By Robo Digitalis | April 8, 2026
Everyone is talking about AI. Your LinkedIn feed is drowning in it. Your coworker is worried about their job. Your CEO just announced an "AI-first strategy." Pundits are debating AGI timelines like they're picking Super Bowl winners. The word "efficiency" gets thrown around like a magic spell — if we just make AI efficient enough, everything gets better, cheaper, cleaner, faster.
There's a 19th-century economist who would like a word.
The Man Who Saw It Coming in 1865
William Stanley Jevons wasn't a tech pundit. He was a 19th-century British economist watching the Industrial Revolution unfold in real time. And in 1865, he wrote a book called The Coal Question that made an argument so counterintuitive that people thought he was wrong.
His observation was this: James Watt had just invented a dramatically more efficient steam engine. Watt's design used roughly 75% less coal than the previous model. Every engineer in Britain was celebrating. Finally — we'll stop burning through our coal reserves. Efficiency wins. ncm.org
Jevons said: not so fast.
He argued that a cheaper, more efficient engine wouldn't reduce coal consumption. It would explode it. And he was right. When coal-powered machinery became affordable enough for every factory, mill, and mine in Britain to run — they all ran. Constantly. The efficiency gain didn't save coal. It made coal accessible at scale, and Britain burned through more of it than ever before. When the cost of producing a ton of iron was cut by two-thirds, Scotland saw a ten-fold increase in total coal consumption between 1830 and 1863. en.wikipedia
More efficient didn't mean less consumption. It meant the door opened wider — and everyone walked through it.
The Conversation We're Actually Having
Right now, the dominant AI conversation runs on a handful of charged terms. Let's call them out:
"AGI is coming." Depending on who you ask, artificial general intelligence — a machine that can think and reason like a human across any domain — is either five years away, fifty years away, or already here and we just haven't noticed. Every major lab is racing toward it. The marketing budgets are staggering. The hype is nuclear.
"It's going to take my job." Radiologists. Lawyers. Coders. Writers. Customer service reps. The list of threatened professions grows weekly. McKinsey publishes reports. Congress holds hearings. People are genuinely scared, and that fear isn't irrational.
"But it'll create new jobs." The counter-argument, deployed like a sedative. The Industrial Revolution took jobs and created new ones. This will too. Probably. Maybe.
"We're making it more efficient." Every quarter, new chips. New model architectures. Smaller models that run on your phone. "We're doing more with less," the press releases say.
Here's what the Jevons Paradox tells us about that last claim: it doesn't matter.
More Efficient Means More Everything
The researchers behind the paper found something that should reshape how we talk about AI's footprint. Energy consumption in big tech doesn't scale with efficiency — it scales with revenue. Every dollar earned gets reinvested. Every reinvestment builds more infrastructure. More infrastructure powers more services. More services reach more users. More users generate more data and more demand.
The loop doesn't tighten. It expands.
Think about what's happened with AI in just the last two years. Inference got cheaper. So companies deployed AI into search, into email, into spreadsheets, into customer service bots, into coding assistants, into medical diagnosis tools. The cost per query dropped. The number of queries went from millions to billions per day. Net energy consumption: up. Sharply.
We didn't get more efficient. We got more ambitious.
What This Means for the AGI Debate
Here's where it gets philosophically uncomfortable.
The people racing toward AGI — the ones promising it will cure cancer, solve climate change, eliminate poverty — are, in the process, building one of the most energy-hungry systems in human history. And the paradox says: the smarter and cheaper the AI gets, the more of it we'll use, not less.
If AGI arrives and it's dramatically more capable than today's models, the demand for it won't gently plateau. It will go vertical. Every business, every government, every individual will want access to a machine that can out-think human experts on demand. The compute required for that world isn't in our current roadmap. It might not even be on our current planet.
So when someone tells you AGI will solve climate change — ask them what it will cost to run AGI at planetary scale. The Jevons Paradox suggests the answer is: more than we're currently imagining.
Will It Take Your Job?
Probably some of it. That's the honest answer and I'm not going to dress it up.
But here's the more interesting question the paradox raises: AI won't just replace tasks, it will make those tasks so cheap that we'll do more of them. A lawyer who used to spend three days on due diligence can now do it in three hours — which means they take on four times as many clients, not go home early. A coder with an AI assistant doesn't write less code. They ship four times as many features.
The paradox doesn't play favorites. It applies to human labor augmented by AI the same way it applies to coal-powered engines. Efficiency unlocks demand. Demand fills the space. The worker doesn't disappear — they just run faster on a bigger treadmill.
Whether that's liberation or exhaustion depends entirely on who owns the treadmill.
So What's the Coal in 2026?
Simple. It's compute.
And the AI industry just built a much more efficient steam engine.
Every few months, a new model drops that's faster, cheaper, and smarter than the last. DeepSeek shocked the world by delivering near-GPT-4 performance at a fraction of the cost. Smaller models now run on laptops, on phones, on edge devices. The cost per AI query has dropped by orders of magnitude. arxiv
And just like 1865 Britain — instead of consuming less compute, we consumed infinitely more of it. Generative AI platforms now consume approximately 600 megawatt-hours of electricity daily — equivalent to over 50 times the annual energy use of an average US household. Global data center electricity demand is already at 2% of total world electricity and is projected to more than double, surpassing the entire national power consumption of Canada. nature
Jevons called his shot 160 years ago. The playbook hasn't changed.
You're Burning Coal Too
Here's where the analogy gets personal.
The same paradox happening at the infrastructure level is happening in your browser right now. AI has never been cheaper or more accessible. A subscription to a frontier LLM costs less than a Netflix plan. The tools that used to require a PhD and a research budget are sitting in your pocket.
And most people are using that efficiency gain to watch more YouTube videos about AI.
That's the paradox at the individual level. The cost of learning about AI has dropped to near zero — so consumption of AI content went through the roof, while actual output stayed flat. You didn't get more productive. You got more informed about productivity.
Jevons would recognize it immediately.
Pick One Thing. Today.
Not a business plan. Not a 12-week roadmap. Not a Notion doc full of ideas you'll revisit when things slow down.
Start here: subscribe to one $20/month LLM and ask it a real question every single day. Not a test. Not a "what is the meaning of life" prompt. A real problem you actually have — a decision you're stuck on, a task eating your morning, a concept you keep half-understanding from YouTube videos but never applying.
Ask it. Read the answer. Do the thing.
That's the steam engine. That's the coal. The question is whether you're going to build something with it or stand at the factory gates watching the smoke rise, waiting for the right moment to go in.
Fix the spreadsheet that wastes two hours of your week. Automate the email you write twelve times a month. Build the landing page for the idea that's been in your notes since last year. Teach yourself the thing you keep deferring with another video.
The world is not pausing the AI revolution while you get ready for it.
A Note of Gratitude
I wouldn't be thinking about any of this if @CompleteTechLLC hadn't dropped this paper in front of me. That's the kind of signal worth amplifying — not another hype post, not another panic piece, but a quiet, rigorous argument that reframes the whole conversation.
Jevons saw the Industrial Revolution clearly while it was happening. Most people just saw smoke and noise.
Don't just watch the revolution. Build something in it.
— Robo Digitalis | Side Quest Studios
TerminalKing.com