AI Concentration Risk: Build a Diversified Portfolio

For the better part of three years, the easiest way to look like a genius investor was to own the biggest names in artificial intelligence. Then came a stretch where the opposite happened: a version of the S&P 500 with the AI-heavy names stripped out reportedly outperformed the standard index. That single data point is worth more than a headline. It is a reminder that "the market" is not a single thing, that leadership rotates, and that the portfolio which wins over the next decade is built on process rather than prediction.

This article is not about calling the top of the AI trade. Nobody can do that reliably, and anyone who claims otherwise is selling something. It is about the money rules that hold up whether the tech trade rips higher, stalls, or frays — how to think about concentration, how to size a theme, and how to build a portfolio that does not quietly become a single bet.

The uncomfortable truth about "market returns"

When people say the index is up or down, they usually mean a market-cap-weighted index. That construction gives the largest companies the loudest voice. If a handful of firms grow to represent a large share of total market value, their fortunes effectively become the index's fortunes.

This is not a flaw — it is how cap weighting works, and it has served long-term investors well because winners keep winning and the index automatically gives them more weight. The trade-off is that the index can become less diversified than its 500-stock label suggests.

What an "ex-AI" index really measures

A benchmark that excludes the companies most tied to the AI theme is essentially a counterfactual: what would the market have returned if that one dominant narrative had not existed? When the ex-theme version wins, it tells you two things at once. First, the excluded group underperformed. Second, the rest of the market was doing more heavy lifting than most investors assumed.

Neither observation predicts the future. But both are useful diagnostics for your own portfolio, because most investors unknowingly own an even more concentrated version of the index than the index itself.

The hidden overlap problem

Consider a typical growth-oriented portfolio assembled over the past few years. An index fund, a tech-focused fund, a semiconductor ETF, a "quality growth" fund, and a few favourite stocks may look like five positions. In practice they often share the same top holdings. The investor believes they hold a diversified basket; they actually hold one idea in five different wrappers.

• Two funds with different names can have substantially overlapping top ten holdings.
• Sector funds and thematic funds often double up on names already owned through a broad index.
• Employer stock, RSUs and options can silently compound the same exposure.

The fix is unglamorous: look through the wrappers, list the underlying companies, and estimate what share of your total investable assets sits in a single theme. Most people are surprised by the answer.

Why narrow leadership feels safer than it is

Concentration feels comfortable when it is working. A rising position produces a psychological reward that makes adding to it feel like discipline rather than risk-taking. Several well-documented behavioural biases reinforce this:

• Recency bias — recent returns get extrapolated into a permanent expectation.
• Narrative bias — a compelling story feels like evidence, even when the story is already reflected in the price.
• Herding — owning what everyone owns feels low-risk because the crowd appears to validate the decision.
• Anchoring — investors anchor to the last big gain and treat any pullback as a discount rather than a re-rating.

The uncomfortable reality is that a theme's risks grow as its popularity grows. When expectations are already high, good news produces a muted reaction and any disappointment produces an outsized one. That asymmetry — not a lack of faith in technology — is the argument for sizing positions deliberately.

A practical framework for investing when one trade dominates

Here is a step-by-step approach you can apply regardless of which theme is currently leading the market.

1. Split the portfolio into core and satellite

Define a core: a globally diversified, low-cost base that you intend to hold through cycles. Then define satellites: deliberate, sized bets on specific themes, sectors or individual companies. The core does the compounding; the satellites provide upside and interest. Write down the split — for example, the majority in the core and a defined minority in satellites — and treat it as a rule, not a mood.

2. Cap single-theme exposure explicitly

Rather than asking "is this a bubble?", ask "how much damage would a permanent write-down of this theme do to my plan?" If a 50% decline in that sleeve would force you to change your goals, the position is too large for your circumstances. Fixed percentage caps, reviewed annually, are crude but effective.

3. Diversify by return driver, not by ticker

Owning ten companies that all depend on the same capital-spending cycle is not diversification. Ask what makes each holding go up or down: consumer spending, interest rates, commodity prices, regulation, credit conditions, the business cycle. A portfolio with several distinct drivers can hold up when one narrative fades.

4. Use weighting rules to prevent drift

Equal-weighting, capped-weighting, or simply rebalancing back to target weights stops a winning position from quietly taking over. This is mechanically contrarian: you sell a little of what has run and add to what has lagged. It will feel wrong in strong trends, which is precisely why it works as risk control.

5. Judge expectations, not just price charts

Two stocks can both trade at high prices for very different reasons. What matters is what a business must deliver to justify its valuation, and how much room there is for disappointment. Without forecasting, you can still notice when a sector's valuation implies near-perfect execution while another sector's valuation implies scepticism. That gap is where risk and opportunity tend to live.

6. Rebalance on a calendar, not on a headline

Pre-commit to review dates — quarterly, semi-annually or annually — and to the bands that trigger action. Headline-driven trading produces activity that feels productive and often destroys returns through timing errors, taxes and costs.

Where capital can rotate if the tech trade keeps fraying

If leadership broadens, the beneficiaries are unlikely to be a single dramatic winner. More plausibly, returns would come from areas the market has treated as unexciting. These are categories to research, not recommendations:

• Equal-weight and broad benchmarks — they reduce reliance on the largest names by construction.
• Defensive sectors such as staples, healthcare and utilities, which tend to have steadier cash flows and often lag during euphoric growth phases.
• Financials, sensitive to credit conditions and interest-rate curves rather than to AI capital expenditure.
• Industrials and infrastructure tied to electrification, construction and logistics spending.
• Small and mid-cap companies, which have often been left behind when mega-caps dominate returns.
• International equities, which provide different sector mixes, currencies and policy cycles.
• Dividend-focused and quality-value strategies, which rely on cash generation rather than multiple expansion.

The point is not to predict which of these wins. It is to own several so that you do not need to be right about the rotation.

Gold, crypto and cash: three answers from outside the stock market

Gold as ballast, not a return engine

Gold has historically behaved differently from equities and has often been used as a hedge against currency debasement, geopolitical stress and negative real rates. It pays no income, so its long-run return has generally lagged productive assets. A modest allocation — sized so you would not be forced to sell during a drawdown — is the common use case. Physical bullion, allocated vaulted products and low-cost exchange-traded products all carry different cost, custody and tax profiles worth understanding before buying.

Crypto as a speculative sleeve

Crypto assets have delivered extraordinary gains and extraordinary losses, sometimes within months. Treat them as a high-volatility satellite rather than a core holding. Practical rules: size the position so a total loss is survivable, use custody you understand, avoid leverage, and be sceptical of yield products that promise stable returns from volatile collateral.

Cash as optionality

Cash and short-dated government instruments do two jobs: they reduce portfolio volatility, and they give you the ability to act when prices fall without being a forced seller. Holding some dry powder is not market timing; it is liquidity planning.

Risk management rules that survive any regime

• Define your maximum drawdown tolerance in currency terms before you invest, then size positions to fit it.
• Count overlapping exposure across funds, stock options and employer equity.
• Separate the decision from the outcome — a good process can produce a bad result, and vice versa.
• Write an investment policy statement covering target allocation, rebalancing bands and the conditions under which you would change your mind.
• Keep costs, taxes and currency in the equation; they are among the few variables you control.
• Review thesis, not price — if the reason you bought has changed, the position should change.

Mistakes investors make when a leading trade stumbles

• Rotating the entire portfolio into last year's laggards after a single quarter of outperformance.
• Abandoning a long-term plan because a headline made them anxious.
• Confusing a cheap valuation with a good business, or an expensive one with a bad business.
• Adding leverage to "win back" losses from a fading theme.
• Ignoring concentration added through employer stock and equity compensation.
• Overtrading, which raises costs and short-term tax bills while rarely improving outcomes.

Key takeaways

• A market-cap-weighted index is not automatically a diversified portfolio; a handful of names can drive most of the return.
• When one theme dominates, the rest of the market can be doing more work than investors realise.
• Outperformance by an ex-theme benchmark is a diagnostic, not a forecast.
• Set explicit caps on single-theme exposure and rebalance on a schedule rather than on news.
• Diversify by underlying return driver — rates, credit, commodities, consumer demand, policy — not just by fund name.
• Gold, crypto and cash play different roles; size each so that a bad outcome does not derail your plan.
• Rules written in calm periods are the ones that protect you in volatile ones.

FAQ

Does an ex-AI index outperforming mean I should sell all my AI-related holdings?

No. One reporting period is a data point, not a verdict, and broad technology exposure remains a legitimate long-term holding for many investors. The useful action is measurement: work out how much of your total portfolio depends on the same theme, then decide whether that figure matches your risk tolerance. Trimming to a pre-agreed weight is very different from exiting entirely.

How much of a portfolio should go into a single theme?

There is no universal number, because it depends on your time horizon, income stability, other assets and tolerance for drawdown. A practical test: imagine that sleeve falling 50% and staying there for several years. If that outcome would change your retirement date, your housing plans or your sleep, it is too large. Many investors keep thematic bets in a clearly labelled minority sleeve and leave the core globally diversified.

Is it better to time the rotation or simply rebalance?

For most people, rebalancing is more reliable than forecasting. Timing a rotation requires being right twice — when to leave and when to return — and the costs of being wrong compound. A written rebalancing rule mechanically trims what has grown and tops up what has shrunk, which keeps risk aligned with your plan without requiring a market call.

Conclusion

The lesson in a headline about an ex-AI index beating the standard one is not that artificial intelligence is finished. It is that dominant narratives are cyclical, and that portfolios built around a single story inherit that story's fragility. The investors who do well across cycles are rarely the ones who predicted the rotation. They are the ones who sized their bets so that being wrong was survivable, diversified across genuine return drivers, and followed rules written before the emotions arrived.

Start with the boring work: measure your true concentration, write down your target weights, set your rebalancing bands, and decide in advance how much a single theme is allowed to matter. Then let the headlines arrive without letting them rewrite the plan.

Risk note and disclaimer: Investing involves risk, including the possible loss of your original capital. Past performance does not indicate future results, and no allocation or strategy described here guarantees a profit or protects against loss. This article is educational content only and is not personal financial, tax or investment advice. Consider your own circumstances, do your own research, and consult a qualified, licensed professional before making investment decisions.