Market on Fire and What to Look for in 2025

Happy New Year.

2024 is now in the rear-view mirror. And what a year it was.

For starters, we turned a sharp corner politically this year. Trump’s landslide victory and Republican control of Congress represent a sharp departure from Biden’s progressive, big-government agenda. That shift carries many implications, including the likely Death of Woke, with DEI and cancel culture fizzling out like an open bottle of Coke.

We also had a stock market on fire. We’ll close the year over 20% higher on the S&P 500 (likely above 23%). That follows 2023’s 24% return for a cumulative gain of over 50% in two years (the power of compounding).

Can we get another 20% plus year in?

It’s happened before. The S&P managed a four-year stretch during the rise of the Internet and the Dot-Com bubble. From 1995 through 1998, each year saw 20% or more gains. 1999 came in just shy of that target, with a 19.53% return.

All told, the S&P rose 219% over those five years. Though it’s far from certain, the AI frenzy could easily propel the market another 20% in 2025.

Unless, of course, AI hits its limits – of which there are many.

Power, for one, could bottleneck growth. An AI search requires 10 times the power of a regular search. And now that most browsers include an AI response in searches by default, energy consumption by data centers has gone through the roof. 

Through 2023, data center electricity demand grew at roughly 2.2% per year. A conservative estimate puts that growth at 10% to 15% for 2024, with an expected 44.7% growth through 2027.

There are some easy ways to meet some of that demand—oil and gas companies like Exxon plan to co-locate natural gas-fired plants next to data centers. As the world’s largest natural gas producer, the U.S. is well-positioned to quickly bring this energy source online.

However, we simply cannot expand electricity production fast enough to satisfy all of the expected growth in the near term. Power projects and infrastructure take too long to implement.

And when demand outstrips supply, the market balances the two through higher prices.

Another limit, which I told you about a couple of weeks ago, is AI itself.

AI needs human-generated data to train models. However, AI-generated content already exceeds 57% of all internet content, and that flood of AI data makes it exceedingly expensive to train new models.

For instance, Open AI’s next model upgrade, codenamed GPT-5 Orion, costs $500 million to train, is months behind schedule, and is still not ready for release.

As 2024 fades in the rear-view mirror, so too do the easy gains in large language model upgrades. Barring a breakthrough in how engineers train models, AI in 2025 looks like a grind of diminishing returns for incremental gains.

I’ll have more to say about what we can expect in 2025 on Thursday. Until then, enjoy the New Year while reflecting on an incredible past year and looking forward to what could be an even better year ahead.

Think Free. Be Free.

Don Yocham, CFA

Managing Editor of The Capital List