AI Isn't Just Hype: Here's How to Actually Invest in It

Here's the thing about AI investing in 2026: the obvious investments are already crowded.

Nvidia is up over 600% in 3 years. Microsoft is part of every institutional portfolio. Google and Amazon have been “AI plays” for almost a decade.

If you’re holding these companies (which I am), I’m not saying they won’t be good investments, because they have proven they are.

What I’m saying is that if you’re only looking as far as the “mega-caps” you’ve made an index bet tilting towards the tech sector.

There’s some more opportunity hiding just a single layer deeper than where you’re looking right now.

I want you to think of AI the way you’d think about the California Gold Rush (1848-1855). The miners got all of the headlines. But it was the people who sold the picks, shovels and denim jeans that built the most wealth.

As it relates to AI, right now the “picks and shovels” of the AI economy are power grids, semiconductor supply chains, fiber optic infrastructure and the cooling systems keeping data centres from melting.

AI doesn’t stop at the chatbot interface, likely how you (and I) have been using it. It runs all the way down to the copper in the ground.

This is the one most retail investors completely overlook. Training just one large language model (LLM) consumes just as much electricity as hundreds of homes in a year. Now take that and multiply it across thousands of data centres across the globe, and you now have a structural energy demand problem that no tech company could ever solve on its own.

This, in its simplest form, is a tailwind for utilities, natural gas producers and nuclear energy companies.

Think Vistra $VST ( ▲ 0.27% ), Constellation Energy $CEG ( ▲ 0.86% ) and Entergy $ETR ( ▼ 0.29% ), which are all benefitting from this trend.

Nvidia gets all the glory. But the semiconductor supply chain extends far beyond it.

Taiwan Semiconductor $TSM ( ▲ 0.8% ) is the one fabricating the chips that Nvidia designs and this will continue for the foreseeable future.

Broadcom $AVGO ( ▼ 0.04% ) provides custom AI accelerators to Google, Meta and Apple at large scales.

These companies are less “flashy” but they are offering something Nvidia can’t right now, and that’s a more defensible moat and a more rational valuation.

Amazon Web Services, Microsoft Azure, and Google Cloud are the rails that AI runs on.

But the physical infrastructure that are building those rails (data centres, server equipment and cooling systems) are where the silent opportunity is.

Let's be real here: AI is here to stay, but many AI-linked stocks are expensive.

Here’s the number that will stop you in your tracks.

The S&P 500’s forward earnings yield, what you earn in return from corporate profits relative to stock prices, is currently near the same level as the 10-year U.S. Treasury yield.

S&P 500’s forward earnings yield: 4.46%

10 Year Treasury Yield: 4.34%

The equity risk premium, in simpler terms the extra return investors demand for taking on stock market risk over “safe” government bonds” has shrunk to 0.12% as I right this.

For AI stocks specifically, the premium is even more important.

Many stocks trade at 50–100x forward earnings, pricing in growth that must compound perfectly for a decade to justify current valuations.

History shows that even great businesses can be terrible investments if you overpay for them.

The smart move here isn't to avoid AI.

It's to evaluate whether a theme is worth the price tag.

And that starts with asking 3 questions:

1) What growth rate is already priced in? Work backwards from the current P/E or EV/Sales multiple. If the stock requires 40% annual revenue growth for 7 years to justify the price, ask yourself how confident you are in that happening.

2) Where is the moat? AI software can be copied. AI infrastructure is much harder to replicate. Focus on companies with moats over pure software plays.

3) Who are the real customers paying real money today? Hype gets headlines. Revenue gets valuations justified. Focus on companies with enterprise contracts, growing backlog, and improving free cash flow margins. Stay away from companies that are just presenting exciting demos.

Let's say you've done the research, you believe in the AI infrastructure thesis, and you've found two or three companies you genuinely have conviction in.

How much should you allocate?

This is where most investors get it wrong.

They either go too small to matter, or they concentrate so heavily that one bad earnings call wipes out years of gains.

Here's a framework that works:

1) Set a ceiling of 15–20% of your portfolio. All AI-related positions combined, whether ETFs, infrastructure plays, or direct tech companies, shouldn't exceed this. This gives you exposure without too much risk.

2) Layer in: 50% via ETF, 50% via individual picks. ETFs give you diversified upside with lower single-stock risk. Individual picks let you overweight the highest-conviction opportunities. Don't pick one or the other, use both.

3) Buy in intervals, not all at once. High-valuation stocks are volatile, we all know that. Splitting your entry into 3 purchases over 6–9 months smooths out your average cost basis and reduces the risk of bad timing.

4) Set a pre-defined exit trigger, not just a price target. Instead of "I'll sell at $300," decide "I'll sell if the fundamental thesis breaks." This keeps your emotions out of the equation.

Here's the uncomfortable truth about transformative technology cycles:

They almost always take longer to play out than the bulls expect, and they almost always go further than the bears believe.

The internet in 1999 was real technology with real long-term potential. But the valuations were also insane.

Investors who bought the best businesses at reasonable prices in 2002 made incredible returns.

The ones who paid peak-bubble multiples in 1999 waited a decade just to break even.

AI is not a bubble in the sense that it's fake.

But concentration risk is real, valuation risk is real, and the path from "transformative technology" to "transformative investment returns" is almost never a straight line.

The investors who survive this market cycle will be the ones who ride it intelligently, not recklessly.

Alex (The Dividend Dominator)
Founder and CEO of Dividend Domination Inc.
Follow me on Twitter and Instagram