AI infrastructure is no longer just a story about Nvidia chips and Big Tech capex. Pimco’s private placement push shows you that the financing stack is becoming part of the advantage.
If you're still watching the AI buildout only through stock prices, you're missing where some of the real pressure is now showing up. The Financial Times reported on June 28 that Pimco, the Newport Beach bond manager with about $2.3 trillion in assets, has been taking large positions in private placement deals that sit awkwardly between public bonds and private credit, including financings tied to Meta and Oracle data center projects.
That is the useful part of the story. AI has become so capital-hungry that the old line between public debt and direct lending is starting to look less important than speed, size and certainty of execution. A company building a multi-billion-dollar data center does not only need GPUs, power and land. It needs someone willing to write a very large check before a normal bank-led process has neatly found every buyer.
According to the FT, Pimco has recently helped finance wartime debt for Middle Eastern governments, lent to a Blue Owl private credit fund after its shares sold off, and contributed a major share of funding for large data center projects for Meta and Oracle. The firm is using its scale to negotiate directly with borrowers and anchor transactions, then in some cases use the 144A market to distribute pieces later. That is not a small procedural change. It moves power toward the institutions that can commit capital first.
Meta’s Hyperion project in Louisiana is the cleanest example. MarketWatch reported in October that Meta and Blue Owl formed a $27 billion joint venture for the Hyperion data center, with Blue Owl funds owning 80% and Meta holding 20%. Morgan Stanley advised Meta, and Blue Owl funded part of its investment through debt sold to Pimco and other bond investors in a private securities offering.
Barron’s, citing Wall Street Journal reporting, said Pimco was the biggest buyer in that financing at about $18 billion, while BlackRock bought more than $3 billion of the bonds. Meta contributed land and construction assets, received a $3 billion distribution at closing, and structured the project so the debt sat at the joint venture rather than directly on Meta’s balance sheet. You can call that financial engineering if you like. Frankly, it is also a sign that AI infrastructure is now large enough to reshape the plumbing of credit markets.
The physical project is not abstract either. Hyperion is being built in Richland Parish, Louisiana, and is expected to require up to five gigawatts of power over time, with an initial two-gigawatt target by 2030, according to Barron’s. When one data center campus starts looking like a utility-scale financing problem, equity investors are no longer the only audience that matters.
Oracle shows the other side of the same trade. Business Insider reported in February that major banks had struggled to sell pieces of $38 billion of debt connected to two data center campuses anchored by Oracle, as lenders worried about the company’s AI spending and credit rating pressure. Oracle later said it would raise up to $50 billion from stock and bond offerings to maintain an investment-grade balance sheet, according to that report. The demand for AI capacity is real, but the credit market is already sorting winners from borrowers who have to pay up or explain themselves.
Cheap capital is becoming infrastructure
This is why founders and investors should care, even if they never buy a bond. In AI infrastructure, access to capital is no longer just a finance-office detail. It decides who can sign long leases, secure power, order chips and survive the time lag between construction spending and customer revenue. If Pimco, Apollo, KKR, Blackstone or Blue Owl can anchor a deal quickly, the borrower gets a different kind of advantage from a startup waiting for bank syndication or another equity round.
The risk is just as concrete. Business Insider reported that Blue Owl was unable to arrange third-party financing for a $4 billion data center project in Lancaster, Pennsylvania, tied to CoreWeave, whose credit rating was B+ at S&P Global. Blue Owl said the project was fully funded, on time and on budget, but the market reaction was telling. Investors are not blindly buying every AI data center story that reaches their inbox.
The private credit backdrop makes this sharper. The FT put the private credit industry at more than $3 trillion, and large asset managers are increasingly using their mandates to lead deals that once would have moved through more conventional public debt channels. At the same time, recent redemption pressure at some private credit funds has reminded everyone that private money is not magic money. It still wants yield, collateral, covenants and a believable tenant.
For AI companies, the next phase will reward more than model quality. You need power contracts, credible customers, balance-sheet room and investors who believe the site will still matter when the first leases start paying. Meta can lean on its cash flow and advertising machine. Oracle can go to public markets and defend its credit profile. CoreWeave has to keep proving that growth financed with expensive capital can turn into durable cash flow.
Pimco’s move does not mean banks are irrelevant. It means the largest bond managers have realized that AI infrastructure needs capital in sizes and structures the normal channels cannot always deliver cleanly. If you are building around AI, watch the financing as closely as the chips. The companies that get cheaper, faster capital will not just build first. They may be the only ones that can afford to finish.
Also read: The BIS says the AI boom has become a financial stability risk • The Iran conflict turned crypto into a geopolitical risk barometer and the market hasn't fully recovered • Hong Kong's AI-fueled IPO boom is rewriting where Chinese tech capital goes to grow
