As AI chatbots become a daily tool for everything from drafting emails to spotting financial fraud, their hidden energy cost is rising. Data centers already consume about 1.5% of global electricity, and demand is set to double by 2030. While a single query uses only a fraction of that power, the sheer volume adds up.
Choose Necessity Over Convenience
First, ask yourself if AI is truly needed for a task. “Asking ChatGPT ‘What should I wear today?’ is like taking a Concorde to the grocery store,” says AI expert Günter Klambauer. Simple queries can often be answered with a quick web search or personal knowledge.
Turn Off AI‑Enhanced Search Overlays
Search engines that auto‑generate answers, such as Google’s AI overviews or Bing’s Copilot, add extra processing. Selecting “Web results only” or adding “-ai” to your search can cut unnecessary computation.
Use Smaller, Specialized Models
For frequent tasks like translation or summarization, opt for compact models that focus on a single function. A 2025 UNESCO study by Ivana Drobnjak showed these models use 15‑50 times less energy than large, general‑purpose LLMs while delivering higher‑quality results for their niche.
Limit Output Length
Large language models consume the most power when generating long responses. Instructing a model to be brief or setting a word limit can halve its energy use, according to Drobnjak’s research. Keeping prompts short yields only modest savings (about 5%).
Select Efficient Modes
Some models have “reasoning” modes that produce many more words and consume significantly more power. Use these modes only for complex problems; otherwise, stick with the standard chat mode.
Leverage Energy‑Saving Algorithms
Tech companies report that newer versions of their models are far more efficient. Google’s Gemini now uses 0.24 watt‑hours per median‑length query—a 33‑fold improvement over its 2024 version.
By applying these straightforward habits—questioning necessity, disabling AI overlays, choosing smaller models, and limiting output—users can collectively reduce the growing energy demand of AI tools while still enjoying their benefits.
Original reporting: KEYT (Ventura/Santa Barbara) — read the source article.
