Physix Frontier · News Briefing Card (Hacker News · Nov 20, 2026)
Blogger estimates personal AI environmental impact: $5,871 in six months
KEY FACTS
- The author used the Kimi K3 model to estimate power consumption at about 0.14 kWh per million tokens.
- On that basis, each million tokens consumes about 0.35 L of water and emits about 0.05 kg of carbon.
- The author assumed environmental impact scales with model price to estimate figures for closed-source models.
- The author tallied AI coding spend from May to November, about 6.3 months, at a total of $5,871.
- The spending was spread across 8 models, including Fable 5 and Opus 5.
KEY DATA
about 0.14 kWhPower per million tokens
about 0.35 LWater per million tokens
about 0.05 kg CO2Carbon per million tokens
5871 USDTotal personal AI spend
PHYSIX OBSERVATION
Converting environmental impact into a per-dollar metric is a practical way to make ordinary users aware of AI energy consumption, but the assumption that price is proportional to energy use is quite crude, and the true overhead of closed-source models may be underestimated or overestimated. For users, the real signal is this: model choice and cache hit rate directly affect the bill and indirectly affect the carbon footprint.
Source: Hacker News report
Physix Frontier