Our view on AI and capital allocation 

Utilising a total portfolio approach is part of our DNA – for over 80 years we’ve been a trusted partner for institutional investors on how to help solve their overarching financial objectives and investment constraints. This requires a culture of collaboration, where asset class specialists have a much greater awareness of total portfolio outcomes rather than siloed single-asset success measures, an integrated approach to assessing opportunities and exposures at the total portfolio level and adaptability to help ensure portfolios can be best positioned for the evolving market environment.

At Mercer’s global asset allocation committee, not only do we draw on the single and multi-asset class specialists for the insights of Mercer’s 500-person research & portfolio management team¹, but we consult a broad set of buy-side and sell-side firms globally for additional perspectives to test our thesis and thinking.

As part of our evaluation for the outlook for equity investing in the context of total portfolios, we have been assessing whether, in light of AI’s transformation potential, traditional valuation metrics, such as P/E and Shiller P/E, remain appropriate for today’s investment environment.

Analysis from our global multi-asset research team shows that:

• In over 150 years of data, using these traditional valuation metrics, the US stock market has rarely been more expensive (except for the pre-Dotcom crash). We are also at one of the most extreme levels of concentration in the US stock market ever, driven by the magnificent seven.^2,3

Source: Mercer Multi-Asset Group research, as at 30 September 2025.

Source: Mercer Multi-Asset Group research, as at 30 September 2025.

• Historically, pockets of concentration have tended to unwind, and today’s winners rarely sustain their position in the top ten in the years/decades that follow. We believe this makes intuitive sense for two reasons:

1. Pockets of concentration often result from a run-up in P/Es to very extreme levels e.g. in dotcom, many of the largest positions had forward P/Es of > 60 (compared to c.30s today).
2. Companies that have been able to make materially above market earnings attract new competition, thus reducing profitability to more normal levels.

• Context matters though. We do not think the US stock market (overall) is as expensive as traditional metrics imply due to three factors:

1. Profits. While P/Es have been elevated for the AI-related stocks for an extended time period, they have consistently delivered strong earnings, helping to justify the market’s optimism⁴. In fact, Price-to-Earning Growth figures for the Magnificent Seven are in line with the broader market. Looking forward, we believe AI is a tailwind that deepens their moats because it enables them to better leverage their data, computer power and distribution that could help them sustain their historical strong earnings growth.
2. Existing Moats. Many of these large companies are deeply embedded as platforms across multiple solutions for their customers, making it harder for them to be competed against, whilst AI enables toolkit expansion to support growth of existing customers, reducing the necessity to place greater emphasis on new customer acquisition.
3. R&D Capacity. The enormous free cashflow enables capex (as has been hugely evidenced in the last year) which further delays competitive threats.

• As a result, we believe that on balance, today’s market valuations whilst elevated, may not be as extreme or unsustainable as they seem on face value using traditional valuation metrics. We believe that the best way to navigate concentrated and expensive US equities is through the discipline of a globally diversified opportunity set that includes international (ex US) equities, emerging market equities and ideally small cap equities. In addition, a well-structured active management programme can embed higher quality active decisions than the alternative of top-down calls that naively underweight the US market.

Source: FactSet, Mercer, as of October 2025

Many private equity firms are spending money on AI and leveraging it in various ways. The depth and breadth of Mercer’s market coverage provide us with an extensive view of different ways general partners are using AI and getting value from it (at the portfolio company level and to enhance their own processes).

While most agree that AI will have a large impact on society, the speed and degree of adoption and implementation vary significantly across markets, asset classes and regions. The current opportunity set largely remains centred on the first order effects with the second order and lateral impacts likely to be determined over years to come. In our view, this may create potentially multi-decade cyclical and secular winners and losers across nearly every sector. 

We believe hedge funds are well positioned to take advantage of these dynamics while managing through the variable speeds of technological adoption and evolution. For adoption and implementation of AI within the investment processes, we continue to see wide dispersion in application. 

• Systematic Macro managers have long been at the forefront of technology adoption with big data, alternative data, machine learning and now AI which encompasses and integrates many of those inputs. Many systematic managers are exploring stand-alone and holistic AI/machine learning strategies. However, the majority remain in testing phase or with trial amounts of live outside capital. If proven successful, these should represent additional diversification.
• For more fundamental based managers it remains early, but primary applications emphasise speed of insights and forecasts with a focus on data processing, synthesising, and analysis. 

We are aware of concerns in the media and wider markets as to whether AI represents a market bubble. While we do not believe it is a market bubble at present, we do believe that valuations of AI and AI-adjacent companies are priced for continued strong earnings growth. In our view, this is all the more reason to help ensure that portfolios are well-diversified and applying a total portfolio approach, taking into account the fact that AI risk may be present across equities through the Magnificent Seven, private markets through the AI-driven venture capital boom, and infrastructure through data centre and data centre-adjacent investment, driven by the need for computing capacity to power the development of AI.