{"inputs":{"queries_per_day":25.0,"usage":"simple","boundary":"comprehensive","days":250},"per_query":{"energy_wh":0.34,"water_ml":1.39},"per_day":{"energy_wh":8.5,"water_ml":34.8,"energy":"8.5 Wh","water":"34.8 mL"},"range":{"energy_wh":[637.5,6375.0],"water_ml":[2613.75,26137.5],"water":"2.6 L – 26.1 L","applies_to":"total"},"total":{"energy_wh":2125.0,"water_ml":8712.5,"energy":"2.1 kWh","water":"8.7 L","energy_equivalents":{"tv_100w":"21.2 hrs of TV","phone_charges":"141.7 phone charges"},"water_equivalents":{"teaspoons":1778.1,"bottles_500ml":17.42,"shower":"58 sec of showering"}},"assumptions":["Usage tier: Simple text – chat, search, summaries (0.34 Wh per prompt).","Boundary: Onsite cooling plus water consumed generating the electricity (~4.1 mL/Wh; US grid average per Li et al.).","Chat-style prompts (one prompt, one response). Agentic sessions can be far higher – use estimate_session_footprint.","Uses the most conservative (highest) of the published onsite water intensities. Excludes model training.","Likely range 0.3x–3x the central estimate: a judgement-based summary of how much published per-prompt energy figures and provider/grid water intensities vary, not a statistical interval."],"methodology_version":"2026-10","reviewed":"October 2026","source":"https://www.pharmatools.ai/water-consumption-calculator","method_note":"https://doi.org/10.5281/zenodo.23417340"}