Wall Street Wants to Own AI Data Centers. Who Gets Paid Before OpenAI Makes a Profit?

Key takeaways

• Wall Street funds AI infrastructure through long-term leases
• Blackstone and Blue Owl target predictable rental income
• Data center owners face growing tenant and financing risks

OpenAI may be building the future of artificial intelligence, but some of the companies positioned to profit from its expansion are not developing AI models at all. They are buying land, financing buildings and collecting rent from the enormous facilities needed to power them.

Blackstone, Blue Owl Capital and Ares Management are increasingly treating AI data centers like traditional infrastructure investments, according to an October 29 Reuters report. Their strategy depends on long-term contracts with major technology companies rather than directly betting on whether individual AI products become profitable.

The opportunity is enormous. Ares estimates that AI-related capital expenditure could exceed $5 trillion through 2030, including roughly $2 trillion for physical data centers and adjacent power infrastructure.

How Wall Street Turns AI Demand Into Rental Income

Blue Owl says it owns or is developing more than 130 data centers, representing nearly 14 gigawatts of capacity, with another 21 gigawatts in its pipeline.

The firm's typical leases last 15–20 years and include annual rent increases. These agreements can give investors relatively predictable contractual income even while the broader AI industry struggles to establish sustainable profits.

Ares follows a similar model, developing facilities that are pre-leased to investment-grade technology companies under contracts lasting 12–15 years.

Blackstone says 96% of its data-center counterparties carry investment-grade credit ratings, while most leases in its portfolio initially run for 15–20 years.

The arrangement helps explain why institutions are increasingly interested in AI infrastructure financing rather than simply buying shares of AI software companies.

Why Investors Can Earn Money Before OpenAI Turns Profitable

Consider a hypothetical $10 billion data center financed by investors and banks, then leased to a major cloud provider.

The building owner can collect rent once contractual payment obligations begin, even if the AI developer ultimately using the computing capacity continues reporting losses.

Similarly, banks can receive scheduled interest payments, while equipment suppliers earn revenue when chips and networking hardware are delivered.

However, contractual income is not the same as profit. Owners must still cover financing costs, maintenance and other expenses, while lenders depend on borrowers meeting their obligations.

The distinction becomes particularly important with OpenAI, whose major computing agreements involve infrastructure providers such as Oracle and CoreWeave.

These companies may commit billions to facilities before collecting all the revenue associated with their AI contracts.

What Happens if the AI Boom Slows?

Long-term contracts reduce uncertainty, but they do not eliminate financial risk.

In September, Oracle issued a force majeure notice connected to Project Jupiter, a New Mexico data center development backed by Blue Owl. Construction delays threatened to postpone higher returns expected after completion.

More recently, concerns about CoreWeave's approximately $50 billion in broader debt exposure sent some of its bond yields toward 15%–16%, demonstrating how quickly financing conditions can deteriorate when investors question repayment prospects.

Those concerns have contributed to the broader AI debt debate, particularly as increasingly large projects depend on private-credit funds and institutional lenders.

There is also a difference between renting to Microsoft or Amazon and relying on a heavily indebted cloud provider with concentrated customers.

If a tenant cannot pay, rental contracts alone cannot prevent losses. Owners may need to find replacement customers, restructure financing or accept lower asset valuations.

This is why the identity of the tenant matters as much as the building itself, and why investors have become increasingly concerned about CoreWeave's borrowing costs.