AI’s Stock Market Impact: What You Need to Know

AI Bubble? No. Volatile? Yes.

Artificial intelligence (AI) has arguably been the driving force behind financial markets since the public unveiling of OpenAI’s ChatGPT generative AI (GenAI) model in late 2022. Our clients have asked if the major surge in AI stocks has already occurred. We have also heard suggestions that AI stocks are experiencing a “bubble” like the dot-com market of 1999-2000.

We can’t know what the stock market will do tomorrow or next year. But we can speak with conviction about company financials in this space. Based on our understanding of the evolving business opportunity, we believe we’re in the early stages of an AI computing revolution. Despite the recent solid gains in AI-related stocks, we think tremendous financial and economic benefits have yet to be realized from the technology.

However, we think these equity returns will likely be uneven and volatile. The sell-offs in AI-related stocks in August and October 2024 highlight this volatility. Additionally, the economic benefits of AI may not be evenly distributed across the economy. This article will discuss the potential hurdles that could impact AI's growth.

AI Market Growth: The Trillion-Dollar Question

The financial press often mentions that the AI market could grow to $1 trillion by 2027. For example, a June 2024 AI point/counterpoint from Goldman Sachs argues that we could expect $1 trillion in corporate spending on AI. However, it also claims that the return on this investment is highly uncertain.¹ Management consultant Bain & Company’s 2024 Technology Report features a chapter titled “AI’s Trillion Dollar Opportunity.”²

Similarly, NVIDIA CEO Jensen Huang recently mentioned this figure when talking about the data center and hardware replacement cycle.³ Because computing power has doubled every two years since the 1960s, a principle known as “Moore’s Law,” there’s a natural cycle for tech upgrades. This creates a consistent demand for chips and technology to update data centers. Servers are typically refreshed every three to five years.

Meanwhile, technology consulting firm Gartner says demand for new data centers to accommodate AI workloads is experiencing explosive growth. Gartner estimates that global spending on data centers will rise by almost 25% to more than $290 million in 2024.

We believe companies’ existing capital spending commitments suggest that the trillion-dollar number is real and achievable. As a result, we’re finding attractive opportunities in data center and semiconductor companies as well as firms that make equipment for manufacturing microchips and integrated circuits.

Expanding our view, a recent IDC research report suggests that AI will have a total economic impact of nearly $20 trillion through 2030.⁴ No matter how you measure AI’s potential, we’re confident that we’re just at the beginning of a technology revolution.

The Current State of AI

Media headlines often present a strange contrast: Some declare the imminent arrival of AI dominance, while others express skepticism about the AI revolution. Critics argue that AI hasn’t yet significantly transformed business and our lives and question whether capital spending is justified.

We believe the truth lies somewhere in the middle. Generative AI (GenAI) technology is potentially transformative, but we’re in its early stages, like the “dial-up” days of the internet. There’s a lot of potential, but we haven’t seen its full capabilities yet. Since it’s so new, the success of different companies will likely change over time. This uncertainty means that stock prices could be quite volatile.

Some of this volatility is natural and even healthy. The run-up in prices of AI chipmakers in the last few years has been met with occasional short, sharp declines. We interpret this volatility as a natural part of the technology's development cycle. Such periods of adjustment reflect uncertainty around the payoff of AI investments. They also help correct overvaluations and set a more sustainable growth trajectory for the future.

We want to identify potential sources of volatility so that you can better understand the legitimate risks to AI development. Given the early stage of AI buildout and adoption, we think these risks virtually assure continued stock volatility.

Six Potential Challenges for AI Growth

In our view, the future of AI looks promising, but we see six potential hurdles that could impede growth.

1. Scaling Challenges May Limit GenAI Gains

“Scaling laws” aim to quantify the relationship between the size of AI models, the size of datasets and the computing power needed for GenAI large language models. Essentially, the larger the model and dataset, the more accurate the model becomes. In industry parlance, bigger models and datasets result in a lower “loss rate,” meaning fewer errors. As models and datasets grow, they also require more computing power.

However, increasing the size of AI models and datasets yields diminishing returns. Doubling the model size or amount of data used cuts errors by less than half. This predictable pattern is a scaling law, but unlike the law of gravity, it’s not a fundamental law. It’s an observed relationship similar to Moore’s Law and can change over time.

This means that even with significant investment in AI infrastructure, improvements in GenAI capabilities may not be proportional to the resources spent. Significant advancements will likely require the development of more efficient algorithms and hardware.

2. Surging AI Energy Demand

Another sizable challenge is the growing energy demands of large data centers. In fact, we’ve dedicated an entire article to addressing the energy and infrastructure challenges that arise from developing this technology.

As the technology scales, so does its energy consumption. Morgan Stanley estimates that data center power demand will grow from 1% of the current U.S. power load to 3% by 2027. Tripling data center energy demand is remarkable. But even more so when you realize that the U.S. is projected to increase total power generation by just 0.8% through 2028.

This reality raises GenAI’s operational costs and poses environmental concerns. Innovations in energy-efficient computing and sustainable power sources will be crucial to address this issue. The burgeoning demand for low- and non-carbon-emitting power helps explain the late 2024 deals that Alphabet and Amazon made with nuclear power producers.

However, we should be cautious not to exaggerate the energy issue. Since the onset of the personal computer and internet revolutions, many industry observers have predicted that power demand would outstrip supply. In reality, gains in energy efficiency have reduced energy intensity across the economy. Over time, we’ve been able to achieve more economic output per unit of electricity.

Of course, there’s no guarantee we’ll achieve the same efficiency gains in the future. But it’s worth noting that we’ve heard these dire forecasts before and have always been able to meet our energy needs.

3. AI Data Limitations: Better Data Mean Better Outputs

The availability of new data to train AI models is essential for the continued advancement of the technology. We’ve already discussed how data is one of the three legs of the GenAI stool. The more data there is, the higher the data quality and the better the model’s outputs will be. However, acquiring high-quality, diverse data sets can be challenging.

Data privacy regulations like the EU’s General Data Protection Regulation (GDPR) further complicate corporate data collection and usage. The GDPR has the benefit of standardizing and clarifying data privacy laws for EU member nations. Companies must navigate these regulations while seeking innovative ways to gather and utilize data responsibly.

4. Rising Costs Could Limit AI Access

The rising cost of building more powerful GenAI systems presents a serious financial hurdle. From advanced hardware to specialized talent, the expenses associated with AI development are substantial. We can cite the enormous capital expenditures of Alphabet, Amazon and Microsoft as evidence.

Smaller companies may struggle to keep up with the investment requirements, potentially leading to a concentration of AI capabilities among a few large entities. Strategic partnerships, open-source collaborations and government support could help mitigate these financial barriers.

5. Sensible AI Regulation Is Crucial

In our view, the successful creation and adoption of AI technology require a solid regulatory foundation. Consider some of the questions this tech raises:

• Navigating intellectual property issues around the content used to train GenAI models.

• Identifying, discouraging and dealing with disinformation/deepfakes.

• Avoiding unintended biases and discrimination.

There are also national security considerations. In October 2024, the Biden administration finalized rules restricting U.S. companies' AI and related investments in China. Nevertheless, we suspect U.S. policymakers will be reluctant to aggressively regulate the space for fear of disadvantaging U.S. companies relative to AI researchers in other countries.

In addition, the massive demand for electricity impacts power generation and distribution (highly regulated industries) and the environment. These considerations introduce a high degree of uncertainty and complexity to the future development of GenAI.

6. Enterprise-Level Hurdles to AI Implementation

The shortage of data scientists and AI-focused engineers also poses a major challenge for businesses seeking to adopt GenAI technology.  We think this suggests a tremendous opportunity for enterprise software companies, which already enjoy domain-specific knowledge across various essential back-office functions. They have the data, the necessary talent and deep industry expertise to develop solutions that individual companies might find hard to achieve independently.

AI Opportunities Are Highly Differentiated

We believe we’re at the beginning of a potentially radical economic transformation. However, significant challenges remain to implementing this technology.

Given the outsized potential gains for winners and penalties for losers, we conduct careful fundamental research when selecting stocks in this space for our portfolios.

In other words, AI is unlikely to benefit all companies equally. Instead, it’s likely to create vastly different outcomes for competing companies, like the internet did for Amazon and Borders^® or Netflix and Blockbuster.