This Might Be The Most Overlooked AI Investing Play Out There

When it comes to investing themes, artificial intelligence is without question the latest and greatest. AI is poised to disrupt and remake companies across entire sectors – heaping a huge number of opportunities right at your feet. Some of the earliest (and potentially most overlooked) opportunities are likely to come from the data center industry. Let’s take a look…

Why data centers though?

These are more than just a bright spot in the otherwise drab world of commercial real estate. See, the overall commercial real estate market has been dogged lately by the economic slowdown, tighter lending standards from banks, high debt loads in a higher interest-rate environment, and increasing vacancy levels. But not data centers: they’ve done well.

Demand for data storage and processing has led to the explosive growth of these centers over the past 20 years. In more recent years, cloud infrastructure, streaming services, remote working, and 5G networks have helped drive demand from customers across a wide range of industries, including finance, tech, and healthcare companies.

Data center growth, in megawatts, over the past eight years. Source: CBRE.

And even with all that expansion, the capacity build-out so far for the industry is still insufficient to cope with the demand that’s expected to come down the pike with AI. And it’s not just that there hasn’t been enough build-out: AI will require a different computing framework with centers located in specific places with specific functionality.

Basically, the technology has two separate functions and needs that are going to drive further growth in this space:

1) AI “training”. The technology needs gigantic computational firepower to continuously ingest data points and refine its understanding. So it also needs the most advanced graphics processing units (GPUs), which Nvidia makes. Data centers used for AI training purposes suck up huge power resources and need to be based close to renewable energy supplies. These centers can be located in rural areas, where land is cheaper, because they don’t have to be near end users.

2) AI inference. These centers serve up the predictions, solutions, and results that you and I see when we ask questions of OpenAI’s ChatGPT or Google’s Bard. These centers have to produce high-performance answers, and do it quickly, so they’re often located in an urban environment, near end users.

While existing hyper-connected data center networks can be adapted to meet this next-level connectivity, they will need upgrades for the enormous processing capacity that’ll be required.

What kinds of firms will increase spending on AI?

In a survey by S&P Global, 84.6% of respondents said their organization’s AI and machine learning infrastructure spending is set to increase “slightly” to “significantly”. And market research firm International Data Corporation estimates that global spending on AI will increase at a 27% compound annual growth rate through 2026, from $154 billion in 2023 to $300 billion in 2026, with spending focused on AI-centric systems.

It’s not clear exactly how much data centers will benefit from this AI spend, but the outlook sure seems positive. Some believe that we will see a huge shift from in-house IT (i.e. companies building and maintaining AI applications for themselves) to data center outsourcing, which tends to be more efficient, cost-effective, and flexible.

What makes data centers a winner?

Data center companies look set to enjoy top-line (revenue) growth as AI spending increases across industries, and also appear set to see an improvement in their profit margins as they increase AI use within their own operations.

These companies already were using AI to improve their own operational efficiencies, but they’ll get even better at it as AI becomes more advanced.

One example is energy usage. Data centers use huge amounts of energy, and AI can help companies figure out when to switch their energy sources to (or from) solar, for example, and how to use hardware more efficiently to extend its usable life. These kinds of decisions would usually require time-consuming analysis by humans, but AI will be able to make these decisions really quickly.

Incident response is another. When a circuit breaks or something goes wrong, it can take a human some time to run through a response playbook to find and fix the fault. AI is a lot faster at assessing the situation and planning a response.

And there’s also security: AI can actually reduce the amount of physical security that data centers need.

So, what’s the opportunity then?

While there are a few big-name real estate investment trusts (REITs) in this space – Digital Realty Trust and Digital Bridge are a couple of well-known ones – the one that really stands out is Equinix. It’s a pure-play digital infrastructure investment, with a market cap of $67 billion and a balance sheet that looks a lot better than its peers’. And it’s well-positioned to benefit from the AI revolution.

Equinix’s client base is full of big players: 68% of its customers have revenues greater than $1 billion, and another 13% have revenues between $100 million and $1 billion. This exposure may make Equinix’s business model less susceptible to any economic downturn, since heftier companies are typically more resilient in a recession than smaller ones.

It has a diversified global portfolio, with operations in the US, Europe/Middle East/Africa, and the Asia-Pacific. That strong US and global footprint allows Equinix to beat the competition on global contracts, which is a major strategic advantage. First-quarter 2023 results showed 16% revenue growth over the year-earlier period, plus record net bookings, and still-strong activity levels.

Equinix has an aggressive development plan – 50 projects are underway across 37 markets in 25 countries. There’s been some concern lately about a decline in cloud spending, but demand for Equinix’s interconnection business, International Business Exchanges (IBXs), has held strong.

One concern for data center firms is whether they lease or own the land their centers sit on – because of the potential for renewal and pricing risk. For Equinix, of its 220 IBX centers in 63 major metropolitan areas worldwide, total revenue of 55% comes from owned facilities. And the company’s average lease maturity is greater than 20 years, including extensions, so it’s got limited risk on that front. In fact, overall, the company’s got low leverage, or borrowing rates, with 96% of its debt at fixed rates, with an average borrowing rate of 2.13%, leaving it less vulnerable in the face of higher interest rates.

Stock analysts generally like Equinix: they’ve given it 20 “buy” recommendations and six “holds”, but only two “sells”. Their average 12-month target price is $792 and the highest is $886 – and with its recent price of $723, that suggests about 10% upside. So it’s not like the racier AI stocks, but, then again, Equinix is expected to be more of a slow-and-steady beneficiary of the AI transition.

Let’s zero in a bit further. Analysts and fund managers typically use adjusted funds from operations ratio (AFFO) to measure a REIT’s financial performance. The AFFO considers the maintenance costs of the real estate property over its life and subtracts those costs from the outlook. Equinix’s first-quarter results came in at $8.50 per share, about 8% ahead of estimates, showing the strength of the current business. Current Wall Street estimates are for AFFO per share of $31.74 for this year, $34.48 for 2024, and $37.7 for 2025. Assuming the company meets expectations for AFFO growth of about 10%, Equinix’s share price should continue to perform well over the coming years.