The True Cost of AI Tools

Every time you Google, generate, or prompt — something real is consumed. Energy. Water. Money. And most of us have no idea how much.

I've been thinking about this a lot lately. As someone who helps leaders build intentional AI strategies, I keep coming back to the same question: are we using these tools consciously, or just conveniently?

This isn't a guilt trip. It's a framework. Because the leaders I work with don't want to be reckless — they want to be informed. And right now, most of the information is buried in sustainability reports and academic papers that nobody has time to read.

So I read them for you.

---

The three costs most people ignore

When we talk about the "cost" of AI tools, we usually mean the subscription price. But there are actually three layers:

1. Compute cost — the energy and processing power needed to respond to your prompt. This varies enormously: a Google search uses a fraction of a watt-hour; an AI video generation uses hundreds. It never shows up on your bill.

2. Subscription cost — what you pay per month, amortised per use. A $20/month subscription costs you roughly $0.03 per daily use — but near-zero if you're only using it twice a week. Most people are over-subscribed.

3. Environmental cost — water for server cooling, energy for data centres, carbon. Text queries are modest. Image and video generation is 10–100× heavier. Almost entirely hidden from users.

---

What does one use actually cost?

Here's the data, pulled from published research and company sustainability reports:

Gemini figure is Google's own disclosed measurement. All other water figures are researcher estimates. Energy figures represent median text prompts.

---

The water reality check

This is the one that stops people in their tracks.

• Google Search: ~0.1 ml — an eyedrop
• Gemini (AI Pro): 0.26 ml — five drops
• ChatGPT / Claude (text): 5–10 ml — a teaspoon
• Canva AI / Midjourney (image gen): 50–100 ml — an espresso shot
• AI Video, 5 seconds: 500–1,000 ml — a full water bottle 🚨

Text AI uses water equivalent to a teaspoon per query. Image generation jumps to an espresso shot. AI video is a full water bottle for five seconds of output.

Infrastructure choices matter far more than individual usage — but knowing this changes how you make decisions.

---

Which tool, when? A decision guide

"I need to look something up quickly" → Use Google Search. Fastest, cheapest, lowest environmental impact. Don't fire up an LLM for something a search can answer in three seconds.

"I need to write, analyse, or strategise" → Use Claude or ChatGPT. Text LLMs offer enormous value at modest cost. Claude excels at long documents and nuanced writing; ChatGPT for multimodal, voice, and image tasks. $20/month is excellent value if used daily.

"I need to design branded content" → Use Canva (non-AI features). Canva's template library is still one of the best-value design tools available. Use AI features only when they genuinely add value — the 300%+ price hike means you're paying for AI whether you use it or not.

"I need a custom image or visual" → Use Midjourney or DALL-E intentionally. Image generation is ~10–20× heavier than text generation. Ask: could a stock photo, a Canva template, or a text description serve this purpose instead?

"I need to create a video" → Use AI video sparingly. A 5-second AI video clip uses the same water as a full water bottle. For social content, consider whether real footage, screen recording, or a still image achieves the same result more sustainably.

"I'm not sure if I need AI at all" → Apply the Wonder Test. Ask: does AI make this genuinely better, or just faster? The most sustainable use of AI is intentional use — choosing the right tool for the right task, with a clear outcome in mind.

---

5 ways to use AI more consciously

01. Right tool, right task. Don't use an LLM when a search will do. Save AI capacity for tasks that genuinely need reasoning or generation.

02. Audit your subscriptions. $20/month is only good value if you use it daily. Occasional users may be better off on free tiers or API pay-per-use.

03. Lean into text over visuals. Text generation uses a fraction of the energy and water of image or video generation. Default to text where possible.

04. Batch your prompts. One well-crafted prompt uses the same resources as one lazy one — but produces far better output. Fewer, sharper requests = less waste.

05. Ask who's transparent. Google disclosed 0.26 ml/query + 0.03g CO₂. Most AI companies publish nothing. Support providers who are honest about their footprint.

---

The big picture

Your individual usage matters less than demanding better infrastructure. AI companies desperately need users' buy-in — that's your leverage.

The leaders who will thrive in the next decade aren't the ones who use AI the most. They're the ones who use it best — with intention, with discernment, and with a clear understanding of what they're trading.

That's what I call the Wonder Mindset.

---

Sources: Google 2025 Sustainability Report · MIT Technology Review · Epoch AI · UC Riverside (Li et al., 2023) · Goldman Sachs · BestBrokers 2025

Want to know your own AI fluency and capability score? Take the free Wonder Audit →