AI is expanding the virtual world infinitely, but making real-world labor more expensive than ever.
Real-World Lessons from a Shattered Glass
Last month, the floor-to-ceiling window in my living room shattered unexpectedly. When I got a quote from the manufacturer, I was surprised:
The glass itself costs only 500 RMB, but the replacement fee is as high as 2,500 RMB.
The expensive part isn’t the glass—it’s the transportation, lifting, labor, and coordination, all the “non-glass” aspects.
For a 3.2-square-meter double-pane tempered glass, the material cost may be just 20% of the total fee. The remaining 80% is people, time, and the friction of the real world.
This reality reminded me of a recent a16z article, Why AC is cheap, but AC repair is a luxury . In China, we’re experiencing the same phenomenon: materials are getting cheaper, but labor is getting more expensive.
Jevons Paradox: Higher Efficiency, Greater Consumption
British economist William Stanley Jevons proposed the famous Jevons Paradox in 1865: when a technology becomes more efficient and cheaper, people actually consume more of that resource.
In China today, there are many real-world examples:
- The unit price of smartphones, solar panels, chips, and AI inference keeps dropping;
- Yet our consumption of smart devices, data centers, electricity, and computing power keeps rising.
Take artificial intelligence (AI, Artificial Intelligence) as an example: the cost of model inference is rapidly falling, but the number of calls is exploding exponentially.
The cheaper the computing power, the more it gets overused. This is the modern Jevons Paradox.
The essence of Jevons Paradox is not “saving,” but “expansion”: Efficiency improvements lower marginal costs, ultimately expanding total demand.
Baumol Effect: Lower Efficiency, Higher Cost
In contrast, the Baumol Effect (Baumol’s Cost Disease) reveals another, more subtle inflation mechanism.
When some industries experience a surge in productivity and wages, other less efficient sectors must also raise pay to retain workers.
For example:
- Tech and finance companies have high per capita output and high salaries;
- As a result, electricians, carpenters, and nannies must also see wage increases—because they compete for labor with programmers and AI engineers.
My glass repair case is a typical example:
Glass manufacturing is highly automated and prices are nearly transparent; but installation still relies on manual labor, lifting, and coordination—efficiency hasn’t changed, yet costs keep rising.
The Double Effect of AI
With the arrival of AI, Jevons Paradox and the Baumol Effect occur simultaneously. The table below summarizes their manifestations in key areas:
Here’s a comparison of Jevons and Baumol effects across different fields:
| Sector | Jevons Effect (Cheaper, More Usage) | Baumol Effect (Less Efficient, More Expensive) |
|---|---|---|
| Computing & Models | Inference costs drop, usage explodes | GPU electricity, data center maintenance costs rise |
| Content Production | Copywriting generation is nearly free | Human review and compliance costs increase |
| Manufacturing | Automation boosts output | Installation, transport, and after-sales labor costs rise |
| Education | AI teachers improve efficiency | Offline tutoring and private lessons get pricier |
So, we’re entering an interesting era: AI is driving the digital world toward zero marginal cost, but making the real world more expensive.
You can generate a 3D renovation plan in seconds, but hiring workers to install glass or wire electricity still takes days and thousands of RMB.
Reflexive Baumol Effect: The High Price of the “Last 1%”
After AI automates 99% of a process, the remaining 1% of work that must be done by humans becomes the new high-value bottleneck.
Examples include:
- Radiologists: AI can read scans, but only humans can sign off with legal responsibility;
- Autonomous driving: AI can drive, but human safety supervisors are still required;
- Software systems: AI can generate code, but architecture review and production approval remain human tasks.
This is the so-called “Reflexive Turbo-Baumol Effect”:
When AI automates almost everything, the last 1% of human labor becomes a scarce resource and regulatory bottleneck.
The Chinese Context: From “Cheap Labor” to “Expensive Labor”
For the past two decades, China’s economic growth was built on cheap labor and technological expansion. Now, AI is making intellectual labor cheaper, highlighting the scarcity and irreplaceability of physical labor.
You can have AI write a book or generate a report in minutes, but fixing a piece of glass, installing a window, or replacing a water heater still requires several people, hours of work, and hundreds of kilometers of logistics.
This is a structural reversal:
- AI enables unlimited expansion of “virtual work”;
- But “physical labor” is becoming a luxury.
Future inflation won’t be in factories, but in real-world services. This may be the new normal for Chinese society in the next decade.
The “Expensive Labor Industry” in the Age of AI Abundance
When we talk about the productivity revolution brought by AI, maybe we should ask:
Who will do the last 1%?
AI is driving the digital world toward zero cost, but exposing the true cost of “human collaboration” in the real world.
In the next decade, what’s truly scarce won’t be computing power, but human last-mile capabilities: those who understand machinery, can make house calls, work with their hands, and take responsibility.
Perhaps then, “the person who fixes glass” will be the true noble worker.
Summary
AI is reshaping the structure of productivity, driving the marginal cost of the digital world toward zero, but making labor and services in the real world more expensive than ever. The combined impact of Jevons Paradox and the Baumol Effect will profoundly influence China’s economy and social division of labor. We need to re-evaluate the value of “human labor,” especially those last-mile capabilities that cannot be automated.