Is everything becoming AI – and how can you respond without selling equities?
Value equities can play an important diversifying role, but only if done properly. Passive approaches will fail to deliver.
Autheurs
AI offers potential transformational change on the scale of the Industrial Revolution. It may well deliver but it is equally reasonable to ask: what if it doesn’t? Or, even if it does, what if that success is already priced into today’s lofty share prices? Your investment portfolio is more exposed than you realise and many of the traditional places you could have sought shelter are now also hitched to the AI-bandwagon.
One area that stands out as offering genuine diversification potential is Value equities but, importantly, only if done properly. And, by properly, this means actively. Most passive value equity benchmarks are stuffed full of the same technology names that you’re seeking to diversify away from. If your aim is to “AI-proof” your portfolio, they are a one-way ticket to disappointment.
Is everything becoming AI?
Duncan Lamont, Head of Strategic Research
Most investors in US equities have roughly half their money in technology stocks, once you include Amazon and Tesla (Consumer Discretionary), and Alphabet and Meta (Communication Services). That proportion has marched higher in recent years. Nvidia, with its $4.5 trillion free-float market capitalisation and 5.5% weight in the MSCI World index, is now larger than every other developed stock market on the planet.
Given that the US now makes up over 70% of the global developed market, investors in global equities fare little better. Long-held beliefs that a low-cost passive portfolio of global stocks delivers diversification – and avoids having too many eggs in one basket – demand urgent re-evaluation.
Weight of IT sector + Alphabet, Amazon, Meta, Tesla
MSCI USA vs MSCI World
Past performance is not a guide to the future and may not be repeated
Data to 31 December 2025. Before September 2018, Alphabet and Meta were in the IT sector but sine then they have been in the communication services sector. Over the period shown in the charts, Amazon and Tesla have always been in the consumer discretionary sector. Source: LSEG Datastream, MSCI, Schroders
Even this understates things. Here are four additional examples of areas with AI dependency.
Utilities: From defensive plodders to AI‑dependent growth stocks
Utilities have morphed from being a sleepy "defensive" (slow-growth/high-dividend) investment to a growth-oriented sector driven by unprecedented energy demand. The correlation is two-fold: AI requires massive, reliable power, causing utilities to become essential "shovels" in the AI boom, while simultaneously adopting AI to modernise ageing grids. US utilities have risen 25% and 16% in the past two years — almost identical to the tech‑heavy broader US market.
The link with AI is particularly strong for unregulated utilities near to data centre hubs or those with a nuclear component. The top US performers are up 320%, 441%, and 626% cumulatively over the last three years alone. Past performance disclaimers should be flashing brightly at this point.
Meanwhile, dividend payout ratios have been cut (the sector payout ratio is comfortably below its five and 10-year average) as capital is diverted into capex, to meet AI-driven electricity demand. If that demand trajectory disappoints, the sector could face excess‑capacity problems later.
Real Estate: data‑centre dependence
Real estate is another sector that in the past investors have turned to for diversification. Here too, AI is becoming entangled. Specialist data centre REITs are one of the largest sectors of the US REIT market. They’re about 10% of the market today vs 4% 10 years ago. Adding in other, more diversified, REITs which have meaningful exposure to data centres pushes that figure closer to 20%. AI demand is embedded deep into the REIT complex.
Software‑driven platforms
There are also other companies such as Airbnb, Uber, Doordash, Netflix, Disney which are software-driven platforms. Their share prices embed an expectation that AI will deliver productivity gains for them (such as content development) and/or an improved user-experience for their customers.
Financials too
Payments companies — established names like Visa and Mastercard and newer entrants such as Block — are structurally tied to online and digital ecosystems. Again: AI exposure.
Put together, investors in US equities have done exceptionally well out of these exposures but now are likely to have a vast, concentrated bet on a technology/AI narrative. This is not simply about correlations going to 1 in a crisis — these are fundamental linkages.
Can you do anything to manage this risk?
The aim here isn’t to claim AI will fail, or even that AI-related companies are overvalued for what they might deliver, but to explore how investors can guard against this risk without giving up equity exposure.
2025 showed the benefits of global diversification. Europe ex UK, UK, Japanese and emerging market equities returned 37, 35%, 25%, and 34% in USD terms vs 18% for the US. International diversification is an option that everyone should consider.
But we also have to remember that AI is everywhere. Stocks with a hint of an AI link in non-US markets have also performed exceptionally in recent years and make up significant proportions of their domestic/regional stock markets.
The world’s largest chipmaker, TSMC, for example, is 11% of MSCI EM and 58% of Taiwan’s market. The next four largest EM stocks – Tencent (AI platforms), Samsung Electronics (memory), Alibaba (cloud and AI services), SK Hynix (memory) – are all explicitly AI‑linked. Many large companies across developed markets have also benefited meaningfully from the AI wave.
Non-US markets remain cheap relative to the US, and catch-up potential exists. But that is not the same as saying they will shelter investors if an AI-driven sell off comes. Many would also be heavily exposed.
Traditional sector tilts are equally complicated, for reasons already outlined.
Value equities as a hedge against AI-risk (without sacrificing equity exposure)
Value investing remains underappreciated in this context. In the US, the correlation between value equities and the “AI trade” — using semiconductors and equipment as a proxy — has recently been low and falling. During the Dotcom selloff it even turned negative.
This suggests value may offer valuable and significant diversification benefits.
Value equities have had a low correlation with AI-stocks
Rolling 24-month correlation: S&P 500 pure value vs S&P 500 semiconductors & equipment index
Past performance is not a guide to future performance and may not be repeated.
Value is S&P 500 pure value total return index, semis are S&P 500 semiconductors and equipment total return index. Semiconductors used as a proxy for AI-stocks. Data covers 30 June 1996 (inception of S&P 500 semiconductors & equipment index) to 31 December 2025. Source: LSEG Datastream, S&P and Schroders.
Even more importantly, if we isolate those quarters where semiconductor stocks fell, value equities were roughly flat, on average, versus a 12% average decline for semiconductors. In deeper drawdowns (semiconductors down by 5% or more), semiconductors fell 15% on average. Impressively, value stocks fell only 1%.
Value equities have delivered significantly better outcomes in down-markets for AI-stocks
Median quarterly return in quarters where semiconductors fall
Past performance is not a guide to future performance and may not be repeated.
Value is MSCI pure value total return index, semis are S&P 500 semiconductors and equipment total return index. Semiconductors used as a proxy for AI-stocks. Data covers 30 June 1996 (inception of S&P 500 semiconductors & equipment index) to 31 December 2025. Chart isolates those quarters where semis fell in value; non-overlapping periods are used. Source: LSEG Datastream, S&P and Schroders.
One reason for this is the “margin of safety” that value investors benefit from, when buying companies on cheaper valuations. By only paying low prices relative to conservative appraisals of earnings or asset values, investors avoid areas of the market that require large growth to justify the price. So, when those areas of the market that have enjoyed a speculative boom retreat, value investors are often much more immune to those falls.
This analysis is based on US value stocks but, in today’s environment, the case is even stronger for value outside the US. If investors turned off US stocks specifically, because of high valuations or any other reason, US growth and value stocks would likely both suffer (to varying degrees). Non-US value should be more resilient in such a scenario.
Buyer beware: passive “value” probably won’t help
The shift from active to passively managed strategies has been one of the biggest changes in investor behaviour of the past decade. While there may be some justifiable reasons, this specific situation is one where passive investing is likely to disappoint. Most value indices contain large allocations to the very same AI-names investors are seeking to avoid.
Most value indices have significant exposure to Magnificent-7 and/or other technology names
Largest five holdings
MSCI USA Value | MSCI USA Value Weighted | MSCI USA Enhanced Value | Russell 1000 Value | S&P 500 value | S&P 500 Pure value |
Alphabet A | Apple | Micron Technology | Berkshire Hathaway | Apple | Ford |
Meta Platforms | Microsoft | Cisco Systems | JP Morgan | Amazon | Bunge Global |
JP Morgan | JP Morgan | Intel | Alphabet A | Exxon | General Motors |
Berkshire Hathaway | Exxon | General Motors | Amazon | Walmart | Mosaic |
Exxon | Amazon | AT&T | Alphabet C | Tesla | Centene |
As at 31 December 2025. Source: MSCI, FTSE Russell, S&P.
This is not an accident. These value indices are often designed to avoid taking large tilts vs the broader market, as that is what many investors desire. Some also allocate a stock partly to a value index and partly to a growth index based on its characteristics rather than purely in one or the other. This leads to a lot of overlap.
I am confident that many investors in passive value strategies are unaware that they are allocating significant sums of money to the likes of Alphabet, Amazon, Meta, Microsoft and even Tesla. They could be in for a rude awakening.
The “pure value” index — used in the earlier analysis — is different: it selects only the cheapest S&P 500 stocks and weights them by value characteristics, not size. This creates a much more differentiated portfolio and stronger diversification benefits. But it also takes no account of whether any of these companies are cheap for good reason. Just because a company is cheap vs its historical valuation does not automatically make it a good investment. There is a significant risk that some of the “pure value” index constituents turn out to be duds.
Why active value makes most sense
Active approaches can deliver the diversification benefits of pure value — low correlation to AI, resilience in AI related selloffs — while avoiding exposure to value traps. The goal is not simply to be cheap, but to be cheap relative to fair value.
This creates a portfolio that:
• has a credible case for performing well if AI disappoints,
• maintains the long term return potential of equities, and
• can still perform if the AI driven bull market continues.
The view from the frontline
Simon Adler, Head of Value Equities
As Duncan makes clear, AI may well deliver, particularly operationally. But as we’ve seen, time and time again, investor returns are often materially different (and frequently worse) than the explosive growth experienced by disruptive new technologies. There are echoes of the dot com era here, in that the explosion of the internet and related businesses and services was indeed revolutionary, but resulted in too much capacity and capital chasing too few opportunities, with disastrous consequences for investors who bought in at peak tech bubble enthusiasm. Value investors did much better in this period, incidentally, which could serve as a useful precedent.
Value investing offers a margin of safety which AI stocks long ago abandoned
Value investors tack a different route. A disciplined value selection process will steer investors to precisely where peak market euphoria is not. At Schroders, we screen for, and solely analyse, stocks in the cheapest parts of the market. It’s worth noting this now includes some companies and industries that have already sold off largely because of AI-related fears. So “AI fever” is providing opportunities for us, too.
It takes courage, consistency and a data-led approach to eschew investing in “new paradigms” in which stocks or even entire industries have supposedly reinvented themselves as high-growth, high-returning businesses. Our requirement for a large margin of safety every time we invest steers us well away from some of the sector examples Duncan refers to. We have no issue with investing in utilities, real estate or financials based on a prudent assessment of their earnings power, and we do invest in each of these sectors across our range of strategies. But we will not increase our assumed earnings power, multiples or fair values merely because companies in these sectors are exposed to AI which might deliver in the future.
In our view, many investors, either deliberately or unwittingly (via passive exposure) are sacrificing most, if not all, of their margin of safety by paying up for AI exposure. A disciplined value approach to stock selection, by contrast, will allow for a wide range of outcomes that could still make attractive returns under various scenarios. This is in stark contrast to the range of AI-related themes, sectors and stocks where investors are increasingly pricing these businesses for perfection.
Avoiding value traps
Just as we would never pay up for the “next big thing”, we also don’t naively buy all of the cheapest cohort of stocks in a given market. This is a key differentiator vs. passive value funds which make no distinction between “value gems” and “value traps”.
We spend more time thinking about what we shouldn’t invest in than what we should, as part of a disciplined and repeatable process.
Seeking out and avoiding value traps is in obvious contrast to passive value which “never learns”. We can even see a quantifiable (albeit hypothetical) consequence of this if we look at the performance of our stock recommendations vs. the broader market, and particularly vs. the cheapest segments of the market (as a proxy for passive).
The table below shows that over three year periods from 2016-2024 the median return was 17% for the MSCI ACWI, while the value “passive proxy” (the cheapest 20% of the market) returned 15%. Looking at our recommendations, however, our “passes” were lower still at 13%, but our “buys”, at 28%, were significantly better than both our passes and the value pool. This suggests that hard work and analysis really does add value over simple passive value exposure.
Median returns from 2016 to 2024
Past performance provides no guarantee of future results and may not be repeated. The value of investments and the income from them may go down as well as up and investors may not get back the amounts orginally invested. Exchange rate changes may cause the value of investments to fall as well as rise. Source: Schroders, Refinitiv. As at December 2024.
Another way of assessing “value add” over passive is to look at an overall assessment of stock risk which passive value won’t consider. We assign every stock we look at a risk score from 1 (least risky) to 10 (most risky) to give an indication of the likelihood of permanent capital loss (our definition of risk). These scores are not intended to be an exact science, but when we map the risk score assigned at the time of research against the subsequent maximum share-price drawdown over the next three years, a relationship does appear to exist. Companies scored ≤3 have roughly half the median drawdown of those scored ≥8.
The relationship becomes even more compelling when we split the samples by buys and passes. Again, this suggests an enduring competitive advantage over passive value funds.
Conservative and dull wins in the long run
When it comes to modelling, valuation, and subsequent investor returns, being conservative on medium term estimates might sound a little dull, particularly in the context of the “new paradigm” narratives associated with AI. But objectively, our approach shows some serious efficacy.
Indeed, as one of many examples of using our huge bank of data to enhance and test our process, we measured the accuracy of our estimates of normalised earnings for the companies we analyse vs. consensus. We found that we were more accurate over multiple time periods, as shown below. Our estimates were also, on average, a little too high vs. achieved outcomes but by far less than consensus, especially the further out you look. And this is for the cheapest shares out there: imagine how over-excited consensus might be for the growth companies! Only time will tell, but if consensus proves to be similarly over-optimistic for some of these AI-exposed segments of the market, the consequences for those companies’ investors is unlikely to be positive.
Source: Schroders, Refinitiv. As at December 2024. Based on median accuracy for all stocks the Global Value Team have made EBITA estimates on since 2015. Consensus is taken on the same date the team entered their estimate for FY ending 3, 4 or 5 years later. Outcomes are based on reported EBITA 3, 4 or 5 years after the estimate. 1,154 valuations were made by the Global Value team. Not all stocks have consensus coverage. Coverage declines over longer horizons as some valuations lack 4- or 5-year projections. Our forecast and estimates are based on assumptions within the bounds of what we know and are not guaranteed.
Could Schroders Global Value team ever invest in AI-related stocks?
Absolutely; when they’re cheap enough! As noted above, we’re already analysing companies that have sold off largely because of AI-related fears. Looking back over recent decades, we ended up buying six out of the ten biggest stocks from the dot com boom over the following 20 years. Obviously, this was only after their shares prices fell dramatically. We’re sufficiently open minded to consider virtually any investable business, sector, region or theme – at the right price.
This is where the power of sophisticated screening and portfolio management tools come in, as they have no preconceived biases about a given stock or sector. An example of this is Schroders’ portfolio heat map, as shown below, which allows us to compare the composition of portfolios to that of various fundamental valuation screens. The primary output of the tool is two heat maps divided by region or country and sector, one reflecting the composition of a portfolio and the other the composition of the portfolio’s opportunity set. The tool provides a visually quick way to identify potential “white spots” in the make-up of portfolios – and stress test whether there are potential pockets of diversification available to us that haven’t been explored.
An example screen using the screen heat map (upper grid) and portfolio map (lower grid). Shown for illustrative purposes only. Our proprietary tools are designed to enhance the research and evaluation process but do not guarantee favourable investment results.
The intention isn’t to mimic the make-up of the screen, but rather to stress test whether we have explored all the areas available to us – while maintaining our style discipline – right from the start of the process.
At some point in the future we could see AI-related stocks and sectors appear on both our screens and these heatmaps, if – and only if – their valuation meets our strict screening criteria.
Conclusion
AI may transform everything – or not. What is clear is that most equity portfolios are now running a far larger implicit bet on the AI narrative than investors appreciate. And many of the usual diversifiers have all become entangled in the same theme.
If investors want to reduce their dependence on a single powerful narrative without selling equities, value investing stands out. The historical data show that value has tended to hold up well when the AI trade stumbles. But most passive value strategies won’t help: they are full of the same mega‑cap technology names that dominate the broader market.
An active value approach offers a way to maintain equity exposure, reduce AI concentration risk, and build a more resilient return profile. In an investment world increasingly shaped by a single theme, that kind of diversification is worth a great deal.
Subscribe to our Insights
Visit our preference center, where you can choose which Schroders Insights you would like to receive.
Autheurs
Topics