Artificial intelligence has been around since before 2000, but became a central investment theme in the spring of 2023. ChatGPT was released late in 2022, but a pivotal moment in the birth of the AI investment theme sprang from the Nvidia earnings call in May 2023, where the company reported a huge sales increase for its microprocessor chips used to perform the massive computations needed to service the large language models behind AI. The news led to a rise in the stock price of more than 24% the next day and prompted a years-long rally that continues today.
That Nvidia earnings call contained a quote from CEO Jensen Huang that displays a remarkable level of forward-thinking insight…“A trillion dollars of installed global data center infrastructure will transition from general purpose to accelerated computing as companies race to apply generative AI into every product, service, and business process”. At the time, it seemed as though Huang’s goal was lofty hyperbole. As we reflect on that today, he may have missed an “s” at the end of trillion.
We estimate that roughly 45% of the US stock market value is tied to companies that operate in the AI space, conducting business in six major categories:
1. Large hyperscalers:
- Representing 21% of the market
- Companies that build and own the data centers, such as Microsoft, Google, Tesla, and Meta, among others.
2. Semiconductors
- Representing 15% of the market
- Chip designers and chip equipment manufacturers like Nvidia, Broadcom, and AMD.
3. Software
- Representing 4% of the market
- Application creators focused on writing code to automate tasks and for security from companies like Oracle, Palantir,
and Salesforce.
4. Non-technology infrastructure construction
- Representing 2% of the market
- Contractors, integrators, and material suppliers such as Eaton, Emerson Electric, and Johnson Controls.
5. Non-semiconductor technology hardware
- Representing 2% of the market
- Companies that connect and configure servers like Cisco Systems, IBM, and Amphenol.
6. Energy providers
- Representing 1% of the market
- Organizations that provide the electricity production supply chain. [GE Vernova, Constellation Energy, Williams Co]
As fundamental investors, we derive an investment thesis from the success of a company’s financial statements. This helps us reduce our exposure to short-term fads. We follow the numbers (and cash) to see how revenues, margin structure, and capital expenditures have changed. If there are no material changes, the news may be too optimistic. We have integrated this focus on tangible results into our AI decision-making process.
While the hyperscalers, semiconductor makers, and software designers garner most of the AI-related news, we have been more bullish on the infrastructure construction, technology hardware, and utilities segments. These companies make up a relatively small part of the market but have been some of the best investments over the past three years as money flowed towards creating data center capacity. Hundreds of billions of dollars have been spent on infrastructure, including pouring cement, installing cooling systems, and connecting servers, before a data center comes to life. Companies that operate in this hardware installation have reaped the rewards, as evidenced by their financial statements.
We often invoke the 1849 California gold rush to explain our outlook. As hordes of people moved west for opportunity and dig for gold, more found failure and hardship compared to the few who struck pay dirt. However, some industries performed well regardless of the miner’s success. Providers of pick axes, shovels, and other mining supplies satisfied a persistent demand for equipment as thousands of risk takers invested in the hardware needed to realize their goals.
AI promises to take automation to another level and has demonstrated striking progress so far. Still, the difference between getting a question answered on ChatGPT on where to eat dinner that night is widely different than having a large language model determine the best ingredient combinations for a successful cancer drug. Proponents claim that AI will achieve PhD-level intelligence, and hallucinations and errors in responses will dwindle. We do not know if those goals will be realized, and we certainly don’t know if they will be realized “on schedule” and without hiccups. We do know that the only way to learn these goals is to first spend billions of dollars on further expanding the underlying infrastructure.
The stock market is fundamentally a forward-looking entity, and trading occurs based on incrementally available data. Even these “Picks and Shovels” companies must serve investors who expect the high demand for data centers to continue. But what happens if more and more gold miners can’t find gold and become discouraged, or when data center capacity finally matches demand? The suppliers of equipment find themselves without a market. Right now, prevailing sentiment is that we are woefully below the necessary supply of data centers; there are not enough chips to process all of the data. Eventually, the supply of chips will catch up to demand, and each data center owner will have the necessary equipment.
Unlike gold, whose supply is limited, the creation of data center capacity is growing. Facilities take years to construct, making the investment horizon less volatile. We have participated significantly in this buildout, closely monitoring the fundamentals of our holdings as new projects are announced and previous ones reach completion. Within the past three months, we have trimmed several names back due to their strong stock performance and approach of relatively high valuation levels. We expect the expanding universe of AI applications will continue to provide profitable investment opportunities for the foreseeable future.
Disclosure: This is for informational purposes only and any reference to a specific company or industry does not constitute a recommendation to buy or sell that company or industry. The reader should not assume that an investment in the security or industry identified or described, was or will, be profitable. For a complete list of disclosures, please click https://mitchcap.com/disclosure/
