Sign in Subscribe

Is AI the Problem - Or Just the Headline?

Kristie De Garis

• 5 min read
Is AI the Problem - Or Just the Headline?

When one topic starts dominating headlines, especially with urgency, I pause. When the focus narrows, I get suspicious. That’s when I start asking questions.

Public attention is often managed: not always by lying (although increasingly by lying!) but by directing focus.

Right now, that focus is on AI.

If you think about it, we’re rarely, if ever, given a clear picture of what’s harming the planet. We’re handed fragments: scattered facts, competing headlines, partial data. Rarely one coherent message.

But lately, there has been a surprisingly coherent message: that AI is a threat to the climate. That it’s energy-hungry. Thirsty. Bad for the planet.

And all of that is true. But without comparison or broader context, it becomes a distorted picture. Incomplete narratives are misleading ones.

This isn’t a defence of AI. And it’s definitely not an attempt to rank one technology above another. It’s a challenge to the way we gather information, and the impact that has on how we see and respond to the world.

AI absolutely deserves scrutiny. From environmental impact to bias and surveillance to exploitative labour and extractive data practices, its risks are real. But those risks require precise, not panic-driven, critique.

This piece is based on publicly available data. It took me ages to research. Not everything about AI or streaming is disclosed, but what we do have paints a consistent picture.

Let’s talk watt-hours. They’re one of the clearest ways to understand energy use, not just how powerful something is, but how long it runs. A 2,000W kettle might sound high, but if it’s only on for two minutes, that’s around 60 Wh. Watt-hours aren’t something most of us think about, but once you see what they actually mean, it changes how you view everyday energy use.

Before this, I didn’t know how much energy anything used. Finding out shocked me and that’s why I’m sharing it. Because I realised just how much we’re not being shown.

To begin, we need a clear baseline.

Common Household Activities

·Boiling 1 litre of water in a kettle – ~112.5 Wh (based on 1500W kettle used for 4.5 minutes) 

Electric shower (10 mins) ~ 1,670 Wh

Tumble dryer (1 load) ~ 2,500 - 4,000 Wh 

Dishwasher (1 cycle) ~ 1,170 Wh

Oven cooking (1 hour) ~ 2,300 Wh – 

Fridge freezer (per day) ~ 1,300 Wh –

Television (LED, 4 hours) ~ 280 Wh –

Digital & AI Use
Streaming energy use varies massively depending on device, resolution, and infrastructure. Some newer estimates suggest 80 Wh/hour, while others put it closer to 800 Wh/hour for higher-res streams. For this piece, I’ve used a mid-range figure of 300 Wh/hour to reflect more typical use.

Streaming HD video (1 hour) ~ 300 Wh

Streaming 4K movie (2 hours) ~ 600 Wh

ChatGPT (20 queries/hour) ~ 6 Wh

To make these numbers easier to grasp, here’s what you could do instead with the same amount of energy. Not to minimise AI’s impact, but to show how easily we overlook the cost of the familiar.

What could you do instead with the same energy? (Based on average ChatGPT queries)

  • Boiling a kettle (112.5 Wh) = 22.5 minutes of Netflix (HD at 300 Wh/hr) or 375 ChatGPT queries
  • A 10-minute electric shower (1,670 Wh) = 5.6 hours of Netflix (HD) or 5,567 ChatGPT queries
  • Tumble dryer load (2,500–4,000 Wh) = 8.3–13.3 hours of Netflix (HD) or 8,333–13,333 ChatGPT queries
  • Dishwasher cycle (1,170 Wh) = 3.9 hours of Netflix (HD) or 3,900 ChatGPT queries
  • Oven cooking (1 hour, 2,300 Wh) = 7.7 hours of Netflix (HD) or 7,667 ChatGPT queries

Environmental Impact Notes: Streaming vs AI

Energy Use (per hour)

  • Netflix (HD): ~300 Wh
  • ChatGPT (20 queries): ~6 Wh

Water Use (per hour)

It’s important to understand that the 300 Wh/hour figure for HD Netflix includes the whole system, not just your device, but the global infrastructure delivering the content. Same goes for AI. Any tool used carelessly becomes wasteful. This piece focuses on everyday AI use, not its most excessive or novelty-driven extremes. Just as we wouldn’t judge ovens by someone heating their home by leaving the door open all day.

And then there’s social media. An hour of TikTok use produces around 175g of CO₂. More than Netflix. More than ChatGPT. These figures vary depending on device, network, and location but we rarely question them. Streaming and scrolling are generally seen as harmless.

We often compare one-time training costs of AI with the continuous, global churn of streaming. An unequal baseline.

Training a large AI model like GPT-3 reportedly used over 1,200 megawatt-hours of electricity and hundreds of thousands of litres of water. It’s a significant figure, but it’s also a one-time per version cost spread across billions of uses. And now, newer approaches focus more on fine-tuning than full retraining, making future updates to versions less resource-intensive. Meanwhile, streaming happens constantly, every hour of every day.

Scale and frequency matter. The moral panic around AI doesn’t match the wider reality of digital consumption.

So what’s going on?

Environmental messaging isn’t built for clarity. The facts aren’t hidden, but they’re scattered. Hard to piece together unless you’re looking. This isn’t just about missing context, it’s about how media is structured. Headlines are designed to provoke emotion, not invite thought. Social media rewards speed and outrage, not nuance or accuracy. And when we build our understanding from platforms that profit from distraction, we don’t just get half-truths, we get manipulated.

If we only hold new technology accountable, we let old systems off the hook. Streaming gets a free pass, not because it’s harmless, but because it’s familiar.

Streaming can be meaningful, sometimes even empowering. But structurally, it’s designed for consumption. It’s passive by default, profit-driven by design, and largely unchallenged in terms of environmental cost.

AI isn’t harmless. It is growing at an unprecedented pace. It doesn’t roll out gradually. Once trained, a model can be deployed globally, overnight. That’s destabilising, but it also gives people power. Power to read legal documents, understand policies, navigate knowledge previously locked behind gatekeeping.

AI can be biased, flawed, incorrect. And many of the companies driving AI development have poor ethical track records, on copyright, data privacy, labour practices, and environmental transparency. It shouldn’t be treated as an authority. And it definitely shouldn’t be treated as a toy. Generating nonsense images/content for clicks is a waste of energy and potential.

Used well, AI is a tool of access and exploration. That’s not nothing.

This is why critical thinking matters. If we don’t stop and examine where we’re being pointed, and what we’re being distracted from, we’ll end up policing the wrong things. We’ll feel guilty for using empowering tools, while ignoring high-cost habits we’ve normalised.

None of our energy use is benign. This isn’t about shaming individual choices, it’s about questioning the narratives we’re given, especially when they serve to distract from much larger systemic consumption. We treat energy as infinite because it feels invisible. But it isn’t. And we need to question more than just AI, we need to question who benefits when one thing is blamed while everything else carries on unchallenged.

Because the point isn’t just the carbon. It’s the control. If we want real accountability, for energy, for climate, for equity, we need sharper questions, not louder outrage.

Don’t take my word for it. I'm no expert. Go look for yourself. Read past the headlines. Compare. Contextualise. Follow the trail. And this goes for everything, not just AI! If you’re forming opinions based on tweets, TikToks, and headlines, ask yourself: who wrote them, and why? Who profits when you don’t ask questions?

Thanks for reading. I appreciate you being here.

If you like my words and images, consider sharing on social media or with friends who might enjoy it too. You can subscribe to my newsletter at the 'signup' link below. You can also read more about my book here and if you like the sound of it, you can buy it here.