46% of the S&P 500 is One AI Bet | Kai Wu on Why It’s Likely the Wrong One

Key Concepts

• AI ROI is Critical: The central question is whether AI investments are translating into demonstrable economic returns beyond infrastructure spending.
• Diffusion of Innovation: Broad economic benefits are necessary to avoid an AI bubble; benefits must extend beyond the tech sector.
• Historical Parallels: Past infrastructure booms (railroads, fiber optics) suggest value accrues to users of the technology, not necessarily builders.
• AI Yield as a Metric: A new metric, “AI Yield,” focuses on companies efficiently investing in AI relative to capital deployed.
• Early Adopter Advantage: Companies actively integrating AI are positioned to benefit, though this isn’t yet fully reflected in market valuations.
• Infrastructure vs. Application Layer: Profits are more likely to be found in the application layer built on AI infrastructure, rather than the infrastructure itself.

AI Adoption & Investment: A Deep Dive

The discussion centers on analyzing the current state of AI investment, assessing whether observed demand translates to sustainable economic returns, and identifying potential investment strategies. This analysis is framed around the “trillion-dollar question” posed by Satya Nadella: for AI not to be a bubble, its benefits must be widely distributed.

Current State of AI Investment & ROI

Currently, 10% of US businesses are utilizing AI in production. Analysis of thousands of company earnings calls, categorized by mentions of AI, economic gains (revenue, cost savings, productivity), and quantifiable ROI, reveals a disconnect. While companies mentioning AI-driven economic gains experience a 4.8% annualized outperformance, and those mentioning specific AI ROI see 5.2% annualized outperformance, there is no correlation between AI investment (excluding infrastructure) and market valuation. Early adopters and laggards trade at parity despite differing AI utilization levels. 80% of companies are classified as AI “laggards,” while 10% are “early adopters.”

Historical Context & Potential Pitfalls

The analysis draws parallels to previous infrastructure booms, specifically the railroad and fiber optic cable expansions. These historical examples demonstrate a pattern where infrastructure builders often don’t capture the majority of the value created. Instead, the value accrues to those using the infrastructure. The telecom bust, where the telecom index fell 95% and hasn’t recovered, is a stark example. Fiber optic cable prices fell 90%, with 85% remaining unused, illustrating a pattern of oversupply and limited profitability for infrastructure providers. This suggests a potential scenario where AI infrastructure providers may not translate investment into sustained profits, becoming a utility rather than a high-growth sector.

Is AI Different? The Intelligence Factor

The question of whether AI is fundamentally different from previous technologies is addressed. While AI’s “intelligence” component is a potential differentiator, the prevailing view is that AI will likely follow the pattern of other technologies – a powerful tool integrated into the existing economic structure, displacing some jobs while creating others. Transformative technologies like electricity and the internet also underwent similar integration processes.

The Concept of “AI Yield” & Investment Strategy

A core concept introduced is “AI Yield,” analogous to dividend yield. This metric focuses on identifying companies investing heavily in AI (employees, patents, trademarks) relative to the capital invested. The firm’s investment strategy prioritizes companies with high AI yield, seeking to capture the benefits of AI without the valuation risk of overpaying for established infrastructure companies. This is a “bottoms-up” approach, driven by intangible value assessment rather than a top-down thematic bet. Traditional value metrics (price-to-book) are deemed ineffective in intangible-intensive sectors like AI.

Market Exposure & Index Analysis

The S&P 500 is heavily weighted towards AI infrastructure, with 46% combined with the MAG7 (Apple, Microsoft, Alphabet, Amazon, Nvidia, Tesla, and Meta). Alternative indices (Russell 1000 Value, Equal Weight, MSCI EAFE) offer diversification but risk missing out on AI gains or being overly exposed to “lagard” companies. The firm’s ETFs are positioned with 56-59% exposure to AI early adopters.

International AI Landscape

While AI innovation is largely concentrated in the US and China, opportunities exist in international markets. Non-US companies are more likely to be positioned as users of AI rather than infrastructure providers, with sector distribution differing towards healthcare, consumer discretionary, and industrial companies.

Monitoring Future Performance

Key indicators for assessing the success of AI investments include continued reporting of ROI from AI by CEOs, technological advancements (e.g., Cloud Code), a shift from pilot programs to material financial impact, and migration of companies from "lagard" to "early adopter" status.

Conclusion

The analysis suggests that while AI investment is booming, demonstrable economic returns are still nascent. The historical parallels to previous infrastructure booms highlight the risk of overinvestment in infrastructure and the potential for value to accrue to those using the technology. A focus on “AI Yield” – identifying companies efficiently investing in AI – and prioritizing early adopters appears to be a prudent investment strategy. Continuous monitoring of ROI reporting, technological advancements, and broader AI adoption will be crucial for navigating this evolving landscape and determining whether AI will deliver on its transformative potential or succumb to the fate of past bubbles.