AI is having a moment. Every week brings a new headline about how much energy AI models consume, how data centres are straining power grids, and how the technology might be an environmental disaster. Meanwhile, nobody seems worried about the hours of Netflix they watched last weekend.
But here is the thing: per use, streaming video produces far more carbon than an AI query. The conversation about digital carbon is focused on the wrong villain.
The numbers, side by side
Let us put the main digital activities next to each other:
The contrast is striking. One hour of Netflix produces roughly 42 times the carbon of a single ChatGPT query. Even if you asked ChatGPT 40 questions in a row, you would still produce less carbon than watching one episode of a TV show in HD.
Why streaming is so carbon-heavy
Video streaming is expensive because it involves massive, continuous data transfer. When you watch a show, your device is constantly downloading high-resolution video frames, thousands of them every minute. That data travels from a content delivery network through your internet provider to your device, and every step in that chain requires electricity.
HD video typically transfers around 3 GB per hour. 4K pushes that to 7 GB or more. Each gigabyte needs servers to store it, networks to move it, and your device to decode it. The energy adds up quickly.
By comparison, a ChatGPT query involves a relatively small burst of computation. The model processes your text input, generates a response, and the interaction is done in seconds. The data transferred is tiny compared to video.
So why does AI get all the attention?
A few reasons. First, AI energy use is growing extremely fast. Major tech companies are building new data centres at an unprecedented rate to handle AI workloads, and projections suggest AI could consume 3-4% of global electricity within a few years.
Second, the energy cost per training run for large models is genuinely enormous. Training GPT-4 reportedly consumed the equivalent of thousands of households' annual electricity use. But training happens once (or periodically), while the individual queries that follow are relatively cheap.
Third, AI feels new and unfamiliar, which makes it an easy target. Streaming has been normalised. Nobody thinks twice about watching three hours of video in an evening, even though that single session produces more carbon than a week of casual AI use.
The bigger picture: data centres
The real issue is not AI versus streaming. It is the overall energy footprint of data centres, which currently consume about 1-2% of global electricity and are growing fast. Here is the uncomfortable fact: only 30% of data centre energy comes from renewables today. If that figure rose to 80-90%, the carbon footprint of every digital activity, streaming and AI alike, would drop by more than half.
The technology matters less than the energy source powering it.
A practical tip: lower your streaming resolution
Dropping from 4K to 1080p can reduce streaming emissions by up to 86%. On a phone or tablet screen, you are unlikely to notice the difference. On a laptop, the difference is minimal. This is one of the simplest, most effective things you can do to reduce your daily digital carbon footprint.
You can also:
- Download content over Wi-Fi instead of streaming over mobile data (Wi-Fi is more energy-efficient)
- Avoid leaving videos playing in the background
- Close unused tabs and apps that auto-play video content
The bottom line
AI queries are not the carbon problem they are made out to be, at least not on a per-use basis. Streaming is far more carbon-intensive for the average person. But the real solution is not to stop using either. It is to push for cleaner energy in data centres and make small, practical changes to how we consume digital content every day.