Revisiting asset allocation in an AI world

Artificial intelligence now dominates the U.S. stock market. This is reflected in the current level of market concentration, which we believe represents a fundamental shift in equity market structure. In this environment, investors may need to rethink their approach to portfolio construction and their ongoing asset allocation decisions.

The four major tech companies (Microsoft, Amazon, Google, and Meta) collectively known as the “hyperscalers” are driving an AI investment supercycle. These companies were on track to make over USD 300 billion in capital expenditures in 2025. By contrast, U.S. private fixed investment in other sectors has flattened out, while technology has accelerated, especially in the last year or so (Figure 1).

So companies are pouring money into AI, driving both economic growth and market performance. The flow of capital into AI‑related stocks also has powered a substantial shift in market capitalization, both for the S&P 500 Index as a whole and within the top 10 stocks in the index.

This shift has significant implications for portfolio construction, both from a multi‑asset perspective and from a bottom‑up security selection point of view.

Is this a bubble?

The surge in AI‑related stocks has drawn comparisons to the tech bubble of the late 1990s. However, we think it’s important to contrast some of the key differences.

When we look at return on equity (ROE) and price/earnings (P/E) ratios, we see that in the dot-com era those two metrics were inverted relative to each other. Not only did valuations soar extremely high, but that rise greatly outpaced improvements in profitability. Today, on the other hand, valuations and ROE have risen more or less in tandem (Figure 2), as the AI rally has been led by some of the most profitable companies in the market.

To us, this suggests that while investor enthusiasm may have reached unsustainable levels in certain pockets of the market, AI is likely to be a durable investment theme going forward. That being the case, investors may want to review their current asset allocation policies.

Market cap in the U.S. equity market has grown increasingly concentrated during the AI era. At the end of 2015, the 10 largest stocks in the S&P Index accounted for just under 18% of total index market cap. Over the next 10 years, their share rose to 38% as of mid‑2025 and almost 40% by year-end (Figure 3). 


But it’s not just that the top 10 stocks are driving S&P 500 market cap—and thus index performance. The composition of that group has also changed dramatically, with the mega‑cap tech stocks progressively driving out companies from other sectors.

In 2005, just one of the top 10 S&P 500 stocks (Microsoft) was a tech firm. By 2020, six of the ten were tech-related (counting Tesla as a tech stock, not an auto stock). By the end of October 2025, eight of 10 were tech-related. This has pushed the tech share from just 10.6% of top 10 market cap in 2005, to over 92% as of December 31, 2025.

In other words, the concentration at the top of the market has become even more tilted toward the tech sector. Twenty years ago, the S&P 500 was reasonably reflective of the broad U.S. economy. Now, it is more representative of the tech economy.

A shifting balance of factor risks

The tech share of top 10 market cap in the S&P 500 may well correct downward at some point. But, assuming (as we do) that AI remains a significant economic force, it seems highly unlikely to return to 10%.

In our view, this means the balance of risks in the U.S. equity market has permanently shifted. This is as true for passive investors—those who simply seek to track the performance of a particular equity benchmark—as it is for active investors who attempt to add value through tactical allocation and/or bottom‑up security selection.

This shifting balance, in turn, raises two key questions for portfolio construction and ongoing asset allocation:

• How should investors respond to a heightened level of factor volatility, more extreme factor return spreads, and an increased index weight in more speculative stocks?
• How can strategic and tactical allocation decisions be integrated in order to balance a portfolio in response to these new market concentrations?

A loss of defensiveness

The increased significance of factor risk can be seen in the fact that the change in sector composition in the top 10 S&P 500 stocks—the increased dominance of tech—also has dramatically shifted defensive positioning within the index.

In 2005, the top 10 stocks included Procter & Gamble, Johnson & Johnson, Pfizer, and Philip Morris—all stocks widely viewed as noncyclical, while the more procyclical names included Citi, Microsoft, and GE.

Today, on the other hand, the names that dominate the top 10—NVIDIA, Amazon, Broadcom, Meta, and Tesla—are all higher-beta stocks. So, while the top 10 stocks now represent approximately 40% of the index, on a beta‑adjusted basis they account for closer to 53% of market exposure. Their contributions to portfolio risk are now massive.

Doing the math, this means the aggregate beta for all of the other names in the S&P 500 is just 0.78. The distribution of beta within the index is very different from what it has been historically.

This distribution isn’t just something that active investors need to consider when building equity portfolios from the bottom up. It also shows that “passive” investing isn’t necessarily passive. As the market evolves, as the index evolves, passive investors could end up holding portfolios with very different risk exposures.

Risk management is even more critical now

From a quantitative perspective, the main thing to take away is that active risk management is probably more important now than it has been in the last 10 or 15 years. There are three patterns in the data that support this conclusion.

First, factor volatility remains elevated. We can see this by ranking the stocks in the Russell 3000 Index based on the various factors (size, style, quality, momentum, etc.) and then comparing daily equal‑weighted returns for the highest quintile with the lowest quintile within each factor (Figure 4).

What’s important isn’t so much the specific factors but the pattern over time, which shows that there have been distinct clusters of generalized factor volatility associated with specific market events—such as the dot-com bubble or the 2008–2009 global financial crisis (GFC). Typically, factor volatility has surged during such events and then slowly returned to more normalized levels.

But the factor spike associated with the COVID-19 pandemic broke the mold. While factor volatility did taper off as the market recovered, it did not return to the normalized level seen over the prior 30 years. Essentially, factor volatility has remained elevated for the last five years.

We believe this represents a long‑term structural shift, one driven by the rise in retail investing, the spread of zero‑commission trading, and the growing popularity of investment vehicles, like exchange-traded funds, that make factor‑based strategies more accessible to retail investors.

Why is the structural shift in factor volatility important? Because active portfolio risk has two sources: factor exposure and security selection. Assuming security selection risk remains steady, a rise in factor volatility will increase the contribution of factor exposure to total active risk—unless steps are taken to manage it.

Higher exposure to riskier stocks

The second issue pointing to a need for careful risk management is the level of speculative enthusiasm in today’s markets. To measure this, we developed a proprietary speculation score that seeks to identify companies with especially high valuations as measured by high enterprise value (EV)-to-sales ratios, and high projected sales growth. These tend to be lower‑quality, extremely volatile stocks with high levels of outstanding short interest.

By our measure, as of October 31, 2025, these stocks accounted for 28% of Russell 3000 Index market cap. So even investors who were passively tracking the index were still making an implicit decision to accept very high exposure to riskier stocks.

Extreme factor performance

The third pattern pointing toward rising portfolio risk is the exceptionally high return spread between high‑ and low‑beta stocks shown in Figure 5. The “Long” line tracks the performance of low‑beta stocks. The “Short” line shows returns on high‑beta stocks, while the “Long-Short” line is the return spread.

Over the six months ended October 31, 2025, the high‑beta stocks collectively returned a whopping 47.04%. Lower‑beta stocks were down 2.04% over that same period, producing a negative long/short spread of more than 49 percentage points—although that spread narrowed considerably in the last three months of the year. The last time the spread was that extreme was coming out of the GFC.

But what is anomalous about the recent extreme in relative beta performance is that it didn’t happen while the market was coming off a major bottom—the typical historical pattern. The “Liberation Day” market correction in April 2025 was hardly equivalent to the GFC or the COVID-19 pandemic.

Typically, when high‑beta stocks outperform low‑beta stocks, both groups rally but high‑beta stocks rise faster. Over the last six months, however, high‑beta stocks rose by a massive amount while low‑beta stocks actually lost ground. The bifurcation in the market has been very extreme.

Index betas may be misleading

By definition, an index’s beta with itself will always be 1.0. So a passive portfolio that perfectly tracked the S&P 500 Index also would have an index beta of 1.0. But that number may obscure more than it reveals.

Ranking stocks by their index beta, the two highest-beta quintiles accounted for more than 63% of S&P 500 Index market cap as of December 31, 2025 (Figure 6). The last time their index cap weight was this high was during the dot‑com era. 

Suppose that an active investor constructed a portfolio with a beta of 0.94 or 0.95 to the S&P 500 Index. If they did nothing while the index became more concentrated in higher‑beta stocks over the past few years, their index beta might have fallen to just 0.82 or 0.85. Again, this essentially would have been an active decision. Investors might be comfortable with that choice, but they need to be aware that they have made it. 


Building resilient portfolios

So how should investors think about asset allocation in the AI era? Are there specific adjustments that investors need to consider, to either tilt their portfolios toward areas of opportunity or to better manage risk exposures?

We believe that portfolios should be resilient across time. Typically, we would define this as being globally diversified, avoiding structural style bias in equity, and investing across both core and plus fixed income sectors. That approach has been partially challenged over the past few years, given the dynamics discussed above.

However, we remain convinced of the value of diversification over the long term. We also believe the opportunity set for asset allocators is broadening as markets and financial instruments evolve. Innovation in investment vehicles now allows portfolio managers to deliver more complex institutional‑level strategies with less friction. In our view, investors would do well to use these innovations to further incorporate alternative sources of return and enhance diversification in their portfolios.

Three areas of long‑term opportunity that we see are:

• Alternative sources of return: Option writing strategies, for example, offer the potential for enhanced returns and can complement traditional fixed income plus sector approaches like high yield.
• Better portfolio hedges: Going forward, we believe that the economic risks will be more balanced between inflation and growth than they were in the post‑GFC period. This suggests that hedging strategies beyond fixed income duration, such as options‑based approaches that dynamically adjust portfolio risk or reduce downside exposure, should be incorporated.
• Building better alpha: We also favor more sophisticated approaches to architecting alpha. In a potentially return‑constrained world, alpha becomes increasingly important. Accordingly, investors may wish to consider incorporating complementary alpha sources and avoiding any unwanted factor biases.

Tactical allocation opportunities

We also currently favor a few portfolio themes that are more tactical, based on what we are seeing in the marketplace. In our view, these opportunities provide an additional degree of diversification in a market that is narrowly focused.

Investors may want to consider  maintaining exposure to long‑term structural growth drivers, particularly innovation and AI, while selectively adding risk to underappreciated areas that could benefit from broadening market participation, favorable fundamentals, and policy tailwinds.

For example, we currently favor exposure to market broadening in areas like health care, small‑ and mid‑cap equities, and emerging markets (EM) equities.

• Health care: The sector offers potential AI‑related investment opportunities that extend beyond the tech sector itself. The short‑term negative impact of the Trump administration’s health care policies has weighed on the sector, but we believe the sector stands poised for growth and features attractive long‑term valuation opportunities. We see a number of opportunities for technology to boost innovation, enhance efficiency, and lower costs. Global demographic trends and a growing need for chronic disease management are also key tailwinds.
• Small‑ and mid‑cap equities: Smaller companies stand to benefit from lower interest rates, increased merger and acquisition activity, and stronger earnings growth supported by higher fiscal and capex spending in 2026. Further, while the broad Russell 2000 Index performed well in 2025, earning over 12.7%, the higher‑quality S&P 600 Index lagged materially, trailing the Russell 2000 Index by over 6.8 percentage points. A quality approach may be rewarded going forward.
• EMs: We believe these markets stand to benefit from improving domestic policy, lower inflation, a weaker U.S. dollar, and AI-related growth trends though their exposure to technology and commodities sectors.

Conclusions

Investors should recognize that AI‑driven market concentration reflects genuine economic transformation but also requires active risk management. Structural changes in market participation and trading technology have permanently elevated factor volatility, demanding enhanced risk management capabilities.

Diversification needs to be redefined. Traditional geographic and sector diversification remain important, but investors may need to consider whether traditional techniques need to be supplemented with alternative return sources and positioning portfolios beyond areas of recent leadership.