Adobe Stock
Regardless if you're an AI bull or bear, one thing is for certain: Data center demand isn't going away anytime soon.
Big Tech companies are continuing their data center spending spree, as seen in Q2 announcements of increased AI capital expenditures from Microsoft, Meta, and Amazon.
What exactly are the tech behemoths spending on? Big Tech companies will need to expand their data center capacity and also build their own bespoke data centers to house GPU infrastructure, according to Chris Downie, the CEO of the data center company Flexential. As GPU chips become more powerful, data centers need to be able to handle more power density than ever before, and that requires infrastructure upgrades.
The GPU infrastructure housed within data centers, such as Nvidia's newest Blackwell chips, isn't cheap, either — Downie says they "could be five times the cost of the data center itself."
That's not to mention the energy and cooling requirements for data centers, as GPUs suck up power and dissipate heat.
Even Goldman Sachs' Head of Global Equity Research Jim Covello, a known AI skeptic, doesn't believe data center spending will slow in the foreseeable future. Covello expects aggressive investment in AI capital expenditures such as data centers to continue as Big Tech profitability remains high.
Against this backdrop, Bank of America forecasts that the data center physical infrastructure market will grow to $58 billion in 2026, up from $37 billion in 2023. That's a rate of 16% per year.
For investors looking to ride the data center wave and put their money into the picks and shovels driving the AI revolution, the bank's research analysts shared the following eight stock choices in a note.
There are several components of a data center that the Big Tech companies are shelling out billions for, and many companies stand to benefit.
First, the bank is bullish on traditional data center real estate investment trusts, or REITs, such as Digital Realty Trust (DLR) and Equinix (EQIX). These companies rent out data center space to enterprise clients such as the Magnificent Seven.
But beyond the traditional data center names, companies providing the supporting technology for data centers are also taking off as the Magnificent Seven start building their own data centers.
One crucial component of data centers is networking infrastructure, which allows fast data transmission between servers.
Data centers have reached a newfound scale from AI, according to Downie, and networking needs have spiked as a result.
"In a traditional data center, you measured networking in feet," Downie said. "Whereas in an AI environment, you're measuring it in miles."
As a result, network solutions responsible for large-scale data transfer such as Ethernet are taking off. Bank of America predicts that the total addressable market for Ethernet will grow 73% annually in the next three years, reaching $5.7 billion by 2027. Tal Liani, the head of technology research at Bank of America, recommends Arista Networks (ANET), which is expected to take over half of the AI networking market by 2027, and Cisco (CSCO), which Liani believes has a $3 billion potential pipeline of AI networking demand in coming years.
Cooling solutions for GPUs are another booming part of the data center industry. Newer data centers require increasingly advanced liquid cooling technologies to process their workloads, which is not cheap, according to Downie.
IT hardware research analyst Wamsi Mohan is bullish on Super Micro Computer (SMCI), a provider of server and storage solutions for data centers. Not only has Supermicro established itself as a hardware supplier for cloud infrastructure companies, but it also has a competitive advantage at providing large-scale liquid cooling solutions.
Lastly, building out data center infrastructure will also boost construction companies that provide the equipment necessary to break ground on multi-acre data center campuses. Bank of America's buy-rated construction companies include Caterpillar (CAT), Construction Partners (ROAD), and Aecom (ACM).
