In 2024, the markets delivered strong returns; however, there was a lot of volatility beneath the surface. We saw plenty of debate tied to interest rates, unemployment, war, trade tensions, political unease, and AI enthusiasm. As we enter 2025, the focus for many investors is on assessing the durability of the AI trade. For the second year in a row, the top 10 weights in the Russell 3000, which includes the “Magnificent 7” stocks (AAPL, AMZN, GOOG/GOOGL, META, MSFT, NVDA, TSLA*), have pushed the market to an all-time high due to the significant potential of the AI opportunity (Exhibit 1). Given the market cap size and significant outperformance of the Magnificent 7, what has become of increasing concern is the narrowing of the market with the top 10 weights now representing over 30% of the Russell 3000, an all-time high (Exhibit 2). The narrowing of the market has been driven by investors’ belief that we are at the dawn of a new era of accelerated computing technology (AI). Similar to the Mainframe era, PC era, and mobile/server/internet boom era, there are generally only a handful of companies that capture the majority of value initially and thus offer the biggest returns for investors. Just think about how much shareholder value Apple, Inc. (AAPL) created with the launch of the iPhone in the mobile era (hint: shares +24% annualized since 2008 or a whopping 3,800% overall return). We see similar expectations playing out with the Magnificent 7 capturing significant shareholder returns as they have allocated the most capital thus far and are in the best competitive position to lead this new level of AI innovation.

Exhibit 1: Top 10 Weights Outperformance During 2024

Exhibit 2: Top 10 Weight over time

After some of the initial AI euphoria subsides, we believe the market will begin to show more skepticism on the actual return on investment (ROI) of these significant AI-related capital expenditure (capex) investments made by the Mag 7. The market has already rewarded these companies without much AI project monetization. For now, the ROI on these AI projects cannot be disproven, but the amount of spending is of increasing concern with large-cap technology company capex expected to ramp up significantly, and it far outweighs any previous level of spending from the past 15 years. Capital intensity for the Mag 7 is near 12% of total revenue, amongst the highest in the market. This level of capital intensity is similar to the energy sector’s capex intensity levels back in 2015, which was the peak of the domestic debt-funded shale boom (Exhibits 3 and 4). During the shale boom, the energy industry chased production growth with heavy development spend, but declining free cash flow and returns ultimately destroyed shareholder value. Similar comparisons can be made to the tech bubble and telecom capital intensity levels in 2000. Historically, the Mag 7 companies have been valued on their asset-light business models. But the market may change its view on how to value these more capital-intensive business models, which puts the overall market (and perhaps your portfolio) at risk given the increased concentration of the index.

Exhibit 3: CAPEX

Exhibit 4: Capital Intensity

We believe the main driver of significant capex spending is the “fear of missing out”. Several of these companies missed out on the mobile opportunity or the cloud opportunity and they are not missing what could potentially be a significant long-term driver of revenue growth. Many Mag 7 CEOs have discussed their fear of underinvesting in this opportunity. But are AI expectations too high? We expect Nvidia's data center revenues to exceed $200 billion annually by 2026. Suppose Nvidia's GPUs represent about half of a data center's costs, leading to over $400 billion in AI data center capex. If AI software companies aim for at least 60% profits, they would need $1 trillion in AI-related revenue to cover the AI infrastructure costs. For context, the entire S&P 500 software industry generated $500 billion in annual revenue in 2022, before AI became a major trend. In the near term, there is a disconnect because companies are spending significantly on AI infrastructure, but the revenues from AI applications are still low. However, we believe this gap will close as enterprise AI adoption and monetization becomes more meaningful in 2026 and beyond.

Given the performance of the Mag 7 and narrowness of the market over the past two years, and our underexposure to the top 10 names in the index, our strategy has underperformed. We believe in the AI opportunity but initially struggled to invest where free cash flow and ROI visibility were low. We do not own Nvidia Corporation (NVDA) due to its high valuation, or Meta Platforms, Inc. (META), given its relentless spending on Reality Labs, or Tesla, Inc. (TSLA), where the valuation is difficult to comprehend, i.e., its market value is larger than the top 20 global auto OEMs combined. To be clear, we are not AI agnostic! We own stocks with AI exposure such as Marvel Technologies, Inc. (MRVL), which designs AI application-specific semiconductors; Cadence Design Systems, Inc. (CDNS), which provides software needed to design AI semiconductors; and ServiceNow, Inc. (NOW) and Palo Alto Networks, Inc. (PANW), which provide enterprises with AI software applications. As with any new technology, there are risks, but we believe AI will improve the rate of productivity going forward as AI is already transforming certain business functions including recommendation engines, customer services, developer productivity, cyber security applications, and healthcare (i.e., drug discovery), to name just a few.

The Russell 3000 is currently trading at 22x P/E, the Mag 7 at 32x P/E, and the Russell 3000 ex-Mag 7 at 19x P/E (Exhibit 5). The Mag 7 stocks should trade at a premium, given their above average earnings growth and their historical ability to generate free cash flow, but a significant portion of the AI enthusiasm appears priced in. Earnings growth for the market overall remains robust and earnings are expected to broaden out, with positive earnings growth expected for every sector. However, the elevated concentration level of the Mag 7 remains a concern, given the recent performance, higher valuation levels, and the possibility that overall client portfolios may have more exposure to these names than they realize. While we are mindful of the economic backdrop, we remain intently focused on stock selection and feel confident in our current positions. Our experience across various market conditions and cycles has informed the belief that high-quality companies, operating models with deep competitive advantages, under the guidance of ROIC-oriented leadership teams, and at the appropriate valuations, are those best positioned to provide attractive riskadjusted returns over the long run.

Exhibit 5: Russell 3000 forward P/E