The valuation gap between the biggest stocks (the megacaps) and the rest is unlikely to persist indefinitely.
“By 2005 or so, it will become clear that the Internet’s impact on the economy has been no greater than the fax machine’s,” declared a former winner of the Nobel Prize in Economics in 1998. This famous quote clearly demonstrates how even the sharpest analytical minds can struggle to forecast the pace of technological progress.
It is now increasingly consensus that a booming artificial intelligence (AI) industry is driving the next technology revolution. For investors, however, the most important question to address is whether the expectations embedded in financial markets today project a realistic path ahead.
AI stocks and index concentration
The current composition of the S&P 500 highlights just how important a small number of companies linked to AI have become. While each of the companies in the Magnificent Seven are geared differently to the AI theme, this group of stocks now make up nearly 35% of the S&P 500 market cap and have driven over 70% of returns since the beginning of 2023. This outperformance has also seen valuations expand. While the rest of the S&P 500 trades on a 12-month forward earnings multiple of 19x, the largest 10 stocks in the index now trade on 29x (Exhibit 9).
The valuation gap between the biggest stocks (the megacaps) and the rest is unlikely to persist indefinitely. If the broad AI ecosystem generates sufficient revenues to justify the earnings expectations already assumed for a handful of companies, the “rest” should catch up over time. If instead, the broader corporate universe does not see the clear use case of these technologies and are unwilling to pay for them, then a “catch down” scenario is more likely.
The strong fundamentals of the megacaps, both relative to other parts of the S&P 500 today, as well as relative to the 2000s tech bubble, provide some comfort that a major “catch down” is unlikely. Collectively, the Magnificent Seven have cash on their balance sheets to the tune of around $460 billion, according to their most recent set of earnings reports. The current spread of just over 10 basis points on Apple’s corporate bonds maturing in 2032 is one example of how the market perceives the company’s quality.
Another key difference relative to 2000 is how delivered earnings growth has supported much of the move in stock prices (Exhibit 10). Take Amazon, whose P/E ratio has declined from 48x to 35x over the past 12 months, but where huge earnings growth has resulted in a 46% one-year return. Or Nvidia, the poster child of AI, whose multiples have risen over the past 12 months, but where a 145% change in 12-month forward earnings has played a major part in the stock’s 207% return.
In other words, what we are seeing today is reality over hope, rather than the hope over reality that prevailed during the dotcom bubble (Exhibits 11 and 12).
Understanding the AI value chain
To understand how a “catch up” scenario could play out, it’s necessary to look much further out along the AI value chain (Exhibit 13). In very simple terms, AI companies can be bucketed into five key groups:
- AI hardware (for example, Nvidia in the US, ASML in Europe and TSMC in Taiwan), which are the companies that drive the design and manufacture of the semiconductors that are key to generating computing power;
- AI hyperscalers (for example, Amazon’s Web Services business or Google Cloud), which are the companies that provide physical AI infrastructure such as cloud services and data centres, create custom silicon chips, and build large language models that can be used by other companies;
- AI developers, which can range from small app builders to existing enterprise software companies (for example, Adobe or Microsoft) that leverage hyperscaler technologies to provide solutions for end users;
- AI integrators, which are the larger organisations that have sufficient technology functionality to build their own AI solutions, as well as the IT services companies that support them;
- AI essentials, which include companies that are less directly impacted by the technology itself, but provide the resources that enable the whole AI value chain to work, whether that is energy, air conditioning, raw materials or even the data to train models.
A megacap monopoly, so far…
Despite the huge array of companies that will at some point find themselves in one (or more) of these AI buckets, only a handful of megacap companies are viewed as winners today based on current earnings expectations, with the flow of capital expenditure (capex) across this small number of megacaps creating a virtuous earnings cycle over the past two years.
According to data from S&P Global, just five AI hyperscalers are currently projected to spend more than $1 trillion in capex collectively from 2024 to 2027, which in turn is driving massive revenue expectations for AI hardware names. Nvidia is the prime beneficiary, where annual revenues have increased from $4bn in 2014 to an expected $61bn in 2024. At the same time, the top 10 stocks in the S&P 500 are now responsible for over 40% of research and development (R&D) expense, despite only representing 13% of the S&P’s revenues (Exhibit 14).
Return on investment is now under the spotlight
The substantial gap between the revenue expectations of hardware companies and the revenue growth that can be generated by the AI ecosystem is an issue that has already attracted significant attention1. If the developers and the integrators can’t generate sufficient profit, this weakness will eventually spread up the value chain. As the initial hype around AI starts to cool, the question investors are posing is a simple one: “show me the money?”
The good news for technology bulls is that AI adoption appears to be increasing. McKinsey’s Global Survey on AI from earlier this year, for example, showed that the proportion of companies that have adopted AI in at least one business function had jumped from 55% in 2023 to 72% in 2024, with an even greater jump in the proportion of businesses using generative AI.
Quantifying the revenue benefit, however, is much more challenging. For the hardware providers, while forecasts are still uncertain, a large proportion of AI-related revenues are coming from a small number of hyperscalers. But for the developers, whether they’re selling software to a law firm in a bid to reduce headcount, or new technology to a pharmaceutical company to enable accelerated drug testing, these increased revenues are going to be scattered right across the economy. Crucially, this makes it far harder for analysts to gauge the outlook for earnings.
Assessing AI investment opportunities
As AI hype turns to AI reality, where along the value chain should investors be positioning for further upside?
So far, investor excitement has primarily been anchored in the first two of our categories, hardware and hyperscalers, where companies are often located in the technology and communication services sectors. High levels of valuation dispersion in these categories suggest that opportunities for skilled stock pickers persist, but investors must recognise that any earnings disappointment could lead to substantial volatility. These categories are also likely to be the most exposed to escalating trade tensions between the US and China, although the nature of any new policy is yet to be known. Today’s wide margins make tariffs somewhat of a secondary issue, but new trade restrictions that look to restrict the availability of sophisticated technology would pose greater risks.
We find ample opportunities in the AI essentials bucket, where companies often trade on less demanding valuations and many are now seeing a significant acceleration in revenues. Take the utilities sector as one example, where data centre demand for electricity is set to more than double by 2026 relative to 2022 levels.
The biggest AI-driven winners may well be found in the developers bucket over time. The category already has many incumbents in the form of enterprise software companies, who are now integrating AI to enhance their product line-ups. Yet the growth of the internet in the early 2000s highlights just how long it can take to fully understand how transformative new technology can be. Few investors anticipated the success that lay ahead for companies such as Amazon and Uber, which later leveraged the huge amount of capex first invested by others, and the same could yet be true for AI. A diversified approach will be critical given the high degree of variability in the future prospects of the less established developers.
Finally, practically every S&P 500 company will be hoping to achieve “integrator status” over the coming years. The key challenge for equity analysts will be to differentiate between those companies “talking the talk”, and those whose earnings reports demonstrate that they really are “walking the walk”.
Conclusion: Look beyond megacap tech stocks for AI opportunities
In our view the gap between the valuation of mega-cap tech and the broader S&P 500 is unsustainable. However, unlike 2000, we see a ‘catch up’ scenario as more likely than a ‘catch down’ scenario. From here, investors should focus on opportunities that will prevail right along the AI value chain. Investors need to weigh future potential earnings against what is already embedded in the price. Cheaper valuations and less demanding earnings expectations outside of megacap tech stocks suggest that even AI bulls should be positioned for further broadening across sectors in 2025.