The Jevons Paradox: Why AI Is Making ‘Dying’ Industries Bigger — And What Actually Kills Them
- Making something efficient increases demand (Jevons Paradox)
- The thing that kills an industry is always a different technology that changes behavior entirely
- 5 industries are in the Jevons phase right now (growing, not dying)
- 3 industries just crossed from Jevons to Substitution (March 2026)
- First-order thinking — "AI replaces X, so X dies" — is wrong almost every time
The Pattern Nobody Sees Coming
In 1975, Xerox had a problem. The personal computer was supposed to make the paperless office a reality, with cheap printers and digital documents expected to kill paper entirely.
Except paper consumption doubled over the next 25 years.
Cheap printing made it so easy and fast that people printed everything: drafts, emails, web pages, even directions to the restaurant. They printed things they never would have bothered with when it cost money and took effort, and the efficiency gain ended up blowing the doors off demand.
This is Jevons Paradox, named after the economist who noticed in 1865 that making coal engines more efficient actually increased coal consumption. More efficient engines meant cheaper power, cheaper power found more uses, and industry ended up burning more coal than ever.
For investors, the key lesson is that office paper was finally killed by a completely different technology, cloud storage and smartphones, that changed behavior entirely. People stopped printing because they stopped needing paper. Different tech drove different behavior.
Most AI analysis falls into the same first-order trap: “AI automates X, therefore X dies.” The logic is intuitive, but it fails almost every time. Efficiency creates more demand, while the real disruption comes from somewhere else entirely.
Five Times This Already Happened
This isn't a theory. It's a pattern with a perfect track record.
Everyone predicted the paperless office, but consumption exploded because printing got cheap enough that people printed everything. It took 30 years and a completely different technology (cloud storage + smartphones) to actually kill office paper.
ATMs were supposed to eliminate tellers. Instead, cheaper branches led to more branches and more tellers; the total headcount grew. Mobile banking, an entirely separate technology, was what finally cut their numbers.
Home video was supposed to kill cinemas. Box office revenue surged for 20 years because VCRs made people more interested in movies, not less. Streaming plus 65-inch 4K TVs eventually dealt the real blow.
Email was supposed to kill mail carriers. Instead, e-commerce, driven by that same digital revolution, created the biggest package delivery boom in postal history. Amazon logistics is the “second technology” actually reshaping the system.
GPS navigation made drivers more efficient, which created more available rides, lower prices, and higher volume. App-based ride-hailing from Uber and Lyft, an unrelated behavioral shift, is what actually disrupted the taxi industry.
It’s the same pattern, five times in a row: the efficiency tool grows the industry, while the kill shot comes from a completely different direction.
Where It’s Happening Right Now
Our engine tracks 28 industries, using 8 analytical dimensions across 5 time horizons. Five of those industries are deep in the Jevons phase right now: AI is making them more efficient, and demand is exploding. For each one, we’ve also identified the “second technology,” the behavioral shift that could eventually trigger real substitution.
-
Cloud Platforms
AI makes compute more efficient, but demand for inference compute is exploding faster than efficiency gains can offset. Every company needs AI infrastructure. Every product wants an AI feature. The efficiency is real — and so is the exponential growth in workloads.
Watch for: Edge AI and on-device processing — when inference moves off the cloud entirely.
-
Energy
AI optimizes grids beautifully. It also requires data centers that are hitting physical infrastructure limits. Bloom Energy has a $20B backlog. Utilities can't build capacity fast enough. The "AI makes energy efficient" narrative is true — and completely irrelevant to the demand story.
Watch for: Off-grid fuel cells and small modular reactors (SMRs) — when data centers stop needing the grid.
-
Healthcare
AI diagnostics make care cheaper and faster. Cheaper and faster care means more people getting screened, more conditions caught early, more procedures performed. Radiology AI isn't replacing radiologists — it's creating a backlog of newly-detected conditions that need treatment.
Watch for: Preventive genomics — when personalized prevention makes diagnosis itself less necessary.
-
Consumer Services
AI chatbots and automation cut service costs, which lowers prices, which increases foot traffic and volume. Restaurants using AI ordering aren't shrinking — they're serving more customers per hour with the same staff.
Watch for: Robotic direct-to-consumer delivery — when the store itself becomes unnecessary.
-
Food & Beverage
AI optimizes supply chains, cuts waste by 30-40%, and lowers costs. Lower food prices increase consumption. Optimized distribution opens new markets. The industry grows.
Watch for: Precision fermentation — when lab-produced proteins replace traditional agriculture entirely.
If you're short these industries because "AI is automating them," you're making the same bet people made against paper in 1985. The efficiency gains are real, but the conclusion gets the dynamic backwards.
The Three That Just Crossed the Line
Jevons doesn't last forever. Eventually, a second technology arrives, habits shift, and the market flips from growth driven by efficiency to contraction driven by substitution.
In our March 2026 analysis, three industries just crossed that threshold.
For two years, AI assistants made SaaS users more productive, creating demand for more features and more users at the same per-seat price. Agentic AI crosses that line by replacing per-seat licensing models entirely. When an AI agent handles the work of 5 Salesforce seats, you don't need 5 Salesforce seats. The shift from "AI helps me use software" to "AI is the software" is the second technology, and substitution has begun.
AI driver-assist features were classic Jevons: they made driving easier and safer, which led to people logging more miles. But Tesla's Cybercab enters mass production with no steering wheel and no human operator, substituting the driver entirely. Once the vehicle doesn't need a human, car ownership changes at a structural level.
For years, AI-powered analytics made junior analysts more productive, prompting banks to hire more analysts to process more deals. Now, agentic trading systems and autonomous wealth management platforms are replacing those junior roles entirely. Goldman's AI trading desk no longer needs the humans it used to augment.
These industries won't collapse tomorrow, but their Jevons tailwind is ending. Over the next 5-10 years, companies that fail to recognize the shift will be restructured by the market.
The Investor’s Framework
Here's how to actually use this:
- Is AI making this industry more efficient while demand keeps rising? That's the Jevons phase. Bullish for 1-3 years. The industry is growing because of AI, not despite it.
- Has a "second technology" arrived that changes behavior entirely? Watch for the shift from "AI helps humans do X" to "AI replaces the need for X." That's substitution. The 5-10 year trajectory just changed.
- First-order thinking is the trap. "AI automates X, therefore X dies" sounds logical and is almost always wrong in the short-to-medium term. The Jevons phase can last 10-20 years. The money is in knowing which phase an industry is in.
- The kill shot comes from a different direction. ATMs didn't kill tellers — mobile banking did. VCRs didn't kill theaters — streaming did. Watch for the behavioral shift, not the efficiency tool.
Our engine tracks this for all 28 industries, updated weekly. We classify each industry's AI trajectory as Jevons (efficiency growing demand), Transition (second technology emerging), or Substitution (behavioral shift underway). The three reclassifications above are the first moves from Jevons to Substitution in our March 2026 update.
Track All 28 Industries
Across 8 dimensions, 167 cross-industry effects, and 5 time horizons, know which phase every industry is in before the market figures it out.
Related reports: The Pharma Split | 600 Years of Technology Panic | The Human Bottleneck