Tim Hoyle, Chief Investment Officer
thoyle@haverfordquality.com

AI: Bubble and Opportunity

The S&P 500 hit an all-time high six days into a government shutdown. It seems nothing will slow this market down as long as three critical ingredients are in place: 1) easing monetary policy, 2) strong earnings growth, and 3) AI optimism. At least for now, weaker employment data, geo-politics, and weak consumer sentiment just doesn't matter.   

The meteoric growth of AI-related investments leaves us all wondering if the bubble is about to burst, or if there is even a bubble at all.  There is evidence to support both sides of the argument. It is very possible that a speculative fervor (a.k.a. bubble) exists while at the same time the AI theme can still offer strong returns for investors with a long-term horizon. Both can be true. Two recent Barron's headlines tell the story:  a September 30 article was titled "Fears About an AI Bubble Are Fading. That's an Opportunity," while just six days later an article carried the headline, "How Much Longer Can the AI Bubble Last? Inside Wall Street's Great Debate."   

From the beginning of 2024 through September 2025, the S&P 500 is up 43%.  Approximately 72% of the market's return is attributable to the AI theme, with the remaining performance being split between the re-rating in financial stocks and all other stocks.   

While we believe AI represents a transformative long-term opportunity, we also acknowledge bubble-like dynamics emerging.  Our portfolios seek to maintain a healthy, albeit below market exposure to AI leaders, balancing participation in a structural growth opportunity with prudent diversification. Short-term volatility is inherent in markets, especially during periods of rapid innovation and investment. We expect a bear market in these stocks will eventually come, but it might come from much higher levels than today's prices.  

We have mapped our portfolios' exposure to different segments of the market.  We believe Haverford's mix of exposures relative to AI-adjacent investments is more defensive than the S&P 500 while still providing exposure to growth opportunities. Within our AI exposure we are also focused on managing risk through position sizing and diversified industry exposures.  As always, our priority is focused on disciplined risk control while creating lasting value for our clients. 

Prices in markets, both rational and irrational, are a balancing act of both the positive and pessimistic views. Below is our take on both sides of the AI bubble debate.   

Positive View Driving AI Investment Today 

Prices are not that far ahead of fundamentals. The mega-cap tech companies are growing revenues by double-digit rates and trading at earnings multiples of 25-35x. This is not extravagant compared to sectors growing significantly slower.  Mega-cap tech has the cash flow to support capex; the real bubble may be in zero-revenue tech start-ups.   

AI will become a general-purpose technology like electricity and the internet and will eventually drive productivity in life-changing sectors like Healthcare, Finance, and Energy. AI will eventually be able to optimize electricity grids to prevent outages, assist doctors in diagnosing and treating illness, and drive risk-management and fraud prevention in banking.[1] 

Today's AI will be the worst model you use for the rest of your life. AI will only get better, and the rate of improvement has been staggering. The technology is moving so fast it is impossible to predict what new businesses, services, and hopefully time-saving tools it will produce. The internet and mobile computing spawned entire business models and jobs. Uber is an example of a business that was unthinkable before the iPhone, but Uber drivers may soon become a casualty of AI and self-driving cars. 

Pessimistic View of AI at Today's Prices   

The parabolic AI infrastructure spending is far ahead of monetization opportunities. $500 billion of estimated 2025 capital expenditures will need to generate over $1 trillion in incremental annual revenues just to break even.  We have seen no evidence of significant end-user revenue yet being generated. Excess infrastructure investment is nothing new. It occurred in fiber optics in the 1990s, railroads in the 1870s, and the electric grid in the 1920s.[2]   

There may be physical constraints to today's planned data center buildouts. The International Energy Agency (IEA) estimates that electricity demand from data centers worldwide is set to more than double by 2030. Driven by AI use, the US economy is set to consume more electricity in 2030 for processing data than for manufacturing all energy-intensive goods combined, including aluminum, steel, cement and chemicals.[3]  Data center growth may slow down if consumers see energy prices rise more than the perceived value of their favorite chatbot.   

While an argument can be made that Mega cap tech is not that expensive, the bubble is more evident in private-markets. The tech and telecom bubble of the late 90s saw an abundance of IPOs at extreme valuation and sentiment. Today, the IPO market is almost non-existent and private markets are rich with capital. According to the private market investment platform Forge Global, "a basket of seven of the highest-valued private tech companies is now worth $1.3 trillion on paper, almost doubling in the past year."[4]  The "fear of missing out" looks to be driving a bidding war for these assets. 

[1] Harnessing AI for economic growth | Brookings

[2] The AI Boom Just Hit Overdrive

[3] AI is set to drive surging electricity demand from data centres while offering the potential to transform how the energy sector works – News – IEA

[4] OpenAI leads private market surge as 7 startups reach $1.3 trillion