The artificial intelligence boom has sent Wall Street bankers, private equity giants, and debt financiers into a frenzy as they scramble to get a piece of the $1 trillion—and possibly $2 trillion—needed to power AI’s takeover of the world.
Data centers, power grids, and communication networks are the battlegrounds, and nobody wants to sit this one out. Morgan Stanley reportedly hosted a high-profile dinner this weekend in New York to strategize.
The guest list read like a who’s who of finance: Apollo Global, Blackstone, KKR, Ares Management, and Oaktree Capital. But it wasn’t a war room, it was a call for unity. The message was simple — “AI is big enough for everyone, so let’s stop fighting and start collaborating.”
The data center gold rush
Data centers are the backbone of AI. They’re massive, power-hungry, and ridiculously expensive. Building just one of these “AI factories,” as Nvidia CEO Jensen Huang likes to call them, can cost $12 billion. And Wall Street can’t get enough.
Deutsche Bank has worked on $17 billion worth of data-center financing over three years. JPMorgan has created a dedicated infrastructure team to manage its AI workload. Even then, the demand is overwhelming. Bloomberg says one banker admitted his firm is juggling so many data-center deals that they can’t hire fast enough.
Private equity is diving in headfirst. Blackstone shelled out $10 billion to acquire QTS Realty Trust, which owns dozens of U.S. data centers. It teamed up with the Canada Pension Plan to buy AirTrunk, an Asian data-center network, for $15.5 billion.
And they’re just getting started. Developers are eyeing projects worth tens of billions globally, from the U.S. to northern England, where a former battery factory site is being repurposed for a new hyperscale center.
Even BlackRock’s Larry Fink is in on the action. He’s planning to raise $120 billion in debt to build data centers, partnering with Microsoft to fund the infrastructure AI desperately needs. Deals like these have private equity salivating.
Debt markets on fire
The AI boom is also shaking up debt markets. Hedge funds are experimenting with exotic debt structures, and sales of securities backed by data centers have hit $7.1 billion this year, nearing record highs. Fiber networks and other tech infrastructure are pushing those numbers even higher.
Private lenders and banks are throwing cash at anything AI-related, and it’s opening up corners of finance that were once collecting dust.
Utilities and telecommunications, long considered boring investments, are now some of the hottest markets in credit. Companies building fiber networks, satellite systems, and telecom towers are raking in funding.
Banks are willing to lend as much as 80% of a project’s cost if a major tech tenant, like Amazon or Google, signs a long-term lease. Developers love it because it minimizes risk. But lenders are more cautious with projects focused solely on AI.
Speculative development is another headache. Some firms are building data centers without signed tenants, a decision critics are calling “zombie projects.” Lawyers warn these risky bets could lead to unfinished facilities and wasted capital if demand doesn’t materialize.
AI’s appetite for power
AI isn’t just expensive though — it’s energy-hungry. Data centers devour electricity, putting massive pressure on power grids worldwide. In Ireland, the national grid operator warned last year that some large data centers might leave the country because they can’t secure enough energy.
Across the Atlantic, Virginia—a global hub for data centers—set six records for peak power demand in July alone. This surge in demand has utilities racing to expand. Capital spending in the sector is set to hit $200 billion next year, double what it was a decade ago.
Moody’s has loosened its rules for hybrid bonds, allowing utilities to raise cash without risking credit downgrades. Bond sales tied to utilities have soared to $15 billion this year, an eightfold increase.
Nuclear power is making a comeback, thanks to AI. Companies like Carlyle are eyeing nuclear energy as a way to power data centers. Some companies are even considering small modular reactors, a radical and costly solution to AI’s insatiable appetite for energy.
Hyperscalers drive the spending spree
Tech giants (Amazon, Microsoft, Google, Meta, and Apple) are known as “hyperscalers” for their massive spending. These guys dumped $52.9 billion into AI infrastructure in just three months. And they’re not done yet.
U.S. colocation centers, which rent out server space to tech companies, are seeing explosive growth. Construction has increased sevenfold in two years, and rents in these facilities jumped 37% in just 12 months. Real-estate brokers like Jones Lang LaSalle say the demand is “insatiable.”
For developers, the returns can be astronomical. London data centers boast profit margins of 65%, according to Green Street. Banks are eager to fund these projects because long-term leases with blue-chip tenants make them safe bets.
But not everyone is convinced this spending spree will pay off. Some experts believe without a “killer app” like e-commerce or GPS from the Web 2.0 era, some investors worry AI could fizzle out before it delivers on its promises.
And then there’s the issue of scale. Sam Altman, CEO of OpenAI, has floated the idea of five-gigawatt data centers—massive facilities that would require millions of square feet and consume enough electricity to power entire cities. These projects are eye-wateringly expensive, with little guarantee of success.
Still, Wall Street is betting big. The allure of AI seems too strong to resist, even if the risks are high.
