01
Start here, this is the part the video skippedThe receipts: price fell 300x, demand didn't shrink — it ran away
The short version of the Jevons paradox is a story. Here's the data that turns it into a fact you can act on. From late 2022 to late 2024, the price to run a GPT-3.5-class model fell from about $20 per million tokens to roughly $0.07 — a ~280x drop in two years. Epoch AI, tracking the cheapest model that matches a given capability across six benchmarks, puts the median decline at ~50x per year, accelerating to ~200x per year after January 2024. If demand were fixed, total AI spend should have collapsed. It did the opposite.
- Price per unit of intelligence: down ~300x in roughly two years.
- Token volume over the same window: up thousands of percent (numbers below).
- Net result: total AI revenue and spend went up, not down. That gap — price down, total bill up — IS the Jevons paradox, measured.
This is the whole essay in one line: when price falls faster than your appetite is satisfied, the bill grows. AI is in exactly that regime right now.
02
Watch the volume side explode
The price drop is only half of it. The other half is how fast usage climbed to fill the new headroom. As the per-token cost cratered, the number of tokens people actually ran went vertical. These aren't projections — they're reported run-rates from the companies serving the traffic.
| Signal | Then | Now / recent | Source |
|---|---|---|---|
| GPT-3.5-class price | ~$20 / M tokens (Nov 2022) | ~$0.07 / M tokens (Oct 2024) | Epoch AI |
| Alphabet monthly tokens processed | (growth phase, 2024) | ~980 trillion / month (2025) | Alphabet / earnings |
| Microsoft tokens processed | — | 500+ trillion in 2025 alone | Microsoft / earnings |
| China's daily AI token use | ~100 billion / day (early 2024) | ~180 trillion / day (early 2026) | CEIBS reporting |
| Anthropic run-rate revenue | ~$1B ARR (Dec 2024) | ~$30B (Apr 2026) | Anthropic / VentureBeat |
China's number is a ~1,800x jump in about two years. Alphabet's and Microsoft's are quadrillions on an annual run-rate. The cheaper a token got, the more reasons people found to spend one.
03
Why this happens: the mechanism, not the metaphor
Jevons isn't magic and it isn't guaranteed. It fires when demand is elastic — when there's a large backlog of things people would do if only the price dropped. Cheap coal in 1865 met a backlog of factories, railways and ships that weren't worth building at the old price. Cheap tokens in 2026 meet a backlog of dashboards, internal tools, support bots, data cleanups and one-off scripts that were never worth a developer's week. When the price crosses the line where those become worth doing, they all get built at once. That's the surge.
- Efficiency cuts the cost per use (a better engine; a cheaper model).
- The lower cost crosses a threshold — uses that were 'not worth it' flip to 'worth it'.
- Because the backlog of those uses is huge, the new volume swamps the per-unit saving.
- Total consumption — and total spend — rises. The market widens instead of clearing.
The condition that breaks it: a saturated market where people already have all they want. That's not AI. The backlog of unbuilt software is the deepest it has ever been, which is precisely why the paradox is firing this hard.
04
This already played out once — in January 2025
You don't have to take the theory on faith; the market ran the experiment live. When DeepSeek shipped a frontier-ish model at a fraction of the training cost, US AI stocks dropped on the fear that cheaper AI meant a smaller pie. Within a day, Microsoft CEO Satya Nadella posted "Jevons paradox strikes again!" and argued the opposite: "As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of." He was reading the same curve you just saw. Cheaper input, bigger total market.
- The reflex (markets, Jan 2025): cheaper AI = less revenue, sell.
- The Jevons read (Nadella, same week): cheaper AI = use skyrockets = bigger pie.
- Eighteen months of token-volume data since then sided with the Jevons read.
05
Karpathy's tell: even the person who can build anything wants MORE
On June 9, 2026, Andrej Karpathy — now at Anthropic — named the same effect for software directly. Reacting to a major model release, he wrote that "working software increasingly comes out on a tap," so "the Jevon's paradox kicks in" and his own "demand for software [is] growing substantially." Sit with who's saying it. The man can write any tool himself, and his appetite for software is going up, not down. Cheap creation didn't satisfy his demand. It exposed how much he'd never bothered building because it used to cost too much. That backlog — yours, his, every business's — is the elastic demand the paradox needs.
- When code is cheap, the bottleneck moves from 'can we build it' to 'what's worth building'.
- Every bespoke single-use app that wasn't worth a dev-week is suddenly worth an afternoon.
- Real-world echo: per SaaStr, Claude Code went from its May 2025 public launch to a run-rate above ~$2.5B by early 2026 — that's the unbuilt backlog turning into spend.
Karpathy is quoted in short attributed phrases from his June 9, 2026 post. The framing is his; the reads in this guide are ours.
06
The catch the paradox refuses to hide
Read the curve honestly: total demand grows, but it does not get handed out evenly. When coal got cheap, the winners weren't people who admired steam engines — they were people who built mills, railways and ships on cheap power. Same shape now. Cheap tokens reward whoever turns them into something a specific buyer will pay for, not whoever merely has access. Everyone has access. Access stopped being the edge the day the price fell 300x.
- Saturated at the input layer (tokens are a commodity, racing to ~free).
- Wide open at the application layer (the backlog of 'now-worth-doing' jobs is enormous).
- The scarce thing is no longer the model. It's a clean offer pointed at one buyer's problem.
07
What to actually do with this (a worked read)
The paradox rewards motion, so here's a concrete way to use it this week instead of nodding and closing the tab. Pick one narrow job that just crossed the worth-doing line.
- Name a 'now-worth-doing' job. Find a task a specific business does by hand because a developer was always too expensive for it. Example: a dental clinic that re-types every new patient's intake form into their practice software.
- Price the old vs new cost. Old: a custom integration nobody would quote under a few thousand dollars. New: an AI workflow that reads the form and writes the record for cents per run. That price collapse is your opening — it's Jevons at the level of one invoice.
- Build the thin layer between the model and the buyer. Not the model — the workflow, the packaging, the 'it just works' setup and support. That layer is what the buyer pays for; the tokens underneath are nearly free.
- Ship the small version now. One clinic, one workflow, one monthly fee. 'I'll wait until it settles' is how you opt out of a market that is still widening — the volume tables above are the cost of waiting.
Notice the move: you're not betting on AI getting better (it will). You're harvesting the backlog that the price drop just made worth building. That backlog is what 'demand for software growing substantially' actually looks like on the ground.
08
The one-line version to remember
When something genuinely useful gets cheaper, people don't buy less of it — they find a hundred new reasons to buy more. AI is the cheapest it has ever been and the best it has ever been at the same time, and the receipts show total spend rising right through the price collapse. 'Saturated' and 'too late' describe a shrinking market. This one is measurably doing the opposite.
- Cheaper + better → more total demand and more total spend (because demand is elastic).
- The builder/reseller market is widening, not closing — the token-volume curves prove it.
- The risk was never being late. It's mistaking 'everyone has access' for 'everyone has an offer.'
Get the next drop
New AI build guides + the occasional bonus template. No spam, unsubscribe anytime.
By submitting you agree to our Privacy Policy & Terms. Unsubscribe anytime.
You're in — check your inbox to confirm.
Frequently asked questions
What's the actual evidence cheaper AI grew total demand, not shrank it?
Two curves that moved at once. Price: a GPT-3.5-class model fell from ~$20 to ~$0.07 per million tokens between Nov 2022 and Oct 2024 (Epoch AI tracks a median ~50x/year decline, ~200x/year after Jan 2024). Volume: over the same period, token usage ran vertical — Alphabet reported ~980 trillion tokens/month in 2025, Microsoft 500+ trillion in 2025, and China's daily usage went from ~100 billion (early 2024) to ~180 trillion (early 2026). Price down ~300x, volume up thousands of percent, total spend up. That gap is the Jevons paradox, measured.
Did Andrej Karpathy really say the Jevons paradox applies to AI?
Yes. In a public post on June 9, 2026 (after joining Anthropic), reacting to a major model release, he wrote that working software "increasingly comes out on a tap," so "the Jevon's paradox kicks in" and his own "demand for software [is] growing substantially." We quote only short attributed phrases; the economic reads and takeaways here are ours.
What is the Jevons paradox in one sentence?
When a resource gets more efficient (cheaper to use), total demand for it can rise rather than fall, because the lower cost makes many new uses worthwhile. William Stanley Jevons first described it for coal in 1865, after better steam engines led Britain to burn more coal, not less.
Didn't the DeepSeek shock prove cheaper AI hurts the industry?
The market's first reflex said so — US AI stocks fell in late January 2025 on the news. Within a day Microsoft's Satya Nadella posted 'Jevons paradox strikes again!' arguing cheaper, more accessible AI would make usage 'skyrocket.' The eighteen months of token-volume growth since then sided with him: cheaper input, bigger total market.
Doesn't cheaper AI just automate the work away?
Specific tasks get automated, yes — but the paradox is about the total. Cheaper creation unlocks a backlog of things never built because they cost too much. The ATM/bank-teller case is the classic shape: automation per branch rose, yet total teller jobs rose too, because cheaper branches meant banks opened far more of them. The mix of work changes; the total can grow.
Does the Jevons paradox always hold?
No. It needs elastic demand — a large pool of uses people would adopt if the price dropped. For a resource people already have all they want of, cheaper just means cheaper. AI sits firmly in the elastic camp: the backlog of software and automation nobody could justify building before is enormous, which is why the effect is firing so hard.
So is it actually too late to start building or reselling AI?
The data argues the opposite. A market whose core input is getting both cheaper and better is expanding. What's scarce isn't access to AI — that raced to a commodity. What's scarce is someone who packages it for a specific buyer. That gap is the opportunity, and the volume curves say it's still widening.
Sources · Andrej Karpathy on X — model reaction + Jevons paradox framing (June 9, 2026) · A quote from Andrej Karpathy — Simon Willison's Weblog (reproduces the Jevons quote) · LLM inference prices have fallen rapidly but unequally across tasks — Epoch AI (price decline figures, ~50x/yr, ~200x/yr post-2024) · AI Price Index: ~280x drop for GPT-3.5-class ($20 → $0.07 /M tokens, 2022–2024) — TokenCost · Why the AI world is suddenly obsessed with Jevons paradox — NPR Planet Money (DeepSeek, Nadella) · What is Jevons paradox? Satya Nadella on DeepSeek — Fortune (Nadella 'skyrocket' quote) · State of AI 2025 — token usage at scale (OpenRouter) · How China overtook the US in AI token usage — CEIBS (~100B/day → ~180T/day) · Anthropic Just Hit $14B in ARR, Up From $1B 14 Months Ago — SaaStr (Anthropic ARR ramp + Claude Code >$2.5B run-rate since May 2025 launch) · Anthropic says it hit a $30 billion revenue run rate after 'crazy' 80x growth — VentureBeat (Apr 2026) · Toil and Technology — James Bessen, IMF Finance & Development (Mar 2015), ATM / bank-teller data · What is Jevons Paradox? And why it may — or may not — predict AI's future — Northeastern Global News