The AI Boom Is About to Repeat a 160-Year-Old Paradox

England was terrified it would run out of coal...

But in 1865, steam engines were becoming more efficient. Everyone assumed that would solve the problem... Surely the country would use less coal as the technology advanced.

Then, economist William Stanley Jevons stumbled onto something that stunned the British scientific community...

As steam engines became more efficient, coal consumption didn't decrease. It increased.

Factories built more steam engines. Railways expanded. Economic activity exploded.

Jevons realized he wasn't witnessing an energy shortage... but a productivity shock. And it was so powerful, it led to far more consumption of the very thing the country wanted to conserve.

This disconnect became known as the Jevons paradox. And for 160 years, economists have watched this pattern repeat across industries...

• More fuel-efficient cars led people to drive more miles.
• Cheaper lighting meant buildings stay lit all night.
• Faster Internet caused data usage to explode.

And now, the Jevons paradox is about to hit artificial intelligence ("AI") with full force...

The Questions Jevons Would Ask

Most investors are asking, "Which AI model will win?" "What jobs will even exist?"

But those are the wrong questions. Jevons would tell you to ask something different... "What will AI cause us to consume far more of?" "What will it create more of?"

Again, it's just like the steam engine...

Benedict Evans, a longtime Wall Street and venture-capital analyst, recently published a chart about that productivity shock...

His work shows that over about 60 years, the steam engine boosted productivity so much, it gave Great Britain the equivalent of a labor force roughly 5 times its total population. Take a look...

Importantly, the efficiency gains didn't replace workers. They produced the equivalent of 300 million new ones. The higher efficiency led to an economic explosion.

AI is following the same script.

Some folks like to say that AI does the work of hundreds of millions of interns. But we can look at this another way... And if we do, we see that AI is creating a new source of demand – comparable to a massive labor force.

As AI becomes more efficient, it will trigger a historic surge in demand for resources that most investors aren't even thinking about. And it points toward a specific corner of the market...

The Math Is Already Playing Out

Every breakthrough that makes AI cheaper to run only increases demand. It's already showing up in corporate energy consumption...

Goldman Sachs Research projects that data-center power demand will surge 175% by 2030 compared with 2023. That's the equivalent of another top 10 power-consuming country.

In September 2024, Microsoft (MSFT) signed a 20-year deal with Constellation Energy (CEG) to restart the Three Mile Island nuclear reactor for the first time since 2019. Constellation is investing $1.6 billion to bring it back online by 2027.

Think about that. Microsoft – one of the most sophisticated technology companies on Earth – is so desperate for reliable power that it's restarting a nuclear reactor.

As I've covered before, the constraint with AI isn't software... It's physical infrastructure.

This also means jobs and spending are not being drained by AI. They're moving into infrastructure instead. Goldman Sachs estimates that it will cost $720 billion in grid spending through 2030 to meet this demand.

And this is only the first wave of this trend. "Edge AI" will also increase consumer demand for AI services...

Apple, Alphabet, Samsung Electronics, and Meta Platforms are racing to put AI directly into devices like your phone, glasses, earbuds, and even your car. That's edge AI. It processes data closer to your devices instead of remotely.

This sounds like it would reduce data-center demand. After all, if AI runs locally, you don't need the cloud, right?

Wrong. With AI embedded in billions of devices, people will use AI constantly. Every interaction generates more data. Every query sparks follow-up queries. Local AI can handle simple tasks – but complex requests still ping the cloud.

We're not looking at a shift from cloud to device... We're looking at an explosion in both.

Consulting firm McKinsey estimates that U.S. data-center electricity demand alone could more than quadruple from 2023 levels by 2030. And that's before the edge-AI wave fully hits.

Where the Smart Money Is Looking

In 1865, the "obvious" bet was on steam engine manufacturers. But the smart bet was on the infrastructure every steam engine needed – coal, steel, and railways.

Today, the obvious bet is on AI developers... But the smart bet is on the infrastructure they need.

Jevons understood 160 years ago that efficiency doesn't reduce resource consumption. It leads to adoption. And adoption requires infrastructure.

And as AI moves to the edge, the winners will shift to picks-and-shovels companies like Micron Technology (MU) and Qualcomm (QCOM) as AI reaches its "solar panel moment."

These aren't flashy AI plays. But Jevons would recognize them immediately.

They're coal in 1865. Oil in 1901. Fiber-optic cable in 1995.

So yes, AI will continue becoming more efficient... And that means demand for physical infrastructure will continue to surge.

Good investing,

Josh Baylin

Editor's note: After marking dozens of new highs again this year, the headlines are calling the AI boom a bubble. But those warnings are dead wrong. According to one market veteran with a 15-year track record, a historic pattern is taking shape that could send $7.4 trillion pouring into select stocks... triggering a Melt Up that could run for years to come.