"Is this really something I need to ask AI?" Over 60% Say Yes

"This isn't really something I need to ask AI... but I'll ask anyway."

Have you ever found yourself asking generative AI a question with this kind of thought?

According to the "Consumer Energy Awareness Survey 2026" conducted by EREX, 61.5% of generative AI users reported having "casually asked AI a question for no particular reason," even when it wasn't strictly necessary.

Looking at the breakdown, 13.7% said "frequently" and 47.8% said "occasionally." By age group, this trend was particularly strong among younger generations, reaching 74.2% for those in their 20s and 67.3% for those in their 30s.

The actual responses to these "casual questions" were quite varied, including things like "What should I do if I can't date the person I like?" and "I wonder if someone would give me 30 million yen?"

In this survey, such "casual questions" are referred to as "AI power waste."

Nearly 80% Don't Care, But AI Uses Electricity!

While users may not actively think about it much, the power consumed by servers and processors in data centers that operate AI is enormous.

However, regarding the fact that generative AI uses a large amount of power in data centers, 78.1% responded either "I didn't know" or "I knew, but I hadn't particularly thought about it."

On the other hand, concerning the increase in Japan's electricity demand due to the spread of AI and data centers, 67.0% responded that they "feel anxious." Among generative AI users specifically, this percentage rose to 69.6%.

The anxieties included "rising electricity bills" (58.1%) and "insufficient power supply" (49.7%).

However, among AI users, only 12.3% said they would "want to reduce AI usage" if power consumption increased, while 45.4% responded that they "want to continue using AI in the future."

The Key is How to Sustainably Support Power Supply

The survey estimates that people aged 20-69 in Japan input approximately 63.5 billion text questions to generative AI annually.

Based on this number of queries and Google's published figure of "0.24Wh per text response" for Gemini, the annual power consumption is estimated to reach approximately 15.2 GWh.

This is a very significant figure, equivalent to the amount of electricity consumed by approximately 1.42 million ordinary households in one day.

Furthermore, this estimate only includes text generation and does not account for image or video generation. Also, 0.24Wh is a figure published for Gemini and does not directly apply to all generative AI. This means that even more power could potentially be consumed for generative AI usage.

So, if we aim to curb rising electricity demand and reduce environmental impact, should we refrain from using AI?

Judging from the results of this survey, user sentiment is not moving towards "not using AI." Rather, the question is how to secure a stable power supply to meet increasing electricity demand and reduce the environmental burden.

As for preferred power generation methods if electricity demand increases, 54.2% cited "renewable energy." Furthermore, 56.8% responded that it is "important" to cover the electricity used by AI and data centers with renewable energy. Among AI users, this figure rose to 67.4%.

Casually asking AI questions. Enjoying its convenience. On the other hand, how will the electricity used for it be generated and delivered?

Precisely because we are entering an era where AI becomes commonplace, thinking about the energy that supports its power will likely become an even more familiar topic than ever before.