Are Google, Microsoft, and Meta Dishonest? Chen Feng-hsin Reveals Nikkei Investigation: Hidden Debts in Financial Reports Reach $53 Trillion

According to statistics from Bank of America Global Research, by early July, new bond issuance by AI-related companies this year had reached $270 billion—nearly double the total issuance for all of 2025. However, veteran media personality Chen Feng-hsin, on her program 'East, West, North, South: Dragon and Phoenix Match,' cited a recent Nikkei investigation into major U.S. cloud service providers. The investigation found that these companies have hidden liabilities in the footnotes of their financial reports—debts not visible in the balance sheet's liability section. These shadow debts amount to $1.65 trillion (approximately NT$53.44 trillion), an eightfold increase over the past four years.

Chen noted that Alphabet claims its AI data centers are continuously burning money but has yet to present a clear monetization strategy. If even Alphabet is in this situation, what about Meta, Microsoft, Amazon, Oracle, and CoreWeave? She also pointed out that former U.S. President Donald Trump is supporting Oracle because its founder is a close friend. Oracle suddenly secured a 10-year contract with the Pentagon, which she believes is an attempt to rescue the company. However, she questioned how long such support can last, emphasizing that the key lies in revolving credit mechanisms.

Chen raised concerns about recent high-profile moves by NVIDIA, including its partnership with SK Group and its financial guarantee for OpenAI. NVIDIA has been involved in at least six investment, financing, and guarantee deals this year, but two recent ones have drawn particular attention. Over the weekend, NVIDIA CEO Jensen Huang partnered with SK Group, the parent company of SK Hynix, to invest $500 billion in AI-related projects in the U.S. But NVIDIA is a GPU company—why is it now building data centers, a task typically handled by its customers? While these facilities will naturally use NVIDIA GPUs, the real issue is accountability: who bears the cost if the investment fails?

Then, NVIDIA provided a $250 billion financing guarantee enabling OpenAI to purchase computing power from a data center built by SoftBank in Ohio, U.S. The problem is that OpenAI is still unprofitable, burning cash, and its IPO next year remains uncertain. If this investment fails and banks demand repayment from OpenAI, and OpenAI cannot pay, NVIDIA, as the guarantor, will be liable.

Chen emphasized that the issue isn't limited to NVIDIA. She highlighted CDS (Credit Default Swap) as a key indicator. CDS functions like insurance: if an investor buys a company's bond but fears default, they purchase a CDS to protect against losses. Rising CDS prices mean investors must pay higher premiums to insure against default, indicating deteriorating creditworthiness. Currently, CDS prices are rising not just for NVIDIA, but also for Oracle, Alphabet, Meta, Microsoft, and Amazon. This shows these companies are not only burning cash to build data centers and taking on debt, but their borrowing costs are also sharply increasing.

The Nikkei investigation reviewed financial reports from major U.S. cloud providers—Meta, Amazon, Google, Microsoft, and Oracle—and examined their footnotes to uncover hidden contracts and obligations. These shadow or hidden liabilities total $1.65 trillion and have grown eightfold in four years. AI is consuming massive capital, but monetization remains elusive. There is still no solid data showing AI can generate profits. Skepticism about this debt-driven growth cycle is growing louder. These two factors have contributed to poor stock performance this year for both the five cloud giants and NVIDIA. NVIDIA, once valued above $5 trillion, is now below that threshold—highlighting investor concerns that are now spreading to the upstream semiconductor supply chain.

Chen asked whether AI development has ended or if the bubble has burst. She firmly answered, 'No.' She believes this cash-burning model can last another one to two years. Even as stock prices fall, companies can continue to raise funds. These firms cannot afford to fall behind—if they fail to prepare in this race, lacking exclusivity will lead to even worse outcomes. If an AI monetization model emerges within the next one to two years, AI could experience another explosive growth phase. But if monetization is further delayed, the situation will become much more severe in one to two years. For now, this is merely a correction.

Veteran media commentator Tang Hsiang-lung stated the core issue is that the market can no longer assess the risks of these AI giants. Revenue is invisible, expenses are hidden, and future revenue models are too uncertain to evaluate due to numerous variables. The market suspects these companies are concealing much of their spending. Even Google, once a highly profitable '躺着赚钱' (lying-down-earning) company, has maximized hidden leverage through capital expenditures. While its financials appear strong on paper, from a risk assessment perspective, it's terrifying. This situation closely resembles the pre-2008 financial crisis—different industry, same pattern: healthy-looking balance sheets but no optimistic outlook for future income.

Tang noted that these AI companies are now collectively bound together, not just in long-term capital expenditures, data centers, AI chip procurement commitments, or financing. When both revenue and expenses are invisible, risk becomes unassessable. All major AI players are now interconnected within the same financing ecosystem. Tang warned that if a crisis occurs, it won't be isolated. Oracle might be the first domino to fall, but if it collapses, many others will follow. Everyone's pockets are already empty. When capital expenditures are leveraged to the maximum through hidden means, his instinctive reaction is, 'They have no capacity to withstand a crisis.' If a sudden crisis hits, these companies will be financially drained and unable to survive.

This situation is severely undermining confidence in the U.S. AI system's ability to withstand risk. The longer the monetization cycle becomes, the fiercer the competition among large language models, forcing price cuts to capture market share. But the cost of new large language models is extremely high, creating a 'burning from both ends' scenario. As the monetization timeline extends, crisis resilience will be questioned. The Nikkei's findings highlight a critical problem: how can the market assess risk when it cannot see these companies' revenues or financials? The inability to assess risk is the market's greatest fear—because no one knows what else these companies are hiding. Even a slight market shift could trigger a massive bubble storm. This risk must not be underestimated. AI giants must return to a more honest stance. They should stop downplaying long-term capital expenditures with verbal promises or calling binding contracts 'letters of intent' to mislead investors. Pretending signed contracts are merely non-binding intentions is deceiving ordinary investors and lies at the heart of current market panic.