Silicon Valley's hidden debt: How Big Tech uses record profits to conceal a trillion-dollar gap in its own balance sheets

Xpert Pre-Release

Available in 27 languages 📢

Published on: August 22, 2026 / Updated on: August 22, 2026 – Author: Konrad Wolfenstein

Silicon Valley's hidden debt: How Big Tech uses record profits to conceal a trillion-dollar gap in its own balance sheets

Silicon Valley's hidden debt: How Big Tech uses record profits to conceal a trillion-dollar gap in its balance sheets – Image: Xpert.Digital

Record profits without real money: The truth about the balance sheets of Google & Amazon

Perfectly legal, but risky: Silicon Valley's secret mountain of debt

Hidden liabilities: The dark side of the balance sheets of Meta, Amazon and Co.

The summer of 2026 seemed poised to be another quarter of unstoppable superlatives for the tech giants of Silicon Valley. Alphabet and Amazon presented the world with record profits, and the stock markets initially reacted with euphoria. But anyone who isn't solely blinded by the headline-grabbing net profits quickly encounters a fundamental contradiction: While profits are soaring to dizzying heights, the actual cash flow – the free cash flow – is collapsing dramatically.

How is it possible that the world's most powerful and profitable companies earn hundreds of billions on paper, yet in reality spend more money than they earn? The answer lies deep in the gray areas of modern corporate finance. Behind the glittering facades of official quarterly reports lies an unprecedented financial illusion. Driven by the gigantic capital appetite of artificial intelligence, corporations are increasingly relying on unrealized book profits and highly complex, off-balance-sheet structures. It is a perfectly legal, but extremely risky maneuver in which trillions of dollars in debt are systematically removed from balance sheets. The following analysis reveals how Big Tech conceals its true financial obligations—and why the foundation of the current AI boom is significantly more fragile than the official figures would have us believe.

The Billion Dollar Fairy Tale: Why Big Tech's Profits Only Exist on Paper

Just eight days apart, Alphabet and Amazon reported record profits in the summer of 2026. Both companies celebrated the highest quarterly results in their corporate history, both share prices initially reacted euphorically, and both simultaneously reported negative or severely strained free cash flow. This seemingly contradictory combination of record profits and dwindling liquidity raises a fundamental question that extends far beyond the balance sheets of two technology companies: Where is the money currently flowing into artificial intelligence coming from, and how solid is the foundation on which the current AI boom actually rests?

To answer this question seriously, it's worth taking a closer look beneath the surface of the official press releases. What emerges is not illegal balance sheet manipulation in the sense of classic fraud cases, but rather a network of perfectly legal valuation methods, accounting rules, and off-balance-sheet financing structures that, taken together, create a risk profile that bears little resemblance to the officially reported figures.

When book profits overshadow operational reality

Alphabet reported net income of approximately $112 billion for the second quarter of 2026, an increase of about 298 percent compared to the same period last year. At first glance, this reads like an unprecedented success story, driven by its cloud computing division, search engine advertising, and generative AI business. However, a closer look at the income statement reveals that the vast majority of this profit growth did not stem from core operations.

Of the $112 billion in net income, approximately $98 billion was attributable to a single line item on the balance sheet, reported as other income or unrealized gains from equity investments. This item primarily relates to the increase in value of Alphabet's stakes in unlisted companies, particularly SpaceX and the AI company Anthropic. Following SpaceX's IPO in the summer of 2026, Alphabet had to adjust the book value of its stake to reflect the market valuation achieved, which alone triggered a valuation jump in the tens of billions. In the same quarter of the previous year, the same position had been valued at just around $1.3 billion.

The crucial point here is that these are largely unrealized book gains. This means that not a single cent of this money actually flowed into Alphabet's coffers. It is a revaluation of shares on the balance sheet that, for the most part, cannot be sold or are not intended to be sold for strategic reasons. The operating profit, that is, the portion of the profit that actually comes from the core business of advertising, cloud services, and other products, was around $40.8 billion in the same quarter, significantly lower than the much-celebrated total profit.

Cash flow as a reality check

While the income statement was inflated by book gains, the cash flow statement painted a completely different picture. Alphabet spent approximately $44.9 billion on investments in data centers, chips, and other infrastructure during the same quarter, while operating cash flow was around $39.1 billion. The difference resulted in a free cash flow of negative $5.9 billion, the first negative quarter in the company's history since its IPO in 2004.

This finding is remarkable because it shows that a company, despite posting record profits in a single quarter, spent more money than it actually earned. The share price reacted accordingly, falling several percentage points the day after the announcement, because professional investors considered cash flow, rather than book profit, to be the key indicator of the company's true financial health.

Amazon's billion-dollar profit from a single investment

Eight days after Alphabet's figures, Amazon followed with a structurally almost identical story. The company reported a net profit of $62.6 billion for the second quarter of 2026, an increase of approximately 245 percent compared to the same quarter of the previous year. Here, too, it's worth examining the composition of this result. According to the company's own statement, of the $62.6 billion profit, approximately $53.4 billion came from a non-operating income item, primarily attributed to its investment in Anthropic.

Since 2023, Amazon had invested a total of approximately eight billion dollars in Anthropic in several tranches, partly in the form of convertible bonds and partly as non-voting preferred shares. Following a new funding round for Anthropic, which saw its valuation temporarily rise to several hundred billion dollars, Amazon was required to revalue its stake according to applicable accounting rules. This so-called mark-to-market adjustment generated an accounting gain of over 50 billion dollars, even though not a single share was sold and not a single dollar was actually received.

If this one-time, non-cash effect is excluded, the remaining net operating profit is estimated at around $20 to $21 billion. Amazon's operating business, driven primarily by strong growth in its cloud division Amazon Web Services and a solid advertising business, thus performed quite well; however, roughly two-thirds of the record profit reported in the media was due to the valuation effect of a single private investment.

Two corporations, one pattern

What we observed at Alphabet and Amazon within a few days of each other is no coincidence, but a systematic pattern arising from the unique structure of the current AI boom. The world's largest technology companies are not only operators of cloud infrastructure and consumer products, they have also become some of the largest private investors in AI startups like Anthropic and OpenAI. These investments often occur in a circular fashion: A cloud provider invests in an AI lab, the AI lab in turn purchases computing power from the same cloud provider, and if the AI lab's valuation increases in a new funding round, the cloud provider records a book profit on its investment.

This situation leads to a peculiar distortion of corporate reporting. The profits of the largest technology companies are increasingly driven by private market valuations, which are themselves difficult to verify and whose volatility is hard to predict. If the valuation of Anthropic or a comparable company falls in a future financing round, the effect would be exactly the opposite, triggering a corresponding book loss that could significantly reduce the profits then reported.

The debt side of the balance sheet and its blind spots

While book profits dominate perceptions on the revenue side, a mirror image phenomenon emerges on the debt side. An analysis by the Japanese business newspaper Nikkei Asia, which sifted through the footnotes of the annual reports of Alphabet, Microsoft, Amazon, Meta, and Oracle, arrives at a startling result. The five companies report a combined total of approximately $1.35 trillion in reported debt on their official balance sheets. Simultaneously, their off-balance-sheet liabilities—that is, payment commitments and contractual obligations that, for various accounting reasons, do not appear as traditional debt on the balance sheet—amount to an estimated $1.65 trillion.

This means that the hidden liabilities significantly exceed the officially reported debt. Therefore, anyone calculating the debt of these five companies based on their official balance sheet figures is working with less than half of their actual total financial burden. According to analysts' calculations, this off-balance-sheet volume has grown approximately eightfold in just four years, which corresponds almost exactly to the growth rate of investments in AI infrastructure.

How debts disappear from the balance sheet

The mechanism by which these obligations disappear from the traditional balance sheet is technically complex, but fundamentally comprehensible. A key component is long-term purchase commitments for graphics processors and other hardware, which are contractually binding, but according to current accounting standards only need to be recorded as liabilities when the corresponding equipment is actually put into operation. A company can thus sign orders worth tens of billions of dollars over several years without this sum appearing as a liability on its current balance sheet.

Another, even more significant component is joint ventures with specialized private equity firms. The most prominent example is the financing of Meta's Hyperion data center complex in Louisiana. For this project, Meta partnered with the private equity firm Blue Owl Capital to establish a separate project company. This company, legally independent from Meta, raised approximately $27.3 billion in debt capital through bond issuances. Blue Owl and affiliated investors hold about 80 percent of the shares in this project company, while Meta itself holds only about 20 percent.

Meta pays a long-term lease for the use of the data centers financed in this way. This lease is contractually fixed for decades and effectively functions like a loan repayment, although it is not technically considered debt capital for Meta itself. Because Meta is not considered the beneficial owner of the project according to the relevant criteria for the consolidation of special purpose entities, the corporation does not have to report the $27.3 billion debt on its own balance sheet. Only the share in the project company and future lease payments, which will only become visible on the balance sheet once the facilities are operational in 2029, are reflected.

According to the Nikkei analysis, Meta has accumulated off-balance-sheet liabilities totaling approximately $420 billion, almost three times the company's officially reported debt. This makes Meta the most extensive user of this financing technique among the five companies examined.

Our US expertise in business development, sales and marketing

Our US expertise in business development, sales and marketing - Image: Xpert.Digital

Industry focus areas: B2B, digitalization (from AI to XR), mechanical engineering, logistics, renewable energies and industry

More information here:

Expert Business Hub

A thematic hub offering insights and expertise:

Knowledge platform covering global and regional economies, innovation and industry-specific trends
A collection of analyses, insights, and background information from our key areas of focus
A place for expertise and information on current developments in business and technology
A hub for companies seeking information on markets, digitalization, and industry innovations

Hidden debts in the AI age: What the footnotes of quarterly reports conceal

Why these constructions are legal, but not harmless

In the accounting scandals of the early 2000s, such as the case of the energy company Enron, debts were also outsourced to special purpose vehicles, but in violation of applicable law and with the aim of actively deceiving investors. The structures used today by large technology companies differ from these in one crucial aspect: they are entirely legal, they comply with applicable accounting standards, and they are disclosed in the footnotes of the respective annual reports.

However, therein lies the real risk. While law enforcement agencies can take action against illegal accounting manipulation, there is no comparable corrective mechanism against perfectly legal but economically high-risk schemes. Essentially, the only recourse is education by independent analysts, specialist journalists, and a critical public that reads, understands, and contextualizes these footnotes. The relevant information is publicly available, but it is scattered across hundreds of pages of quarterly reports, technical notes to financial statements, and separate communications from rating agencies, meaning that its entirety is rarely noticed at first glance.

The role of private credit markets

A key driver of this development is the enormous growth of private credit markets in recent years. Funds like Blue Owl Capital, as well as major players like Pimco and BlackRock, are willing to provide very long-term loans secured by specific projects, offering yields significantly higher than those of traditional corporate bonds issued by the participating technology companies. In the case of the Hyperion financing, the interest rate premium over comparable US Treasury bonds was around 225 basis points, corresponding to an effective interest rate of approximately 6.6 to 6.9 percent, which is considerably higher than what Meta would have to pay for traditional corporate bonds listed on its own balance sheet.

This deal is attractive for lending funds because they receive an implied credit rating close to that of Meta itself, combined with a significantly higher return. The rating agency Standard & Poor's rated the project company's bonds A+, just one notch below Meta's own credit rating, because a so-called residual value guarantee from the group largely mitigates the risk for lenders. Such securities are currently in high demand for institutional investors such as pension funds and insurance companies seeking long-term, high-yielding, and supposedly safe investments.

At the same time, this means that the risks of the AI construction boom are no longer concentrated solely on the balance sheets of technology companies, but are increasingly being carried into the entire financial system, into pension funds, insurance portfolios and other institutional investment vehicles, whose customers are often unaware of this shift in risk.

The bet on ever-increasing valuations

Both the profit and debt sides of current AI balance sheets are ultimately based on the same fundamental assumption: that the valuations of the participating private AI companies and the underlying physical infrastructure will continue to appreciate or at least remain stable. If Anthropic's valuation increases further in a future funding round, Alphabet and Amazon will again be able to record book profits. If the demand for computing power remains consistently high, the data centers financed through special purpose vehicles will reliably generate enough revenue to cover their rent payments and service their underlying loans.

However, should this assumption prove overly optimistic, for example, because the actual economic use of generative AI applications develops more slowly than current valuations suggest, both effects could simultaneously reverse. Falling valuations of private AI investments would lead to significant book losses for investors, while at the same time, contractually agreed-upon lease payments for the newly built data centers would have to continue unchanged, regardless of whether there is actually demand for this computing power.

Several market observers and rating experts have already warned of a misallocation of capital in this context. Critics from the field of credit analysis point out that these financing structures effectively create twenty- to thirty-year payment obligations for a technology whose specific hardware generation could become technologically obsolete within just a few years. This discrepancy between the financing term and the actual economic life cycle of the financed assets is considered one of the key weaknesses of the current structure.

A comparison with previous financial crises

The current situation exhibits structural parallels to earlier phases of financial history, in which complex but technically legal financing structures were initially celebrated as innovations before their systemic risks became apparent. Prior to the global financial crisis of 2007 and 2008, large amounts of risk were also offloaded from the balance sheets of participating banks via structured special purpose vehicles and redistributed to a broad range of institutional investors. Back then, too, the structures used were considered legally sound until the underlying assumptions about stable or rising real estate prices proved false, triggering a chain reaction throughout the global financial system.

Today's situation differs in important details, particularly in that the technology companies involved have exceptionally strong operating businesses, high cash reserves, and, by historical standards, very high creditworthiness. Nevertheless, both episodes share a common basic pattern: the shifting of risks into less transparent structures, the concentration of these risks with specialized but relatively small lenders, and a collective bet on the uninterrupted continuation of the current growth trend.

What kind of objective assessment remains?

The crucial conclusion of this analysis is not a blanket condemnation of the companies involved or a prediction of imminent collapse. Investments in data centers, chips, and AI models are based on real, technologically sound demand, and the corporations involved continue to have solid operating business models that remain profitable regardless of the valuation effects described. Rather, the real message lies in the need for a more nuanced examination of the published figures.

Anyone who assesses the balance sheets of leading technology companies solely based on headlines about net profit and revenue growth systematically overlooks two key sources of risk: the increasing dependence of reported profits on volatile, non-cash valuation effects of private investments, and the substantial volume of liabilities that are kept out of the traditional balance sheet through creative, yet entirely legal, financing structures. Both of these factors distort the picture of actual financial stability in a more positive direction than a complete, consolidated analysis would reveal.

For investors, analysts, and an informed public, this means that traditional key figures such as net profit or reported debt ratios are no longer sufficient to adequately assess the true economic situation of these companies. Free cash flow, the composition of other income, and the off-balance-sheet liabilities and special purpose entities disclosed in the footnotes, on the other hand, are gaining significantly in importance. Those who consistently examine these levels gain a more realistic, albeit considerably less euphoric, picture of the financial architecture upon which the current AI boom actually rests.

Tech giants under pressure: Why traditional balance sheets are misleading in the AI boom

The upcoming quarterly reports will reveal whether the described pattern continues or whether initial signs of normalization emerge. Should Anthropic's valuation rise significantly again in another funding round, Alphabet and Amazon are likely to report exceptionally high book profits once more, while free cash flow is expected to remain under pressure given continued increases in capital expenditures for data centers and chips. At the same time, it is anticipated that other technology companies, including Microsoft, Amazon, and possibly Alphabet itself, will adopt Meta's model of special purpose vehicles with private debt funds for their own data center projects, as their own balance sheet capacity is increasingly stretched to its limits by the sheer scale of the planned investments.

Industry estimates predict that the five largest American technology companies alone will invest well over $500 billion in capital expenditures in 2026, many times the level of just a few years ago. A significant portion of these sums will no longer be financed from the companies' ongoing operating cash flow, but rather through a growing variety of debt financing instruments, ranging from traditional corporate bonds and trade finance to the special purpose vehicles described above. This development is likely to further deepen the interconnectedness between the technology sector and global credit markets in the coming years, thereby increasing the systemic importance of this sector for the stability of the entire global economy.

🎯🎯🎯 Data-driven B2B industry hub as a quasi-in-house solution

The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business - Image: Xpert.Digital

Xpert.Digital is a data-driven B2B industry hub led by Konrad Wolfenstein . The company acts as an external, quasi-in-house solution for industrial partners, closing operational gaps in marketing, content, and sales – without requiring additional resources on the client side.

More information here: