Chapter One, Three Years Later

What I got right in the first chapter of Sustainable Web Design In 20 Lessons, what sharpened with experience, and what the industry is still refusing to face in 2026

I sat down the other evening with a copy of Sustainable Web Design In 20 Lessons in my lap, and I opened to the first chapter. It had been a while since I had last read it. I had not forgotten about it, of course, but when you finish writing something, you tend to let the words live their own life without constantly returning to them. You move on, and you trust that what you wrote was true at the time you wrote it.

Three years is a long time in this industry though, and when I read through Chapter 1 again, I wanted to know honestly how much of it still held up. How much of what I had written was still true. How much had shifted. How much I had simply not been able to see yet, because I did not have the experience to see it. This article is my attempt to answer that question in a way that is honest both to the reader and to myself.

I want to say up front that this is not a correction piece. The spine of what I wrote in 2023 is, in my view, still correct. The nuances around it have deepened though, and the last three years have taught me things I could only have learned by being in the field long enough for the picture to fill in.

The urgency did not disappear, it grew teeth

In the first chapter, I referenced the prediction that the internet would double its greenhouse gas emissions by 2025. I cannot tell you today whether that exact number landed on the nose, because I have not been able to find a clean dataset that directly supports or refutes the claim in that form. I want to be honest about that rather than pretend the number was either vindicated or wrong.

What I can tell you is what the last three years of data are pointing toward, and the picture is not calming. AI alone has added something in the region of 90 terawatt hours of additional energy consumption in data centers by 2026. That is not a small number. If you combine that with the fairly well-established assumption that data centers have been growing their energy consumption by roughly fifteen percent per year, then the consumption since 2023 has grown somewhere between forty and fifty percent, which translates into roughly 150 to 180 terawatt hours added to the load. That is not a doubling in the strict sense, but it is a very significant increase, and the curve is still pointing upward rather than flattening out.

And this is where experience has sharpened the picture for me, because in 2023 I was mostly talking about energy, and energy was already enough to tell the story. Since then, it has become more and more visible that energy is only one side of what data centers actually consume. They also consume water, and not just any water. Clean drinking water, which is often sourced through agreements with municipalities that can be negotiated cheaply, and which in many cases happens to be in regions that are already struggling with drought. That detail alone changes how you think about the cloud, because when a piece of infrastructure consumes fresh drinking water in a drought-affected area to cool servers that are running in large part so people can scroll through content they will not remember tomorrow, you start to understand that the problem goes well beyond carbon.

Then there is the hardware itself. The rare earth materials, the extraction, the manufacturing, the short lifespan of the servers that are replaced far more often than the average consumer realises. Every layer you peel back reveals another layer underneath. So when I read my own words from 2023 warning about the urgency, I do not feel the need to soften them. I feel the need to extend them, because the problem has more faces now than I originally had language for.