A Few Drops of AI: Thinking of the Environmental Cost

I use AI every day in my business to help me write, plan, troubleshoot, and support my clients. It’s become a key part of how I work. But like many small business owners who care about the bigger picture, I’ve found myself wondering: is this tech helping me grow while quietly adding to the planet’s problems?

The headlines are hard to ignore. We’ve all seen the dramatic claims: AI queries using up bottles of water, data centres consuming more energy than entire countries, and a looming environmental crisis driven by our digital tools. It’s enough to make any values-led solopreneur feel conflicted.

So I did what I always do when something feels off:

I dug into the research. And what I found was far more layered than the headlines suggest. It’s a paradox: the AI industry’s overall footprint is a growing concern, but the environmental cost of your day-to-day usage as a small business is, for now, surprisingly small.

Let’s break it down.

1. The Water Debate: Context Matters

One of the most common claims is that AI “guzzles water.” And yes, training a large model like GPT-3 reportedly used around 700,000 litres of freshwater. That’s a huge one-off cost, but it’s spread across billions of queries.

When you zoom in on a single prompt, the numbers shift dramatically, and not all sources agree.

• Google estimates that a typical Gemini AI query uses just 0.26 millilitres of water, about five drops.

• A UC Riverside study, quoted by The Register, puts on-site water use at 2.2 millilitres per query, and total water use (including electricity generation) as high as 47.5 millilitres, nearly a shot glass per prompt.

• The World Economic Forum suggests that a single ChatGPT query could use the equivalent of a 500ml bottle of water for every 10–50 responses, depending on the data centre’s location and cooling method.

The difference often comes down to what’s being measured. Google’s figure only includes on-site cooling, while others factor in the water used to generate electricity. As Shaolei Ren of UC Riverside puts it:

“Google’s claims are misleading because they draw a false equivalence between onsite and total water consumption.” — The Register

Ground News adds another layer: the time of day you use AI (say, 3 AM vs noon) can affect its carbon intensity, depending on how strained the grid is. These metrics are already influencing infrastructure decisions.

So what does this mean for us? The per-query water footprint is tiny, but the wider system it sits within is complex and worth keeping an eye on.

2. Putting It in Perspective

Let’s say you use AI for 30 queries a day, five days a week. That’s 600 queries a month.

Using conservative estimates, your monthly AI usage would consume:

• Water: 156 millilitres (less than a cup of coffee)

• Energy: 144 watt-hours (less than running a desktop computer for two hours)

Now compare that to everyday activities:

• A 10-minute hot shower uses around 60,000 millilitres of water, equivalent to 380 months of AI usage.

• Driving one mile produces the same carbon emissions as 13,500 AI queries

So yes, the industry’s footprint is real. But your personal contribution? A drop in the bucket.

3. The Flip Side: AI as a Sustainability Tool

Here’s the part I find most compelling: AI isn’t just a resource consumer — it’s also a powerful tool for environmental change.

• Optimising Supply Chains: AI systems like UPS’s ORION have saved millions of miles in travel, cutting fuel use and emissions.

• Conserving Water in Agriculture: AI helps farmers use satellite data and soil sensors to reduce water usage by up to 30%.

• Building Smarter Energy Grids: AI predicts energy demand and helps balance renewable sources, reducing waste and supporting clean energy transitions.

The energy demands of AI are also pushing tech companies to invest in renewables and decarbonisation — creating ripple effects across the infrastructure we all rely on.

Even platforms like Nebula are hosting creators who explore the hidden lifecycle costs of AI — from mining rare earth metals to the energy intensity of inference vs training. Their commentary reminds us that the environmental impact isn’t just about per-query stats, but about the global systems that support them.

4. The Hidden Costs of Digital Habits We've Used for Years

Even simple digital actions like sending an email or posting on social media come with a tiny environmental footprint. It’s not the message itself that uses water, but the infrastructure behind it, the data centres that store, transmit, and process our online activity. These centres generate heat and rely on water-based cooling systems, either directly on-site or indirectly through the power plants that supply their electricity.

A standard email might use around 0.1–0.3 millilitres of water, less than a single drop. But if you’re using AI to generate that email, the footprint increases. According to UC Riverside, an AI-generated email could use 2.2 millilitres for cooling and up to 47.5 millilitres when you factor in electricity, nearly a shot glass per prompt. TechRepublic estimates that if 16 million people used ChatGPT to write one email per week for a year, it would consume 435 million litres of water, equivalent to Rhode Island’s daily water use.

For small business owners, the individual impact is negligible. But it’s a reminder that even our most routine online habits are part of a much larger system.

5. Is It Really That Different?

If sending an email or posting on social media uses water, even just a drop, is it really so different from using AI?

At first glance, AI seems like the bigger culprit. It’s newer, more complex, and often framed as energy-hungry. But when you look closely, the infrastructure behind both is surprisingly similar. Whether you're sending a message, uploading a video, or generating a blog post with AI, you're tapping into the same global network of data centres, all of which require electricity, cooling, and maintenance.

The difference lies in intensity and scale. AI queries, especially those involving large language models, demand more processing power per interaction. But that doesn’t mean your everyday digital habits are impact-free. A simple email might use less than a drop of water, while an AI-generated one could use a shot glass, yet both rely on the same systems.

So the question becomes: how do we weigh the value of what we’re creating against the resources it consumes? If AI helps you write more clearly, serve your clients better, or reduce waste elsewhere in your business, does that justify the slightly higher footprint?

This isn’t about guilt, it’s about awareness. And perhaps, about choosing tools that help us do more good than harm.

Conscious Use, Not Avoidance

The environmental story of AI is complex. But for small businesses, the path forward isn’t about guilt or avoidance, it’s about conscious, strategic use.

By choosing cloud-based platforms (which are more efficient than local servers), and using AI to streamline your own operations, you can turn a personal concern into a positive force.

AI’s biggest contribution to sustainability may not be in reducing our individual usage, but in helping us solve the bigger problems, together.