Three AI Megadeals Are Breaking New Ground on Wall Street -- WSJ

Dow Jones
18 hours ago

By Matt Wirz

Tech giants need so much money for their artificial-intelligence ambitions that Wall Street is developing new ways to get it for them.

Details of some of the biggest AI infrastructure deals, including those involving Meta, OpenAI and xAI, are coming into focus, revealing lucrative, innovative -- and in some cases risky -- funding schemes.

Exhibit One is the deal fund manager Blue Owl Capital struck with Meta in their joint venture to build a giant data center in Louisiana called Hyperion. Blue Owl is buying private equity in the deal and is receiving a debt-like guarantee from Meta if the partnership falls apart, an extraordinary protection, according to several people involved. Another deal involving OpenAI and Oracle involves a lending syndicate of more than 30 banks, according to deal materials reviewed by The Wall Street Journal.

Tech titans are offering sweeteners in their deals because they need to offload risk as the cost of the AI arms race soars, threatening even the strongest competitors. Meta's market value dropped by around $300 billion in a few days after Chief Executive Mark Zuckerberg warned about higher spending on AI, spooking shareholders already worried about an AI bubble.

Banks and fund managers are writing big checks for now, but many are worried about how the complicated deals being signed today will perform when the AI frenzy calms down. Another concern is that each time tech companies take on lots of new debt, their cost of borrowing rises.

"It's been a very long time since we've had a sustained period of economic weakness and that tends to breed more complexity and complacency," said Dan Ivascyn, chief investment officer at Pimco, which also invested in Hyperion. The debt deals now hitting the market are far larger than in past credit cycles, he said.

Here's a close look at some of the biggest AI megadeals so far:

The Hyperion deal is a Frankenstein financing that combines elements of private-equity, project finance and investment-grade bonds.

Meta needed such financial wizardry because it is already borrowing by the bucketload to build AI, issuing a $30 billion bond in October that roughly doubled its debt load overnight.

Enter Morgan Stanley, with a plan to have someone else borrow the money for Hyperion. Blue Owl invested about $3 billion for an 80% private-equity stake in the data center, while Meta retained 20% for the $1.3 billion it had already spent. The joint venture, named Beignet Investor after the New Orleans pastry, got another $27 billion by issuing bonds that pay off in 2049, $18 billion of which Pimco purchased. That debt is on Beignet's balance sheet, not Meta's.

The notes bear a 6.58% interest rate, much higher than the 5.5% yield on Meta's comparable corporate bond, and have an A+ credit rating, one notch below Meta's AA-.

That's the simple part. Meta will pay rent to use the data center for its AI products and that cash will go to bond interest and principal payments, as well as dividends for Blue Owl. But Meta retains the ability to exit its lease every four years so that the rental agreement won't count as a long-term liability on its balance sheet.

Which is where the unusual guarantee comes in. In exchange for the option to walk away, Meta agreed to make investors whole if it ever did so. In that event, Blue Owl would sell the center, using proceeds to first pay outstanding bonds, then itself. If the sale fetched less than bondholders were owed plus what Blue Owl had invested with a modest profit, Meta would pay the difference.

"It's a fixed-income risk with an equity-like return," said Alexey Teplukhin, a managing director at Blue Owl. "That's what we're aiming for."

This financing for Stargate -- the data-center partnership between ChatGPT maker OpenAI, Oracle and SoftBank -- would be fairly straightforward if it weren't so big.

Vantage Data Centers -- a developer and landlord -- is building two Stargate projects in Texas and Wisconsin, respectively, for $38 billion. Oracle has signed a 15-year lease on the centers but the end user will be OpenAI, which has signed a $300 billion purchase agreement with Oracle.

As a startup, OpenAI can't borrow the money and Oracle has a low credit rating for a tech giant. Instead, banks will provide the money through project-finance loans secured by the data centers, much like a mortgage. Oracle's lease payments will go toward paying off the debt.

Such arrangements are standard for large construction projects. But the scale of the deal -- named Jacquard by its top lenders JPMorgan Chase and Mitsubishi UFJ Financial Group -- is off the charts.

Typically, no more than a dozen banks arrange a project-finance loan, but the Jacquard "syndicate" includes more than 30 institutions, according to the deal materials reviewed by the Journal. Members range from global banks like BNP Paribas to regional players like U.S. Bancorp and they are making a full-court press to reduce their exposure by selling the debt to investors.

The loan comes due in five years and pays interest of about 6.4%, almost 2 percentage points more than the yield of Oracle's comparable corporate bond. Since the loan is so big, banks are pitching it to almost every type of buyer, including insurers, asset-backed bond specialists and corporate debt funds.

Not everyone can buy the debt though, reflecting the difficulty measuring risk in deals involving an unknown technology like AI.

JPMorgan only got one credit rating for the Stargate loan from Kroll Bond Rating Agency, a relatively small firm. That means the banks won't be able to sell the loans to managers of collateralized loan obligations, or CLOs, which control about $1.3 trillion but require ratings from one of the top three firms, Fitch Ratings, Moody's Investors Service and S&P.

News Corp, owner of the Journal, has a content-licensing partnership with OpenAI.

Elon Musk is asking investors to fund xAI, the startup he hopes will overtake OpenAI.

Musk is also using cash from other parts of his corporate empire, like SpaceX, and xAI has sold shares and raised debt worth at least $10 billion. But that is a drop in the bucket. xAI needs $18 billion just to buy the 300,000 microchips required to run Colossus 2, its second giant data center in Tennessee, the Journal previously reported.

Earlier this year Musk ally Antonio Gracias used his investment firm Valor Equity Partners to indirectly bankroll the chips. Valor committed to buy private equity in a financing vehicle that would borrow billions more debt from private-credit funds, all to be used to purchase chips from Nvidia.

Valor is recruiting other funds to buy shares in the vehicle, called Valor Compute Infrastructure, which would allow it to borrow more debt. The investment firm also tapped private-credit powerhouse Apollo Global Management for help arranging the debt, according to people familiar with the deal.

The deal could end up raising $7.5 billion of equity and $12.5 billion of debt, but might be smaller if fewer shares are sold, the people said. While the debt will be paid off in five years with rent charged to xAI, returns on the equity will largely depend on what the chips are worth after that. Some analysts worry that circularity between Nvidia and its clients is creating a market bubble.

Apollo is pitching investors debt that would bear a 10.5% interest rate and potential additional payouts if the value of the chips exceeds certain thresholds, the people familiar with the deal said.

Write to Matt Wirz at matthieu.wirz@wsj.com

 

(END) Dow Jones Newswires

November 11, 2025 14:47 ET (19:47 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.

At the request of the copyright holder, you need to log in to view this content

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10