How to invest in AI in 2026: ETFs, stocks, and the bubble question

Investors looking at the AI boom in 2026 face an unusually difficult set of questions at the same time: how to get exposure, whether the leaders are already overvalued, how to think about ETFs versus single stocks, and how to size the allocation without taking concentration risk that can hurt them in a drawdown. This guide works through each of those decisions in the order they actually come up — paths to exposure, how to evaluate the vehicles, the honest both-sides bubble question, and a practical framework for sizing and managing the position over time.

The three primary paths to AI exposure

There are three real ways to invest in AI, and most investors end up using two of them rather than choosing one.

1. Individual AI stocks. The most direct path, with the highest single-name risk. The roster splits into chipmakers (semiconductor designers and equipment makers), hyperscalers (the cloud platforms running the largest AI workloads), and application-layer software companies embedding AI features into existing products.
2. AI-themed ETFs. A basket approach that bundles 30 to 100 names under a single ticker, weighted by some methodology. Better diversification than single stocks, but methodology matters enormously — two ETFs with similar names can hold very different companies.
3. Broad tech or index exposure. Owning a Nasdaq-100 fund or an S&P 500 fund already gives meaningful AI exposure, because the largest AI beneficiaries (Nvidia, Microsoft, Apple, Alphabet, Amazon, Meta) are also the largest index holdings. The exposure is indirect but not trivial.

Most diversified investors combine path 3 (baseline broad-market exposure) with one of paths 1 or 2 (thematic tilt). Building a portfolio out of single stocks alone is rarely optimal unless you have unusually deep conviction and the time to monitor positions.

What "AI exposure" already lives in a typical portfolio

Before adding more AI, audit what you already own. As of early 2026, the so-called Magnificent Seven — Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia, and Tesla — represent roughly a third of S&P 500 market capitalization, per S&P Dow Jones Indices public index data. Five of those seven are direct AI beneficiaries. A 60% U.S. equity allocation in a typical balanced portfolio already carries substantial AI exposure through that concentration.

This matters because investors who add a 10% AI ETF allocation on top of a tech-heavy core often end up with 25-35% effective AI exposure once you look through to the underlying holdings. That can be the right answer — but it should be a deliberate choice, not an accident.

How to evaluate an AI ETF (the four-dimension framework)

When comparing AI ETFs, four quantifiable dimensions matter more than the marketing copy.

• Concentration. What percentage of the fund sits in the top 10 holdings? Funds with top-10 weight above 60% behave more like concentrated single-stock baskets than diversified theme funds.
• Expense ratio. Thematic ETFs typically charge between 0.35% and 0.75% — a meaningful drag at the high end over a 10-year horizon. Always compare against a broad-market alternative.
• Methodology. Is the fund passively tracking an index, or is it actively selected by the manager? Passive funds offer transparency and lower fees; active funds can adapt but introduce manager risk.
• Methodology drift. Does what the fund holds today actually match its name? An "AI ETF" launched four years ago may now hold companies whose AI relevance is questionable. Read the most recent prospectus and holdings file.

The SEC's Investor.gov ETF guide covers the basics of how to read a prospectus. Our deeper AI ETF evaluation framework walks through the comparison side by side.

Picks-and-shovels vs. application layer

A central debate in AI investing is whether to favor the infrastructure ("picks and shovels") or the applications.

Historically, in early-cycle technology buildouts, the infrastructure providers capture more value than the application layer because the winning applications haven't been identified yet. The classic example: during the railroad boom, rail-equipment makers and steel producers compounded reliably while individual railroad operators went bankrupt at high rates. During the early internet, the networking-equipment vendors and chipmakers captured durable value before the application-layer winners (Google, Amazon, Facebook) emerged years later.

If that pattern repeats, AI infrastructure exposure — semiconductors, datacenter REITs, utilities, networking — captures more of the first wave. Application-layer exposure becomes more attractive once monetization patterns mature and the platform shakeout settles. Many AI ETFs blend both; some specialize in one or the other.

This is one reason a portfolio approach matters more than a single bet: the truth is you don't know yet which layer will dominate, and the answer may be "both, at different stages."

The honest bubble question

There is no way to write seriously about AI investing in 2026 without addressing whether it's a bubble. The honest answer is that some bubble indicators are clearly present, and others are clearly absent — both things can be true at the same time.

Bubble indicators present:

• Index concentration in mega-cap AI beneficiaries is the highest in modern market history, per data tracked by Bank of America Global Research and similar institutional sources.
• Circular capex — hyperscalers buying chips from Nvidia, Nvidia investing in cloud and model providers, those providers committing back to hyperscaler infrastructure — creates revenue interdependencies that can unwind quickly.
• Forward valuations for AI leaders price in years of continued execution at current growth rates.

Bubble indicators absent (vs. the late-1990s comparison):

• AI leaders generate substantial cash earnings today, unlike many 1999 internet pure-plays.
• Infrastructure capex creates physical, depreciable assets (chips, datacenters, power) that retain value even if specific use cases fail.
• Enterprise adoption is producing measurable revenue inside Microsoft, Google, Amazon, Meta — visible in 10-Q filings, not just press releases.

Our Is AI a bubble? page works through both sides in detail, including how to invest such that you're protected if either side is right.

How to size the position

Sizing AI exposure is the decision most investors get wrong, in either direction. Two common errors:

1. Under-sizing. Avoiding AI entirely because of bubble fears, while owning broad-market index funds that already carry heavy AI weight by market cap. The investor thinks they have no exposure; they actually have a lot.
2. Over-sizing. Adding a thematic AI ETF on top of an already tech-heavy portfolio without looking through to combined exposure, ending up with 30%+ in essentially seven names.

A practical sizing framework:

• Audit existing exposure first. Look through your equity holdings and tally what's already AI-adjacent. Many investors are surprised.
• Decide on a target. Most diversification frameworks suggest thematic tilts under 10-20% of equities, depending on conviction and time horizon.
• Pick vehicles. Use broad index funds for baseline exposure, an ETF for diversified thematic tilt, and individual stocks only for high-conviction theses you can defend in writing.
• Plan for drawdowns. AI-heavy portfolios should be expected to have higher volatility than the broad market. If a 30% drawdown would force you to sell, the position is too large.

This kind of sizing question is exactly what tools like Kronos are built for — running your existing portfolio through scenario analysis to see what AI concentration you already carry before deciding whether to add more.

Long-term AI investing: themes, not names

Long-term AI investors generally do better when they think in terms of durable themes rather than specific company names.

Names rotate. The single dominant chipmaker of 2026 may not be the single dominant chipmaker of 2030. Foundation-model platform leaders may shuffle. Application-layer winners are still being sorted out.

Themes compound more reliably. The need for compute, energy, networking, data infrastructure, and software that integrates AI is not seriously in doubt — even bears on specific names tend to grant the underlying secular demand. Investors who stay diversified across the themes, rebalance as concentration grows, and resist the temptation to chase last quarter's winner have historically outperformed those who concentrated and held single names through the full cycle.

The history of past technology cycles is instructive: the railroads, electrification, automobiles, semiconductors, the internet — each produced durable wealth for diversified, patient investors and large losses for those who concentrated at peaks. AI is unlikely to be different in that respect.

Risks investors underweight

Most investors think about valuation risk and forget the others. The full risk list:

• Concentration risk. A handful of mega-caps drive most thematic and broad-index returns. A regulatory action against one or two of them would mark down a much larger basket.
• Valuation risk. Forward multiples already price in significant continued growth. Multiple compression alone can cause 30%+ drawdowns even with strong fundamentals.
• Regulatory risk. Export controls (especially U.S.-China chip restrictions), antitrust action against hyperscalers, and AI-specific safety rules are all moving targets in 2026.
• Energy and grid risk. Datacenter power demand is colliding with grid capacity in several regions. The International Energy Agency's 2024 report on data center electricity use projected datacenter electricity demand could double by 2026. If that demand can't be met, AI growth slows mechanically.
• Methodology risk in thematic ETFs. What you think you own may not be what the fund holds in 12 months. Re-read the holdings annually.
• Behavioral risk. Most underperformance in volatile themes comes from investors selling at lows and buying at highs. The position size needs to be small enough that you can hold through a drawdown.

Where to go from here

Three follow-on resources, depending on what you need to decide next:

• If you're comparing specific AI ETFs: see the AI ETF evaluation framework.
• If the bubble question is what's stopping you: see Is AI a bubble?.
• If you're earlier in the journey: AI investing for beginners walks through the basic terms and a starter framework.
• If you've narrowed it to ARK Innovation versus a pure-play AI ETF: AI ETF vs. ARK Innovation breaks down the difference factually.

A coherent AI allocation isn't about picking the right stock — it's about understanding what exposure you already carry, deciding deliberately how much more you want, and building the position in a way that survives drawdowns. The investors who will own AI well in 2030 are the ones who size correctly, rebalance, and don't let the cycle's emotional swings force them to sell low.

---

This page is educational and does not constitute personalized investment advice. Consult a qualified advisor before making investment decisions. Clockwise Capital is a registered investment adviser; Clockwise and its principals may hold positions in securities or sectors discussed.