It writes blog posts. Answers questions. Designs logos. Translates languages. Builds websites. Suggests which shoes go with your jeans. It finishes your sentences — and then politely rewrites them to sound more professional.
It’s clever. Instant. Magical.
Or so it seems.
Because behind that cleverness is something new.
Not totally new — the internet’s always run on electricity. It’s always had a weight problem.
It just used to be easier to manage.
You picked a green host. You compressed your images. Maybe you avoided autoplay videos. That was enough. The web mostly served up static pages, or the occasional bloated WordPress plugin.
But that version of the web? It’s quietly disappearing.
The new internet doesn’t just serve content — it creates it. On demand. For every user. In every moment.
And every time it does this
- Somewhere, a data centre flares to life.
- A GPU wakes up.
- Electricity flows.
- Carbon follows.
Training the model behind it? That consumed up to 62 GWh — enough to power the entire UK hosting industry for weeks.
This isn’t a glitch. It’s a shift.
Because the internet is no longer powered just by the infrastructure you choose.
It’s now powered by the infrastructure your tools choose.
The AI plugins. The smart chatbots. The content generators. The image makers.
And that means green hosting alone isn’t enough anymore.
It still matters. It’s still the foundation.
But if we don’t pay attention to what’s built on top of it, we risk swapping one kind of footprint for another — and telling ourselves it’s all the same shade of green.
🤖 The Fork in the Power Cable
At first, AI felt like a party trick.A chatbot that could rap like Shakespeare. A tool that rewrote your emails in pitch-perfect corporate nonsense. A logo generator that made your cat food startup look like it had a £10 million brand team.
Fun. Harmless. Smug in a look-what-I-can-do sort of way.
Then it got serious.
People started using it for real work — writing code, generating content, designing websites, sorting data, managing workflows.
It crept into the tools we already relied on: Microsoft 365, Google Workspace, Notion, Figma.
It showed up in WordPress plugins, ecommerce platforms, CRMs, chat dashboards.
Quietly, invisibly, AI became infrastructure — each layer a little smarter, a little hungrier.
And behind every one of those layers?
Servers. Compute. Power.
That slick little plugin writing your product descriptions?
It’s tapping into a model hosted in a data centre that hums like a jet engine and gulps electricity like a slot machine eats coins.
And those machines — racks, cooling systems, processors — don’t just run hot.
They’re starting to strain entire power grids.
In 2023, global data centres consumed more than 415 terawatt-hours of electricity — more than the UK uses in a year.
AI already accounted for nearly 20% of that. By the end of 2025, it could be close to half.
The scale of demand is outpacing the systems meant to support it.
Countries are scrambling to figure out where that power will come from — and how to keep the lights on.
Some are building new generation capacity.
Some are rationing industrial loads.
And in many cases, national power grids simply aren’t built to handle this — not at the speed AI is accelerating. And just as some nations were starting to decarbonise, they’re now being pulled back.
In the UK, we spent over a decade phasing out coal.
Year after year, plants closed. Wind turbines spun up. Rooftops gleamed with solar.
By 30th September 2024, the last coal-fired power station had shut down and for the first time in modern history, the UK grid was coal-free — a huge win.
But the timing couldn’t have been worse. Just as we stepped away from coal, AI’s demand began to spike like never before.
Meanwhile, across the Atlantic, the story took a darker turn.
In March 2025, the United States made its stance brutally clear.
President Trump signed an executive order that dragged coal back from retirement — and handed it a key to the server room.
Coal was reclassified as a “critical mineral”. Federal lands were reopened for mining. Coal-powered infrastructure was prioritised — encouraged — for AI data centres.
Yes — coal. For AI.
Not tolerated. Not transitional. Encouraged.
Because apparently, the best way to power a bedtime story about a hedgehog in space…is with a 400-million-year-old lump of compressed swamp.
None of this is being done with pause or coordination.
The AI boom has triggered a global sprint — to build bigger models, train them faster, deploy them everywhere. And in the race to dominate, energy use has become background noise and even tech giants like Microsoft and Google have admitted their AI ambitions are derailing their own climate targets.
And as for transparency? It’s limited.
So for now, we keep burning the past to build the future.
That’s the fork we’re facing.
One path leads to clean grids, smarter deployment, and long-term benefit. The other? Burning ghosts to train machines.
And right now we’re sprinting down the wrong path — and calling it innovation.
AI isn’t the villain. But it’s not the hero either.
Not yet.
But the plug is still in our hands.
And the next connection we make could decide the story.
🌎 Building the Web We Actually Want
AI is here, it’s growing fast, and it’s consuming power at a rate that would’ve seemed absurd five years ago. And unless something changes, it’s going to set fire to decades of progress toward a more sustainable internet.But it doesn’t have to.
At Kualo, we believe it’s time to stop pretending that AI is somehow separate from the carbon cost of running it. And it’s time to start asking harder questions.
Because the future of the internet is being shaped by what we build today — and by what we choose not to build.
So what does it look like to build better?
It starts with infrastructure. Renewable energy should be the baseline — not the marketing headline. But we also need to go further: ultra-efficient cooling, intelligent load balancing, smart code, less bloat. It’s not about abstaining from progress. It’s about choosing the version of progress that doesn’t cost the planet.
Our East London hosting facility pushes for PuE figures that compete with the best in the world. It uses systems that use the air outside, instead of refrigerants inside. We’ve made cold air containment not just a feature — but the default.
But we can’t do it alone.
What can you do?
If you’re a business, developer, marketer, or agency using AI — or thinking about it — here are questions worth asking:
- Where is this model hosted? Most AI tools run on centralised hyperscale cloud platforms. Some of those use a lot of renewable energy. Some don’t. Some are transparent. Some aren’t.
- Is there a more efficient alternative? Tools like Claude, DistilBERT, or open-source models can offer lighter, faster, and greener performance for many use cases.
- Do you need AI for this? A good FAQ page might outperform a chatbot. Use AI where it adds real value — not just where it looks good in a press release.
- Can you run it closer to home? A growing number of models can now run on-premise — without relying on external APIs. That means no constant inference calls to third-party clouds, no long-haul data travel, and no unknown carbon footprint. Hosted locally — perhaps even with us — these models offer real control, real transparency, and a dramatically smaller impact.
- Are you supporting digital infrastructure that reflects your values? Hosting on green-powered platforms matters — not just for the base load of your site, but because every watt saved in your supply chain frees up capacity that hyperscalers won’t get to call theirs.
Can AI help solve this problem?
In some ways, yes. AI can help optimise power grids. It can design better wind turbines. It can compress video streams, shrink codebases, automate energy forecasting, and reduce the carbon footprint of other industries.Some models are being trained specifically to fight climate change — from satellite analysis to carbon capture forecasting. And they matter. They’re necessary.
But let’s be clear: those use cases don’t cancel out the footprint of generative AI. They can offset it. They can even justify it. But only if we choose to build AI that fights for something bigger than engagement metrics.
If you care about sustainability, privacy, equity — the future of the internet — then vote with your infrastructure. Choose tools that align with your values. Ask hard questions about where your AI runs. Use it where it makes sense — and don’t be afraid to say no when it doesn’t.
And beyond that: speak up.
Because in the years ahead, energy policy will be tech policy.
What we build — and what we power — will define the web for decades to come.
So… will it ever be enough?
Only if we choose to make it so.
