reference series. Categories are the NDMC definitions (D0 Abnormally Dry, D1 Moderate, D2 Severe, D3 Extreme, D4 Exceptional). Re-use with attribution per the NDMC Web Policy; listing here implies no endorsement by NDMC.
• Disclosure ledger: the same method as the Japan edition's water chapter — we check each operator's official sustainability page by hand and record only whether a disclosure exists, at four levels (WUE figure / total volume / policy only / none or not confirmed), with a link to the page and the date we checked. No operator figure is reproduced. An operator we could not confirm is recorded as "none or not confirmed", never guessed. Annual reports we did not open are noted as such.
• Boundaries: drought as an external hazard sits here (water), not in the disasters chapter; cooling load (degree days) sits in the data-centers chapter.
• Chapter Beta: removed automatically once the drought feed has fetched successfully for 5 consecutive days. Data collection began 2026-09-03 with history back to 2026-01-01 retrieved from the Drought Monitor archive (weekly points).
• What this chapter can say: what percent of the contiguous U.S. (and of the U.S. total area) was in drought category D1 or worse each week, as the U.S. Drought Monitor publishes it; whether each of the 14 cloud / data-center operators in the panel publishes a water disclosure, at which level (WUE figure / volume / policy only / none), with the page link and the date checked; and, for WUE disclosers, the figure only where it was transcribed directly from the primary-source report with its page number; the national water-use figures USGS published in its five-year compilation (ledger, as published).
• What this chapter cannot say: how much water U.S. data centers withdraw or consume, by state or by facility (no public periodic statistic exists; checked 2026-09-03); whether drought has constrained cooling at any specific facility or region (regional drought extent only); whether any drought level or disclosure count is good or bad (no threshold is inferred from the drought categories).
• Sources: U.S. Drought Monitor (National Drought Mitigation Center at the University of Nebraska-Lincoln, USDA, NOAA and NASA; web policy), operators' official sustainability pages (linked in the ledger). Compiled by AI Lifeline.