Building on the Edge of a Cliff: The Unspoken Debt of Google and Meta

In August 2025, someone registered seven companies in Delaware, all with the word "Beignet" in their names.

Beignet is a fried pastry commonly found on the streets of New Orleans, coated in a thick layer of powdered sugar that always seems to fall onto your clothes when you pick one up.

No matter how hard you look, you'd never guess this dessert has anything to do with AI.

A month after the Beignet companies appeared, Meta built a data center in Louisiana named Hyperion, covering an area equivalent to four New York's Central Parks.

To build this massive data center, Meta borrowed a total of $27.3 billion.

But if you look closely at Meta's financial reports, you'll find that on its balance sheet, the entire record related to this project amounts to only a $2.37 billion investment.

The remaining over $200 billion in debt has vanished.

This is not an isolated incident.

On July 22nd, Nikkei reported that US tech giants had hidden up to $1.65 trillion in debt where no one could see it, a figure even surpassing the $1.35 trillion total liabilities on their balance sheets.

We combed through the filings these five companies submitted to the SEC and found the situation to be even more dramatic than that report suggested.

On July 23rd, the day after the report was published, Google's parent company Alphabet filed its next quarter's documents. Its purchase commitments jumped from $332.4 billion three months prior to $811 billion. A year earlier, that figure was $62.1 billion.

In one year, it multiplied thirteen times. This also pushed the $1.65 trillion figure cited in the Nikkei report up to $2.13 trillion.

Over the past year, the debt created by Microsoft, Google, Amazon, Meta, and Oracle on data centers rose from $710.8 billion to $1.55 trillion. If hardware procurement and construction contracts for items like GPUs are included, it surged from $1.02 trillion to $2.86 trillion. That's nearly double in a single year.

And the portion actually sitting on their balance sheets accounts for only a quarter of this final figure. Some $2.13 trillion in "data center debt" has disappeared from the giants' balance sheets.

Where did all this debt go?

The Bank for International Settlements noticed this long ago. In a quarterly report from March 2026, it gave this practice a name: Shadow Borrowing. The report stated that these arrangements were economically equivalent to debt, yet mostly remained off the companies' balance sheets.

Three months later, in its annual report, the BIS took the rare step of listing the AI bubble and circular financing alongside sovereign debt as top risks to the global financial system.

By reviewing various filings from the five companies, at least five different methods can be identified: SPVs, credit wrap derivatives, finance leases, residual value guarantees, and leases not yet commenced.

And those helping them create this shadow debt are turning this into a whole new business.

Tech Giants' Collective "Return to Poverty"

For the past two decades, tech giants were the most comfortable type of company in the US capital market. They were profitable, had cash on hand, and bought back their own stock.

In Q4 2021, Microsoft, Google, Amazon, Meta, and Oracle collectively repurchased $48 billion worth of shares, with Meta alone spending $20 billion. With such abundant cash, shareholders didn't have to worry about them running short of money.

But in Q1 2026, the total buybacks among the five plummeted to $4.6 billion.

Over the past two years, tech companies' capital expenditures related to AI have more than doubled, but operating cash flow growth is less than 60%. At this growth rate, by mid-2027, tech giants will "collectively return to poverty," reverting to an era of losses.

Morgan Stanley estimated that by 2028, tech companies would need to spend approximately $2.9 trillion on AI, but they can only generate about $1.4 trillion internally. The gap of $1.5 trillion must be found outside of cash flow.

So they started borrowing. From 2020 to 2023, the five companies issued an average of about $31.3 billion in bonds annually. By July 2026, this figure reached $189.7 billion, six times the historical average.

The massive amount of corporate bonds is already straining the market. Over the past nine months, Amazon's bond oversubscription ratio has been consistently declining. It was 5.3 times in November 2025 but dropped to just 1.6 times by July 2026.

The cost of issuing bonds has also increased. This year, in the entire investment-grade bond market, an average of just 4 basis points extra is needed to sell a new issue. But for Amazon's July issuance, it took an extra 18 to 21 basis points to sell.

Google and Oracle went even further, raising nearly $80 billion through stock offerings and subsequently announcing a total $60 billion ATM offering plan.

Financial media outlets have widely criticized tech giants for breaking the "unspoken covenant" with investors. In the past, the market bought these companies' stocks based on the implicit understanding of net cash, low debt, and continuous buybacks. Now, they are heavily issuing debt and halting buybacks.

But the real trouble is written on the giants' balance sheets.

The more debt on the balance sheet, the more cautious rating agencies become, and the narrower the pool of potential bond buyers. Oracle was the first among the five to hit this wall. Its capital expenditure jumped from $21.2 billion to $55.7 billion in one year. In July 2026, S&P downgraded it from BBB to BBB-, just one notch above junk status. A further downgrade would force global insurers and pension funds to sell Oracle bonds under regulatory requirements.

So, the giants don't just need more money; they need money that is more discreet, longer-term, and carries less onus. And this kind of money simply can't be found in the public markets.

After SaaS Faltered, Wall Street is Also Looking for New Business

Just as tech companies were worrying about money, another part of Wall Street was also looking for an exit.

In the first half of 2026, a fund under Blue Owl received redemption requests approaching 40% for two consecutive quarters, but only managed to pay out just over 10%. The company's stock price fell from $24 to $9.

Blue Owl is one of the world's largest private credit managers, overseeing over $310 billion in assets. The fund facing the redemption run specializes in software, with software loans comprising over 60% of its portfolio.

Unfortunately for it, in 2026, the last thing Wall Street wants to hold is software loans.

From its peak in October 2025, software stocks have fallen nearly 40%. The market's explanation is simple: AI will kill software. Investors were previously willing to give software companies high valuations because customers renewed annually, making revenue seem like it could grow indefinitely with contracts. Now, whether customers will continue to renew has suddenly become a question.

In reality, the fundamentals of software companies aren't that bad. Microsoft 365 subscription revenue growth accelerated from 15% to 19%. Revenue growth for software companies like ServiceNow, Salesforce, and Snowflake has accelerated for five consecutive quarters. Gartner even raised its forecast for global software spending.

But in the financial world, confidence often matters more than fundamentals.

Blue Owl acknowledged in a shareholder letter that market concerns about AI's impact on software companies have significantly affected how investors view software credit exposure.

Wall Street urgently needs a new narrative to get investors back on board. That narrative is the data center.

Software loans bet on whether customers will renew next year. Data center loans bet on whether AI companies will need computing power. The former question is increasingly difficult to answer, while the latter seems almost rhetorical. The stronger AI becomes, the more valuable the server rooms.

Currently, data center loans are the hottest business on Wall Street. In December 2025, Blue Owl rejected Oracle's data center project in Michigan, citing it didn't meet underwriting standards. Shortly after, PIMCO snatched up the deal at a higher price. Such competitive bidding happens almost monthly on Wall Street.

On one side are cash-starved tech giants; on the other, asset managers with nowhere to deploy capital. A match made in heaven.

Their first major creation is Beignet—the dessert mentioned at the beginning.

How Meta Hid $28 Billion in Debt Inside a Pastry?

That Hyperion hyperscale data center in Louisiana, co-built by Meta, is initially registered under Laidley LLC. The entity operates the campus and also signed the 15-year power supply contract with the local utility company.

Hyperion's footprint compared to Manhattan, NYC. Source: Bloomberg

Laidley belongs to Project Beignet Holdings, a joint venture. The ownership of the data center resides here.

The major shareholder of the joint venture is Beignet Investor. It issued the $27.3 billion in bonds.

The debt isn't placed on the company that owns the data center; it's placed on its shareholder.

Further up is Beignet Pledgor. It wholly owns Beignet Investor, and pledges all equity of Beignet Investor to a trustee.

This provides creditors with a very straightforward form of collateral. In case of default, the trustee doesn't need to first go to Louisiana to appraise, sell, and wait through lengthy litigation over a data center. It simply executes the equity contract as agreed. The campus continues operating, the lease continues running, only the recipient of the rent changes.

Above Beignet Pledgor are four more companies, including one named Beignet Net Lease Aggregator. At the very top is Blue Owl's net lease REIT called OSNL, along with co-investors contributing capital.

All seven companies are registered in Delaware. LLCs there are not required to disclose their members or ownership percentages.

The $27.3 billion in debt was also not issued publicly. It went through the 144A channel, sold only to Qualified Institutional Buyers (QIBs), without a public prospectus. To review the transaction terms, one must first sign an NDA. It's also a perpetual 144A, meaning it won't convert into registered public bonds.

Searching for "Beignet" in the SEC's full-text filing system yields only one narrative hit: a subsequent events footnote in a Blue Owl quarterly report. By the annual report, it disappears along with "Meta" and the county name of the project, reduced to a single line summarizing "net lease data centers."

This is already dizzying, but it's just the legal isolation.

SPVs are not new. The real estate and infrastructure industries have used them for decades, and accounting standards have long known people would try to park debt in them, setting up two hurdles for consolidation. Whether an entity should be consolidated into a company's financial statements depends not just on ownership percentage, but on who directs the most significant activities and who bears most of the risks and rewards.

Meta is the sole tenant of Hyperion, provides capital and credit, and is responsible for construction and property management. By this standard, the debt should logically be consolidated into Meta's own statements.

But it kept 20%.

Blue Owl's OSNL fund and co-investors, via Beignet Pledgor and four other holding companies, own 100% of Beignet Investor. Beignet Investor then holds 80% of the joint venture. Meta holds the remaining 20%.

80 and 20 are the two recurring numbers in this story.

Meta's explanation in its earnings report is that the company does not have the power to direct the activities that most significantly impact the joint venture's operating performance, thus it is not the primary beneficiary and does not consolidate the entity. Since the joint venture isn't consolidated in Meta's statements, the $27.3 billion in debt naturally doesn't appear either.

The debt hasn't vanished. It's just changed where it's stored.

Synthetic Leases: Silicon Valley Giants' "Art of Debt Repayment"

Debt off Meta's books doesn't mean Meta doesn't have to pay.

Meta has a subsidiary lessee called Pelican Leap. It signed a four-year lease with Laidley. Starting in 2029, Pelican Leap pays rent to Laidley each month. The money flows from Laidley to the joint venture, to Beignet Investor, and is then used to repay the bond investors' principal and interest.

After this circuitous route, the rent still ultimately comes out of Meta's pocket.

The $27.3 billion bond carries a coupon of 6.581%, maturing in May 2049, using a fully amortizing structure. Unlike typical corporate bonds that repay principal in a lump sum at maturity, this one pays down a bit with each installment, stretching over twenty-four years.

It looks more like a mortgage.

The total rent Meta pays in the first four years amounts to $12.3 billion, averaging $3.08 billion annually. This amount just covers the year's principal and interest payments, with an extra 12% going to the equity investors.

The real drama begins with the lease term.

The bond lasts twenty-four years. The initial lease is only four years. Starting in 2029, it has attached renewal options that can extend the lease for up to twenty years. After 2033, Meta could theoretically choose not to renew and leave.

What happens to the over $20 billion in debt that hasn't been repaid by then?

The answer is hidden on another page of the lease agreement. Besides the monthly rent, Meta provided a residual value guarantee, with a cap of approximately $28 billion—slightly higher than the debt itself. This amount decreases over time. If Meta doesn't renew and the campus value falls below this threshold, Meta must cover the difference.

Putting $28 billion and $27.3 billion together makes it hard not to connect the two things.

So the real backing for the bond isn't just the building or the machines inside it.

It relies on Meta's credit.

This also explains the rating. S&P rated the bond A+, while Meta itself is AA-. The rating agency didn't price it based on a building yet to be operational but downgraded Meta's own credit by one notch.

Meta puts up the money, the credit, runs the operations, is the sole tenant—yet writes in its financial statements that it is not the primary beneficiary.

This clean balance sheet doesn't come cheap. If Meta issued debt of the same tenor in the public market, the cost would be about 5.5%. Through this structure, the cost rises to 6.581%. For the same amount of money, Meta pays nearly $300 million more in interest per year.

Clearly, spending so much money buys more than just a building.

In July 2026, a second similar project emerged, named Sopaipilla—another fried pastry from the American Southwest. The project is in El Paso, Texas, with a minimum bond issuance of $12 billion, also an 80/20 split. The difference is that the 80% stake shifted from Blue Owl to BlackRock.

Meta has used two internal codenames for its next-generation large models: one was Avocado, the other Mango. The financial department's naming clearly has more appetite.

Meta's structure is the most intricate, but it's not the only approach.

Microsoft didn't create a shell company, nor did its book debt significantly grow. Over the past two years, its total debt decreased from $44.9 billion to $40.3 billion. However, during the same period, its finance lease liabilities surged from $27.1 billion to $62.9 billion, more than doubling and exceeding total debt by over $20 billion.

The $62.9 billion is indeed on Microsoft's balance sheet, just not under the "debt" line. It's broken up into "other current liabilities" and "other long-term liabilities." From the most prominent line item, it looks very quiet.

The meaning of a finance lease is quite straightforward. In name, it's a lease, but in substance, it's akin to buying in installments. The lease term covers most of the asset's useful life, so for accounting purposes, it's treated as if the asset is purchased, just with payments spread out. Therefore, it must be fully recorded as a liability but doesn't need to be reported under the debt line.

Google didn't even set up a shell. It provides payment guarantees for data