- The AI market only continues to grow, and so does demand for the data that powers it.
- Data centers are perfectly positioned to take advantage of this trend, according to BofA.
- David Barden breaks down the opportunity for data center REITs, and which companies will profit.
Software. Cybersecurity. Fintech. Semiconductors.
With AI permeating every aspect of our lives, there are a million and one ways to invest in AI.
However, there is another way to bet on the growth of AI that combines technology with the world’s oldest asset class: real estate. Data center REITs, to be specific.
This investment thesis is supported by a simple formula: “As data continues to grow at an exponential rate, more data centers are required to store that data,” wrote Bank of America equity strategist Haim Israel in a note to clients earlier this year.
If, as Bank of America predicts, AI has its “iPhone moment” and grows to a $900 billion market by 2026, there will be a greater need for data centers than ever before — and that means higher profits for a few well-positioned companies.
Two approaches to AI growth
Bank of America senior research analyst David Barden delves into the state of AI and its meteoric rise to the top of investors’ minds in a note to clients dated September 7.
“NVDA’s 1Q23 earnings print sparked an AI frenzy among companies and investors alike, who are looking for potential beneficiaries of recent, rapid advancements that are making Generative AI more powerful and affordable,” Barden wrote.
AI has remained in the spotlight since then. According to BofA’s analysis of earnings call transcripts, mentions of AI in S&P 500 companies have increased 95.6% quarter over quarter in 2023.
Although artificial intelligence (AI) is on everyone’s mind, the technology is still in its infancy. Researchers and coders are still figuring out how to make it work, but Barden has divided the process of developing a generative AI program into two main components: training and inference.
“A hugely compute-intensive exercise where a model learns from a large data set, asking and answering questions in an effort to ‘learn’ (the intelligence),” Barden wrote.
Inference is the outcome of that training, as an AI model answers questions about the data set on which it was trained.
Training and inference are two sides of the same coin, and both necessitate large amounts of data, a location to store that data, and the computing power to process it all.
This has increased global demand for data centers. Barden cited markets around the world where data center vacancies were at all-time lows, while the cost of new leases in those centers was rising.
Because of the simultaneous increase in demand and decrease in availability, data centers now have more pricing power.
“AI will only exacerbate the dynamic, as the explosion of interest in AI creates even more demand for data center space,” wrote Barden.
Two data center behemoths are poised for takeoff.
With data centers poised to benefit from the AI boom, the only question is which data center providers are making the best investments right now.
Barden believes that the two largest data center REITs on the market will benefit significantly from increased AI demand in the coming years, and that each will focus on a different aspect of the AI model creation process.
First, he believes that Digital Realty Trust (DLR) will benefit from the need to train AI models due to its “larger-scale wholesale deployments,” according to Barden.
“DLR’s business model should, however, position it to develop facilities with the power density required to host large AI model Training,” wrote Barden. “We anticipate increased demand for such capability in the coming years.”
Second, according to Barden, Equinix (EQIX) has the infrastructure in place to handle increased demand for AI capabilities without having to invest heavily in the near future. Moreover, unlike Digital Realty Trust, Equinix is better positioned to profit from businesses that use AI models for inference rather than training.
“In the short term,” Barden wrote, “we believe EQIX’s retail-centric business allows it to live closer to its customers and positions it to better serve Inference opportunities.” “EQIX has many smaller enterprise customers that will look to implement AI Inference-related functionality as soon as possible.”
While Equinix can capitalize on AI sooner rather than later, the larger capacity of Digital Realty Trust makes it ideal for “larger, longer-dated AI training model environments,” according to Barden.
However, while each company has a unique set of skills that allows them to profit from different segments of the AI market, Barden believes that both have a huge opportunity ahead of them as demand for AI continues to rise.