Companies that lease facilities to artificial intelligence hyperscalers like Meta, Amazon, Alphabet, and Google were getting stung Thursday amid a selloff spurred by fears of excessive spending on AI infrastructure.
Equinix and Digital Realty Trust stocks both fell around 4%. By comparison, Meta was down around 7% by late afternoon. Meta actually started the panic after disclosing in a regulatory filing that it had committed to spending more than $355 billion in future "data centers, energy-related infrastructure, colocations, and certain network infrastructure." That's up from the $279 billion it said it was planning to spend back in July.
These real estate investment trusts or REITs get all their revenue from leasing or operating data center spaces. That includes not just the building itself but also the electricity and networking that make the data centers hum. (By comparison, neocloud firms like CoreWeave and Nscale provide the actual computing power.)
But Equinix stock has fared better this year, rising nearly 30% though yesterday's close versus Digital Realty's 14% increase.
The difference is explained in part by their business models: While both boast big-name customers like Nvidia, Microsoft, and Oracle, a larger share of Digital Realty's revenue comes from those hyperscalers who bargain harder for the lowest prices. In contrast, Equinix's more than 10,000 customers tend to be smaller companies such as Olympus, Uber, and Air Canada that pay higher rates.
Despite the latest AI spook, Guggenheim analysts wrote in a Wednesday note that "demand remains quite strong" for Equinix as its more diverse customer base "justifies a premium to peers." The analysts rate the stock a buy with a price target of $1,235.
Insulating themselves from the big swings among hyperscalers could be key to the REITs' success. This past summer, Guggenheim noted that Digital Realty's increase in data center contracts for relatively small amounts of power (1 megawatt or less) -- a growing business for the company -- is key because it is both high margin and "less-exposed to hyperscale lumpiness."
Comments