I run Claude Code most of my day. agent loops firing all day, one after another. the usage screen tells me I'm at 26% of my weekly limit. it doesn't tell me what that 26% weighs anywhere else. it feels free.
It isn't.
A single chat reply costs about 0.3 watt-hours — the energy an LED bulb burns in under two minutes. Simon P. Couch measured actual Claude Code sessions and got a median of 41 Wh — 130x a plain message. Simon Willison put his own daily habit at the equivalent of 4,400 typical queries. His comparison: a dishwasher, or a fridge running 24 hours.
KAIST measured agents specifically this year not chat, agents that reason and act across steps. A 70 billion parameter agent — a model roughly the size of what real commercial products run on, not a toy — averaged 348.41 Wh per query. 136.5x a normal query. Scaled up, researchers project close to 200 gigawatts globally just for agents deciding what to do next. A gigawatt is what a small country's whole grid can push out at once.
One dev. One agent open all day. Willison's own number: 1 to 1.5 kWh on a heavy day. Not a lab estimate. His actual habit.
Average Nigerian: 165 kWh a year. 0.45 kWh a day. Lights, a phone charge, a fan through the heat. That's the whole ration.
I'm writing this from Port Harcourt. Friday, I was on a video call with a product manager at Cloudflare — he was walking me through a demo, asking what I thought. It started raining. The light went. The internet went with it, mid-sentence, no goodbye. That's the wire I'm on, talking to someone at a company whose servers never blink.
Building the reconciliation agent for the hackathon that same week, I caught myself mid-build — one stretch, maybe a few minutes, and the token count had already crossed 10k. I wasn't even done. That's one exchange, one agent, one afternoon, running on the grid that dropped a call over rain.
That night I googled it. how much one of these sessions actually costs, in power, not dollars. the number that came back stopped me mid-scroll.
One coding session. One dev. More power than an entire day of an entire person's life, in the country I'm typing this from.
Gartner: global data centers hit 565 terawatt-hours (TWh) this year — the unit whole countries use to measure their power grids. AI-optimized servers alone not data centers overall, just the AI slice — went from 95 TWh in 2025 to 175 TWh in 2026. An 84% jump. One segment. One year.
Nigeria: 40.7 TWh a year, total, country-wide. Less than half of installed capacity reliably reaches anyone.
AI-dedicated servers alone will burn more than four times what my entire country produces. This year.
It's not coming off Nigeria's grid. Different continent, different wire. Nobody dims Port Harcourt to run my prompt.
Cheap for a dev paying $20 a month for Claude Pro. Cheap for a hyperscaler burning gigawatts to keep a chatbot responsive. Nowhere near cheap for the 61% of Nigerians who had grid access in 2022 or the rest who never did.
I needed the work done. I'm not apologizing for the sessions.
I am done pretending they cost nothing.
Every agent loop you fire off casually costs more than someone else's entire day, somewhere, always. You just don't see the meter.
I saw mine. It didn't look like a coding tool anymore.
I'm half sick writing this. And some of you reading it will paste it into an LLM to draft a smart-sounding reply for the comments. That's the loop. That's the whole joke, and it isn't funny.
Top comments (84)
I don't want to be a luddite, but what are really the pros and cons of this whole AI hype, can we have a nuanced (or maybe not so nuanced) discussion about that?
'Pros':
some people get filthy rich (is that really a pro? I've put it here as a "straw man's argument", lol)
we can spit out code 10 times faster with AI (does anyone actually need that much code? are we still able to understand how it works?)
managers at companies are getting so happy, coz they can fire people and cut costs (at least that's what they think ...)
'Cons':
the toll on the environment is insane - not to mention the "mental toll", which has been the topic of quite a few articles here on dev.to ...
computers, smartphones and game consoles have been getting a lot more expensive, because AI data centers gobble up all the RAM chips
widening the already huge gap between haves and have-nots (the point you're talking about mainly)
But I'm sure AI can be beneficial to mindkind - curing cancer, fundamental scientific breakthroughs, solving pressing societal issues ...
It can't be stopped really, but it all comes down to: what will we use it for, and how will we use it?
In fairness. Nothing is stopping you from creating 30 google accounts and switching between them to make the most out of AI Studio and Antigravity. So the have-not's still have, if they're willing to be smart about it.
As for the filthy rich part, yeah... But worth noting those people are the investors in these AI companies and you can always invest yourself. What it has done though, is allow startups to scale FAST and people from grass-roots build a functional business with freebies like Azure and Google Cloud startup trials ($1000 and $300 respectively in cloud credits).
It has opened doors for people, if you were an app dev, but knew nothing about cloud run, you can finally deploy to cloud.
Suddenly language specific tasks become specialty agnostic. You dont need to be able to write a F# function to understand it, it makes sense reading it, but to conform to the strict requirements, it's a massive undertaking to learn (it took me 5 years just to call myself 'proficient', nowhere near expert).
As for the environment, yes, it uses ALOT of power, it takes up alot of real estate and it uses alot of computers, kicking out alot of heat. But consider this... Data Centers use water to cool them, that water is vaporized and emitted into the air. Generally speaking, the piping used is copper for better heat dissipation which isnt all that bad, it's anti-microbial. That water vapor emitted raises humidity and in turn either fixes greenhouse gasses and drops to the earth (imo, marginally better than building up in the outer atmosphere), humidity also raises chance of rain. Right now (I live in a desert), the world needs more rain, so that aspect, it's actually the first time we're producing more heat, but the outcome is actually good?
The way I see it, is AI, is like computers. When they came, it created a divide, with 1 difference... When computers came out, you needed to fork over half a year's income just to buy 1. Whereas to access AI and use it, you dont need a cent, just a phone or a computer. When you look at how the industry is changing, existing companies are leaning down, while startups are scaling up. The shift in dynamics is essentially a goliath slaying event, where small companies scale fast enough that they can actually compete against giants. Take the 12 part series I did, it took me just about 5 days to scale from building an extension to building a full-fledged OS running it's own custom language at a kernel level... That's like the guy with the computer designing a fighter jet in a week, vs pen and paper took a year.
More code isnt necessarily better, but when you look at the displacement caused by it, people are actually doing unit tests now... People are actually securing their endpoints now... People are shifting how they design, because their fatal flaws were exposed by AI. That's global growth, accessible with a free account. The only real burden of it, is whether you choose to embrace it, or ignore it till you're obsolete? How many pen and paper employees rejected computers, till they became the industry standard in every single profession. That's how it's going with AI.
Water vapor is actually a pretty potent greenhouse gas. More water in the atmosphere contributes to runaway climate changes, which, in turn, makes weather phenomena like rain more extreme. Heat waves cause extreme droughts, leading to a sudden increase in water in the atmosphere (rivers and lakes evaporating), which leads to extreme rain fall, which can't be absorbed into the dry ground properly (because it's too much at once), leading to surface flow and - in turn - to inland floods, causing entire cities to be flooded.
Also, the energy consumption isn't free, either: Rapidly increased energy consumption is often met with burning more fossil fuels, because building new nuclear power plants takes ages (not to speak of the radioactive waste produced, the amount of energy necessary to build it and how we get the uranium out of the ground to begin with) and most countries are somewhat slow with implementing solar power and wind power because they don't like how it looks.
The heat generated is the least problem here. It's how we deal with the massively increased energy consumption. The fact that everybody and their mother are building server farms like crazy to cover the immense token consumption doesn't exactly help the climate situation.
Pascal, the backfill point is the one that matters. New demand doesn't wait for clean supply, it gets met with whatever's fastest to bring online, and that's still coal and gas in most grids.
I studied geophysics before any of this . the water cycle math you're describing isn't new to me, just usually not the frame people reach for when they're impressed by a chatbot. Nigeria's 40 TWh a year isn't landing on any of these plants either way.
About this climate issue — I feel like I've been hearing it since I was a kid. Back then the news was flooded with the Antarctic ozone hole, global warming, greenhouse effect. How come you barely hear about them anymore?
Pascal. well said, very well explained - I noticed the remarks about water vapor, and already thought it didn't make sense, but forgot to comment on it ...
I don't doubt the potential benefits of AI, but when are we, as humanity, going to prioritize saving the environment on our precious little planet, our only home?
First things first - AI is great, but right now its environmental fallout is horrible - I wouldn't mind at all if we decide on a moratorium on the AI datacenter craze, and first focus on saving this planet - well I know it ain't gonna happen, but one can dream ...
P.S. gotta disagree with the suggestion that climate change is a fad, just because some disasters which were predicted haven't happened yet - I think we're still underestimating the scale of what's coming ...
Just to clarify — I wasn't saying climate change is yesterday's news. I meant it feels like the internet and the news don't cover it as much as they used to. Or maybe some of it is being deliberately filtered out. Yeah, I know it sounds a bit conspiracy theorist-y.
Well you're right about that, and it's a fair observation - it's just not that much "in vogue" anymore as a topic ... I think that says more about the short attention span of "us mankind", than about the seriousness of the issue!
Those are fair points, though a data center doesnt go anywhere in 100 years (or atleast they dont seem to be anyway). While the sudden change is a shock, it's no different than Mumbai getting clear skies for the first time in years during Covid lockdown. Sudden shocks cause ripples, even if they're good ones. And yes, the immediate result is flooding in deserts, but that's not always a bad thing. Here in Namibia, farmers want a flood, because it disperses the seeds, washes up all the dried manure and causes deposits where grass grows and settles the moisture.
True on the no free dinner for energy, though that's where governments have been pushing hard for green energy, especially for data centers. Places like China use hydropower. Take Namibia for instance, we dont really have any clean power plants, but they want to build a data center here, on the coast, so the climate is humid and cold, there's a water desalination plant right next door and the oceanic fiber line running through here too. The 1 missing link, power... That's 1 of the key points brought up when they brought it up for community feedback and the current push is to force the data center to build their own solar plant to power it. So if a country in Africa with sub 3m people can restrict it to being a clean solution, there's no excuse anywhere else in the world.
UnitBuilds, that's the counter-example that actually matters. But Namibia had leverage most sites don't — fiber and desalination already sitting there, a location worth negotiating over. Nigeria's grid isn't sitting on anything a hyperscaler needs.
That's the difference between a country that can set terms and one that just gets billed for the shortfall. What made the community feedback process actually stick, was it the site value or something in how it was structured??
It's true though, climate change always has a 'big bad' that gets swapped out for the new evil every few years. First it was cars, then it was aerosol cans, then data centers, now ai-data centers. All valid and none have gone away in the slightest. But lets take a moment to appreciate that in the few years AI has been here, it's actually led to discoveries that improved efficiency, whereas all the rest have done nothing but harm? Not saying AI data centers dont do harm, far from it, but atleast there's a silver lining. Take cars for instance, take normal data centers for instance, you telling me that facebook back in the day adding a line of server racks a week had any benefit? Aerosol cans atleast means the DMV doesnt reek as bad and cars mean you get to work on time, but those havent directly improved anything, in terms of net efficiency. You move a 2 ton piece of steel for 5km, to move tiny little 60kg you to work? That's hardly efficient, nor economical. Whereas AI boost your output, whether you're pumping out memes, or discovering a new way to harness the sun. It's by no means a clean tool, but atleast it's a tool that has the potential to solve it's own problem. A car cant derive it's own efficiency shortcomings and improve it, nor can it magically make itself faster. Humans did that over thousands of hours, whereas AI turns those thousands of hours into a few MW. We're in a race against time with climate change, we need to race to undo hundreds of years of bad decisions, all at once... AI is the solution to that. In just a few years, it's made breakthroughs in every single field. That's something nothing else has achieved and it's worth noting that amongst all the AI generated content on TikTok. You cant expect everyone to use a gun to hunt, nor can you expect everyone to use AI for genuine research. But the 10% that do, have done some incredible things with it and those discoveries have improved the situation globally.
Unit, the self-improving part is true and it's also not the same ledger. AlphaFold's efficiency gains sit in the research column.
The 84% jump in AI server draw sits in the deployment column. Nothing says the second column shrinks because the first one grows . right now they're both just growing.
Cars didn't get cleaner because a lab somewhere improved combustion theory either until that theory got mandated into every engine. The breakthroughs are real. Whether they get forced back into the infrastructure that's burning the power is a policy question, not a technology one.
Well doesnt that just mean that the data center should be built in Namibia and not in Nigeria? The reason why it stuck in Namibia, is for the same reason Starlink got turned down. Not because of valid community input, but money... If government needs to provide power, which we buy in from other countries, it costs government more, when all they want is to actually just tax them. So people brought up the cost of up'ing our power buying, vs the labor and taxes on building a solar farm. Easy win, just give government more money...
UnitBuilds, that's sharper than "leverage" . it's a tax stream beating a power subsidy on the government's own ledger. Nigeria's math would flip: no subsidy to protect since the state already under-supplies power. Same instinct as the Starlink call, different asset taxed.
Well, as a case study, look at Norway. Instead of charging for drilling rights for oil, they simply taxed the companies who want to drill it. They took that money, put it into a fund, that fund is reaching close to $2T now... The money is there, the means to get it easy, but lets face it, greedy politicians are what prevents things from being done right. Instead of demanding that the data centers be built self-sustaining, Nigeria put more strain on their own system, which in turn put more strain on the people... Not by popular demand, I imagine. But in Namibia, people have learnt that the way 2 voices outweigh 1500, is money... Not by having money, but by pointing it out. Starlink is honestly something Namibia desperately needs. South Africa opted that instead of forcing local ownership, that Starlink must invest in the country, so they're offering free internet for rural school. Namibia on the other hand turned it all down, just because they want more money out of it. If starlink cant become a local provider, they have to be a provider to local companies, meaning double tax... That's how it got shot down and why we'll likely end up paying double what we should be paying for it. It just takes 1 person to point out the most profitable route and government will see nothing else.
Norway taxed instead of charged and built a $2T fund. Nigeria's oil revenue took the tax route too, and the fund isn't $2T, it's mostly gone. Same mechanism, different guardrails. The lever exists here already .it's just never been pointed at a data center
It's very effective, but the governance is what's the difference. Norway took that money, put it into a fund and had a publicly auditable company manage the funds. If they dont do a good job, they get fired, new people take their place. Africa is unfortunately sorely lacking in transparent government behavior. Money spent wisely keeps receipts and an audit trail. Money wasted is where the ledgers run dry without a sign.
think this one pasted back at me by mistake . what were you meaning to say? 🤣🤣🤣🤣🤣
🤣🤣🤣
copy wrong,sorry
Just wanted to chime in. I'm not talking about some cutting-edge AI breakthrough here — take what X/Twitter did recently. AI basically tore down the language wall. People from different regions, speaking different languages, can now see content on X in real-time, in their own language. So many barriers are genuinely being broken down. It's no longer that clunky garbage translation from before — it can actually convey what the original poster meant. That's the most direct benefit of AI I've seen.
That's actually pretty solid real-world impact tbh, I mean look at Pascal (the other 1), he's french and used AI to translate his thoughts when we were having the discussion that lead to the 12 part series. How'd we have communicated without AI, google translate? We wouldnt have even bothered.
I'm logging off to sleep, everyone. Not sure whether to say good night, good morning, or good afternoon — maybe that's the annoying part about globalization.🤣🤣🤣
@dannwaneri - absoutely, yes, that's why I mentioned new demand being mostly met with burning more fossil fuels. Fast growth no matter the price is usually paired with short-sightedness for long term impact and a "we'll fix this later" attitude, and all these new datacenters are in this category, sadly. Your post is the kind of wake-up call people actually need. Thank you for the calculations and bringing it up!
@unitbuilds you've got valid points for the good that all these developments have brought, just like @xulingfeng did with their remark about the language barrier, but I have a strong gut feeling that most AI traffic isn't being used for research or to actually improve the pressing global issues we have. Often it's simply used to increase stakeholder value. I have no data to back up this claim, though, and I know that anecdotal evidence is no evidence at all, but AI is used for all sorts of things besides research. And that's the core issue. The technology is amazing, don't get me wrong, but what issue does generating tons and tons of pictures of people as Studio Gibli super hero figurines that make them chuckle for 2 seconds actually solve? How do AI generated ads for AI-generated slop knockoff fantasy stories narrated by AIs improve the world?
Re climate change and AI datacenters in general: I personally read this as the main message of the original post. The energy consumption is staggering, and even though it feels "free" to us, someone eventually will pay the price for it.
--
EDIT: Re Good morning/night: tilde.town/~kirch/UGT.html ;D
Pascal, that's the gap the piece didnt mention not whether AI datacenters are wasteful but what the waste is actually producing.
Ghibli-filter traffic and research traffic burn the same grid at different rates of return. No dataset splits that out cleanly, which is convenient for whoever doesn't want it split out.
But if we filtered the 99%, we'd have such a condensed set that it'd be too rigid for truly creative work, or be priced out of reach of most. For the 1% to do great things, they need the 99% of garbage to fund it and keep it accessible.
Spot on.the garbage isn't a side effect, it's the subsidy. But that only holds if the 1% actually needs the 99%'s volume to stay funded not just convenient to have funded.
Stripe doesn't need meme generators to keep the lights on. Whether AlphaFold needs Ghibli filters is a different claim than whether the same company profits from both.
The real cost comes in 2 forms, upkeep (they're losing this 1) and R&D (they're praying this 1 makes up for it, before it collapses). With the only real profit to be had, being investments. Even now, when you look at the combined cost of using eg. Claude Sonnet to write a page of code, outweighs what you pay. But through you, 10 new people join, investors see 11 people using it and invest, because it's an 'emerging industry' and will likely grow considerably over the next 10 years. When you consider that, the 99% is vital, because people dont invest into AI because of research breakthroughs, they invest into AI, because they see 9/10 videos on tiktok are AI. Take ChatGPT for instance, it started off as a fad, I'm pretty sure you were as skeptical as I was about why they made it. We finally had AI, now people just use it to ask stupid questions they could just google... But that's exactly what gave them the coverage they needed to get proper funding to expand.
That changes the whole subsidy point . The 99% isn't funding R&D, it's funding the appearance of adoption that gets R&D funded. Investors aren't pricing AlphaFold.
They're pricing TikTok penetration. Which means the garbage isn't a byproduct of the business model, it's the product the business model is actually selling to investors. The breakthroughs are the side effect people point to when asked to justify it after the fact.
Exactly. Real advancements are funded by marketable fluff. Take RTX graphics, the world sees ray-tracing, while researches see matrix optimized cores. You invest because who doesnt want better looking games, while that investment funded the onset of AI research and development. By making it cheap enough for mass consumption, it opens the door for researchers to make advancements they could never have otherwise afforded. If the launch of RTX wasnt such a massive hit, Tensorflow would have died right there, now it's all but the industry standard for ML.
Well that's true, you have a point there:
"Real advancements are funded by marketable fluff"
More often than not it works like that - also, often, in research and tech (and science), you stumble across inventions or discoveries that you were not looking for (there's a word for that, can't recall it right now) ...
Science actually has way more examples, eg. that company researching malaria treatment that made a skin serum to fund the research.
Or the way Fleming discovered penicillin while he totally wasn't looking for it, classic example ...
Or a bit more of a fuzzy fit, Coca-Cola being derived from a solution to tummy aches.
Fact is if it's a niche product, you have to sell it at a premium markup to make ends meet. If it's a general goods, you can lower your price and scale production to make the money.
Or newspapers actually turning a profit selling ads, not papers. Sometimes waste isnt a byproduct but a means of diluting the cost across a broader spectrum that makes the unattainable accessible. Apple as an example of keeping it 'premium', whereas Nokia halo'd the 'let them have phones' era, where mass-scale turned them into a giant. If it wasnt for that push for scaled adoption, phones would still all be priced at iphone premiums. The same goes for AI, if it wasnt for the general assistant aspect, the simple summaries, the meme generating, etc. The technology would be far too niche to properly fund research and to properly price accessibly. So they sell meme generators to shift it to economies of scale, rather than charging $500 a month for your Max tier subscription. If you look at the time-saving that you get out of $20 of AI subscription, the speedup would have demanded a 10x premium, because it turned a week's work into an afternoon. Logically, it makes sense, it's still a good deal, but by doing that, they'd ostracize every person who wants to just try it and ostracize everyone who cant afford the premium.
Idk if you've ever tried AI via an API and compared work done:price for subscription vs pay as you go? Pay as You Go, is where the companies actually make a small profit, Subscriptions lose them money on day 1 of the month. But that loss is offset by the investment offset from it. The same old economical bubble that caused the Big Short. You over extend to make it accessible, but with no plan on how to recoup the losses, but it's fine, because the market valuation just goes up and as long as adoption continues, investing continues and the 2 together are sustainable... Until people invest somewhere else.
Take Tesla and SpaceX for instance, Trillions of dollars, but produces less cars than Ford? Both make cars, both make EVs, but Tesla is marketed as the 'savior for all man-kind', whereas Ford is the beat-em-up workhorse. Tesla is for 'visionaries' and 'forward thinkers', while Ford is for the everyday worker. We can all afford Ford pricing for AI, 99% of us wouldnt if it were Tesla pricing.
Yeah it's funny how a lot of the "added value" or competitive advantage seems to be in marketing and economics, rather than in purely technical innovation ...
P.S. this article must be in the op few % in terms of generating # of comments!
That's pretty much why we get cloud storage, image generation, code assistant, chatbot assistant, email assistant, etc. bundled together in Google's AI plans. They took the note from Microsoft, you sell a package for everyone, that way everyone gets bloat they paid for, despite the fact they'll never use it. Access, OneNote, Publisher, Teams, PowerPoint, all things you pay for, despite never using. Because they'd need to raise the pricing of Word and Excel if they were to be sold separately. That's why we get bundled meme-generators with our code assists
Well, I'm impressed how you articulated these advantages - whether or not all of it holds up to scrutiny I cannot tell, but as a response it's impressive ...
I'm totally convinced that the potential benefits of AI are huge, it just depends on how we use it - I'm waiting for people to start doing something REALLY useful with AI - medical and scientific breakthroughs, or technical ones that would enable AI to drastically cut its own energy and water usage - let's talk again when we see that kind of thing happening, rather than it only driving up the share prices ;-)
P.S. if you're enthusiastic about what you can technically build with it (and in a pretty affordable way, if you're a bit clever), yeah I totally "get" that, sure it can be powerful ...
Well it already has? MoE LLMs were designed to be more power efficient, while improved membrane designs have increased efficiency for water desalination plants, refined wind turbine designs, mathematical advancements have had some millennium problems solved with AI (Navier-Stokes), AlphaFold has had some incredible breakthroughs too (Protein folding).
It's easy to mistake global upscale with inefficiency, when that's the only reason we havent hit a wall yet.
Well those are indeed great examples, happy to hear that people are doing that kind of stuff with it!
UnitBuilds, efficiency and aggregate growth aren't in tension . they're compatible. Gartner's 84% jump in AI server draw happened on top of MoE-era efficiency gains not despite them. Same with the credits: free tiers move the cost off the developer, not off the grid. That's the gap the piece pointed at not whether AI can do good things, but who's holding the wire when it does.
Fair, I guess it's the same PC dilemma again. Before computers, we used very little power, then suddenly every household has 1, data centers, etc. All using a ton of power. Yet, it brought us the LED light bulb. Small graces along the way hardly offset the gross and the small efficiency gains dont suddenly cascade till they outweigh the cost. But I think it's a fair thought that unlike smartphones that go to landfill, AI might be the first human invention that can turn itself around. MoE for instance was a massive improvement in power consumption, but it's dwarfed by the increase in users and concurrency with multi-agent workflows.
Just a disclaimer, I'm by no means a fan of the power draw, especially not where they draw it from... But I do believe someone has to occasionally play Devil's advocate and shift the blame from the data center to the people using it (more specifically how they use it). If we were to switch to everyone running a local LLM, it wouldnt make the problem disappear, it would just disperse it to where the average household consumption goes up, instead of keeping it nucleated where solutions can be applied to mitigate costs (eg. higher concurrency per server node), to reduce the overall usage.
But in the end, Facebook didnt add servers a day, just because they needed more compute, it was a direct result of increase in users. While data-centers arent the best solution, I do think though that's a local government problem. Take China for instance, some of the largest data centers in the world, hydro-powered and hydro-cooled, which imo is atleast a cleaner solution than running massive AC's and just kicking out dry heat.
Wasn't trying to litigate AI overall, leob. Just what one session weighs where I'm sitting...
I know you weren't, but I am, lol ;-) - was kind of jumping on the opportunity ...
Not blaming AI for anything either - AI is innocent, it's what people are doing with it, especially the hyper capitalists over there in that-place-you-know-what-I-mean ;-)
/end of rant
Fair rant. No argument here...
I don't want to loose my coding skills: not only writing, but reading comprehension. Reasons being:
So I think iA is useful but we do not need it everywhere. It's really crazy how they want to push it to ubiquity, but it just doesn't make sense, and is crazy they convinced us so that we want to help them achieve their commercial imperative.
Is just like IOT in the sense of putting computer everywhere. I don't wan my f..n fridge to compute, it just supposed to cool or freeze its contents. I don't wan my toaster to compute, neither my washing machine nor my car, nothing else other than my PC or wherever it might really be necessary and make sense.
I think it's hard to create something with the tools we were taught (a hammer) and build a sustainable/profitable biz model around it and so we start thinking hard where to use our hammer, in time and trough great struggle we (as creators or entrepreneurs) start to see nails every where and convince everyone that hammering is useful even on the f'n soup.
totally agree....it's a two faced concept(coin).....its what we do with it that matters
The real tragedy here, in my opinion, is that many of us (myself included at times) aren't using the most appropriate tool for the job. We're flying a Boeing 777 200 miles to deliver a single passenger, when a Cessna 172 or a Honda Civic would be equally as effective.
The cloud-default design that has emerged routes surprisingly simple tasks through multiple network hops before they arrive at huge computer power for processing. Things like reading files, making small edits, validating data, running basic code analysis are often workloads that could be handled much closer to where the data already exists.
It makes me even more convinced that "local-first" is not just a privacy and security concern. It's becoming a thermodynamic necessity. A sovereign system escalates computation only when the complexity of the task justifies the cost of the escalation.
If we can run SLMs locally, get good results, and use a fraction of the energy of the API call, isn't that a win? I'm seeing a future in which we aren't cloud-only or local-only but a more thoughtful matching of the physical compute power needed for the task at hand.
Great post, Daniel. Thanks for sharing it.
Ken, the escalation point is the mechanism worth sitting with . most tasks don't need the 777, they get put on it because routing everything through one API is simpler to build than deciding per-task.
That's not a local vs. cloud question, it's a "who pays to make the decision" question and right now nobody's paying for it, so the 777 flies every route....
Another reason why I think using AI for every little task is simply wrong. It's just the same as Taylor Swift going on her plane to Paris to get a dress for a concert. In the end, it all comes down to your own sense of morality and what you can resign from for the bigger sake. Like resigning from fast fashion or makeup to not support sweatshops and child labor. I think AI brought a new level of moral dilemmas.
This trend I also cannot understand.
"You just don't see the meter" is the line. But a lot of that 41 Wh isn't the work, it's the loop being sloppy, re-reading the same file, retrying, dragging full context every step. Token and watt are the same meter, and most agent runs I instrument burn 2 to 4x what the task actually needed. Won't fix the Port Harcourt grid from my keyboard, but that waste is the one thing the person firing the loop controls.
Vinicius, the 2-4x is the part I haven't instrumented and should. But there's a ceiling on how much of this is fixable at the loop level .
you can cut re-reads and retries and still be running a 70B model for a task a 7B could've handled, still routing through us-east-1 because that's where the API lives. Waste inside the loop is yours to fix. The floor under the loop isn't.
Fair. Region I concede flat, you use the API where it lives. Model size is the one I'd still call half yours, right-sizing which step runs on which model is a caller call most loops skip. But the floor is real, and the honest version of my point is just: measure it so the floor stops being the alibi for the loop.
Vinicius, fair enough. model size is the one I've actually skipped not just underused. Most of what I run defaults to whatever's already warm in the session rather than what the step needs.
"The floor isn't the alibi" is the sentence I'd want in front of me next time I reach for the bigger model out of habit instead of requirement.
What if I told you I just used an LLM to translate your article? Would you believe me?
After reading your piece, all I can say is — everyone really is standing on very different ground. There's an old saying where I'm from: Trees die when moved; people thrive when they move. We still need to invest more in infrastructure. If you want wealth, build the roads first.
the translation makes the point better than the article did . you used the tool to close distance, I used it to notice how far the grid still is from doing the same. I'd flip the order tho: roads don't get built because people migrate toward them, they get built because someone pays for the wire before anyone needs it yet.
The title caught my attention.
Nice analogy BTW!
Worldwide, over 51,000 of us have mitigated that waste by up to 20x, keeping over 50 tons of CO2 out of the atmosphere and avoiding the burn of over a third of a **trillion **tokens. And that's only since March.
I won't turn this reply into an infomercial. I'll just offer this: It doesn't HAVE to be this way...
-J. Gravelle
GSF Certified Green Software Practitioner - jMunch, LLC
The part about not apologizing for the sessions but being done pretending they cost nothing — that's the only honest position to take. The discomfort is the point.
I've been building agent tooling and thinking about efficiency in terms of latency, token cost, and accuracy. Energy is almost never in the design constraints. Reading this made me realize that 'cost optimization' in agent design has a dimension almost nobody is measuring at the individual session level.
The 130x multiplier between a chat reply and an agent session is the number that sticks. It suggests most of the energy isn't in generating tokens — it's in the reasoning loops, the retries, the tool calls that go nowhere. If that's where the waste lives, then the efficiency lever isn't smaller models, it's fewer unnecessary iterations. Tighter pre-condition checks, better task decomposition so the agent doesn't wander, knowing when to stop before spending another 300 Wh on a dead end.
This really puts AI usage into perspective. Agent workflows are incredibly useful, but they also remind us that efficiency matters just as much as capability. As developers, we should build tools that solve real problems without wasting unnecessary compute.
At CodeCan.net, we're also thinking about practical development—helping developers ship faster with ready-to-use scripts and applications while encouraging smarter, more efficient workflows. The future of AI shouldn't just be more powerful; it should also be more responsible.
This is a conversation the AI industry needs more of. We often measure AI by intelligence or speed, but resource efficiency deserves equal attention. Smarter workflows, better model routing, and reducing unnecessary agent loops can improve both cost and sustainability. Thanks for bringing attention to the hidden side of AI.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.