Five of the largest technology corporations in the United States hold approximately US$ 1.6 trillion, equivalent to R$ 8.4 trillion, in financial commitments that are not included in their accounting reports. This amount exceeds the US$ 1.3 trillion (R$ 6.9 trillion) in liabilities formally declared by these companies.
According to a Nikkei report, these commitments are being used to fund the expansion of data centers dedicated to artificial intelligence (AI). The research indicates that the volume of debt kept off the accounting records has increased by approximately eight times over four years, involving Alphabet, Microsoft, Amazon, Meta, and Oracle.
Although this practice recalls the model adopted by Enron before its collapse, the study emphasizes that it currently remains within current accounting standards and is considered legal due to the transparency and disclosure requirements established after the American company's scandal.
The study compares this to the method used by Enron about 25 years ago, a period when the company used special purpose vehicles to hide debt, which was later classified as fraud. However, the current analysis points out that the use of these structures is still permitted, provided that all information is adequately communicated to investors.
Analyst Gil Luria commented to Bloomberg Law that 'Enron's crime was not having special purpose vehicles,' but rather 'hiding them.' Companies use separate entities to finance their infrastructure.
The mechanism involves transferring debts related to the acquisition of servers, chips, and electricity to autonomous legal entities, often configured as joint ventures. In this way, the financial commitments are no longer registered directly in the balance sheets of the technology companies.
One example cited is Meta's Hyperion data center project in Louisiana. In this case, Meta and Blue Owl Capital invested in an independent structure that assumed US$ 27 billion (R$ 137.4 billion) in debt. Although Meta is the sole user of the facility, the company argues that it does not need to record this liability on its balance sheet because it is not responsible for finding new tenants if the space becomes vacant.
Additionally, the study mentions that Oracle has US$ 260 billion (R$ 1.3 trillion) in future lease obligations that will be gradually incorporated into its accounting books. Nvidia has accumulated US$ 119 billion (R$ 605.6 billion) in purchase obligations. Alphabet and Microsoft also employ structures kept outside their official records.
The scale of these operations is a point of great attention. Only Meta reportedly has about US$ 420 billion (R$ 2.1 trillion) in off-balance-sheet debt, an amount almost three times greater than the liability declared by the company itself. In Oracle's case, the volume of these obligations has grown by about 30 times in the last four years.
All these investments are part of the intense competition for the infrastructure necessary for AI. It is expected that the sector will inject more than US$ 3 trillion (R$ 15.3 trillion) by 2028 into the construction and equipping of AI data centers, with a large part of this capital being financed by guarantees tied to the chips installed in these facilities.
The topic gains importance at a time when four of the five companies will release their financial results in the next two weeks. The balance sheets will only present the officially registered debt, while the US$ 1.6 trillion in external commitments appear only in the explanatory notes.
The survey also warns about the future consequences of these structures. When a data center begins to operate, its lease contracts become part of the company's balance sheet. If demand for AI falls short of predictions, the facility may depreciate, transferring losses to the creditors and insurers responsible for the financing.
Part of the market is already expressing apprehension about the increase in financial leverage. S&P downgraded Oracle's credit rating due to the high degree of indebtedness, and Morgan Stanley and Moody's have also expressed concerns related to the growth of these obligations.
Accounting consultant Tom Selling questioned, in an interview with The Next Web, the risks inherent in this model, asking: 'What if one of these companies were a house of cards,' and if it were sustaining itself with this accounting treatment?
The study emphasizes that the use of these structures is not illegal, and the companies argue that the future profits generated by AI expansion will be sufficient to cover the assumed commitments. This information about the obligations is made available to investors, although it is located in the explanatory notes of the financial statements. This implies that investors who analyze only the main figures of the balance sheets will see only a fraction of the companies' financial leverage, in a scenario where the market debates the possibility of an AI-related bubble.
Artificial intelligence has transcended the status of mere technological speculation, becoming a driver for vast financial operations. Banking institutions are actively involved in fundraising, lending, and transactions related to the expansion of essential infrastructure for AI systems.
According to Reuters, the investment cycle in this sector could last for several years, although there is market debate about the sustainability of technology company valuations. The pursuit of greater computational capacity has already resulted in billion-dollar operations in the financial sector.
Notable examples include the offering of American Depositary Receipts (ADRs) from chip manufacturer SK Hynix, valued at US$ 26.5 billion (approximately R$ 143 billion), and the IPO of SpaceX, which reached US$ 86 billion (about R$ 464 billion).
Banks have also been financing infrastructure projects, focusing mainly on building data centers and increasing the processing capacity required by artificial intelligence applications. David Solomon, CEO of Goldman Sachs, noted during an earnings conference call that AI infrastructure is still in its early stages and that this investment cycle will continue to foster strategic activities, financing, and capital formation.
Solomon characterized the sector as being in the middle of an 'AI investment supercycle,' where companies seek various forms of financing to support their endeavors.
Despite banking enthusiasm, there is apprehension among investors. In July, technology stocks, especially semiconductor manufacturers, faced pressure due to doubts about the high valuations of the segment.
However, infrastructure spending projections continue to rise. Ted Pick, CEO of Morgan Stanley, indicated that estimates for data center investments have increased in recent months: the forecast for 2026 jumped from US$ 575 billion (about R$ 3.1 trillion) to around US$ 850 billion (R$ 4.6 trillion); the expectation for 2027 increased from US$ 700 billion (R$ 3.8 trillion) to US$ 1.3 trillion (R$ 7 trillion); and the projected investment for 2028 could reach US$ 1.5 trillion (R$ 8.1 trillion). Furthermore, Morgan Stanley estimates that AI-related spending could total US$ 10 trillion (R$ 54 trillion) over several years.
Artificial intelligence has also become a central topic in discussions among financial sector executives. Jane Fraser, CEO of Citi, commented that the technology is 'dominating many conversations,' mainly due to the growth in spending on technology, data centers, energy, and defense.
Bank of America provided a credit line of US$ 520 million (approximately R$ 2.8 billion) to OpenAI. The same institution stated it has helped AI-related companies raise nearly US$ 500 billion (about R$ 2.7 trillion) since 2025.
Stephen Biggar, Director of Financial Services Research at Argus Research, observed that the investment cycle is already impacting various areas of the market, stating that the 'AI-driven investment supercycle has benefited stock issuances, mergers and acquisitions activities, and debt financing.' Jeremy Barnum, CFO of JPMorgan Chase, highlighted that the expansion of data centers also stimulates secondary sectors, such as construction and specialized services.
In summary, Reuters points out that the increase in investments in artificial intelligence has opened a new front of opportunities for Wall Street, although the market continues to monitor whether this pace of expansion will be maintained in the coming years.
