rewrite this title DeepSeek Accelerated America’s AI “Manhattan Project”

Earlier this week I flew from my home in Florida to attend The Microcap Conference in Atlantic City, which I’m returning from today.

As I boarded the flight, I was thinking about Monday’s $1 trillion market meltdown and how the big AI companies weren’t the only ones who were hit hard by the news that China had developed a more efficient AI.

In a single day, energy companies lost over $40 billion in value as investors rushed to sell their shares of energy stocks.

Companies focused on nuclear energy were hit especially hard. Constellation Energy, the biggest U.S. producer of nuclear power, dropped 19% on Monday.

And I understand why.

When investors heard the news about China’s DeepSeek-R1, they worried that these energy companies would lose money because AI wouldn’t need as much power to run.

After all, what’s the point in building out a nuclear energy infrastructure in the U.S. if we don’t need all that power?

But as I sat in my seat watching wave after wave of passengers board the flight after me, it occurred to me that these investors might have made a mistake by selling so quickly.

I believe they might have overlooked something important: a principle called the Jevons Paradox.

My packed flight was proof that this paradox is still in play.

Here’s what I mean…

The Jevons Paradox

This idea of the Jevons Paradox comes from the British economist William Stanley Jevons back in 1865.

It suggests that when something becomes more efficient and uses less resources, people often end up using more of it, not less.

Jevons first noticed this pattern with steam engines and coal.

When more efficient steam engines were invented that used less coal, coal use didn’t go down.

Instead, it went up.

This happened because the more efficient engines were so useful that people started using them everywhere.

I remembered this idea as I sat on the tarmac on Tuesday waiting for my packed flight to take off.

Because the airline industry is a clear example of the Jevons Paradox happening today.

Per the IPCC, between 1960 and 2016, the per-seat fuel efficiency of jet airliners tripled or quadrupled, reducing the cost of flying by over 60%.

But despite these significant improvements in fuel efficiency, overall fuel consumption actually increased during that time due to the rapid growth in air travel demand.

Combined with population growth and rising incomes, the increased affordability of flying drove a 50-fold increase in global annual air travel…

From 0.14 trillion passenger-kilometers in 1960 to nearly 7 trillion by 2016.

This is similar to the paradox that Jevons observed back in 1865.

But instead of steam engines and coal, this time improvements in aviation efficiency have paradoxically led to greater overall resource consumption due to increased demand.

So here’s the good news if you’re still shellshocked from the events of this week…

The same thing could happen with AI.

Here’s My Take

Again, I understand why investors got out of AI and energy stocks on Monday.

When DeepSeek came out with a fast, efficient AI model that was apparently trained for only around $6 million, it upended everyone’s idea of what it takes to build and run an AI.

But dig a little deeper, and the story becomes clearer.

To scale an AI model, you train the model, then you use it to generate data. Then you train that model on the new data and use it to generate more data. And so on.

That’s how these Al models keep getting better and better.

But it seems that DeepSeek was able to “hack” this normal way of scaling by having a better model generate the data for them.

That way they were able to make a model comparable to OpenAI o1 at a fraction of the cost.

To be clear, I’m simplifying the training process. But that’s essentially what seems to have happened here.

And that’s why I believe a “Manhattan Project” for AI is more necessary now than ever.

We need to build an infrastructure in the U.S. that is capable of handling rapid growth in this sector.

Because the Jevons paradox tells us that with cheaper AI becoming available, we should see an increase in its use.

Financial experts at Morgan Stanley agree, saying that as AI becomes less expensive to operate, its use will likely increase dramatically.

And as more businesses and researchers start developing and using AI technology, it could actually lead to more energy use overall, not less.

That’s great news for energy companies… and their investors.

Best wishes,

Ian KingChief Strategist, Banyan Hill Publishing