Is the free lunch for AI over?

Over the past two years, the market has assumed that someone would always foot the bill for AI infrastructure. This assumption is now being rejected by the credit market.

On July 28, South Korea's KOSPI fell 10.84%, triggering its circuit breaker for the eighth time this year. NVIDIA dropped 5%, SK Hynix fell 7.47%, and Japan's Kioxia plummeted 18%.

On the same day, Apple rose 1.2%, reclaiming the title of the world's most valuable company. The Dow Jones Industrial Average gained 0.51%, and major European indices all closed higher.

On the same day, within the same news cycle, movements were completely opposite. The market did not panic—it voted. The assets voted out were those deeply tied to AI infrastructure, while the assets voted in were those not reliant on the AI infrastructure narrative. The signal triggering this vote did not come from the stock market, but from the bond market.

Who Is Saying No

On Monday, NVIDIA's five-year Credit Default Swap (CDS) spreads surged by 14 basis points in a single day to 82 basis points, marking the largest single-day increase for the contract. Oracle's CDS rose to 215 basis points, Alphabet's to 67 basis points, while Broadcom, Amazon, and Meta all hit record highs. SpaceX's newly issued bonds in June broke issue price in the secondary market.

The pricing logic of CDS is entirely different from that of stocks. Stock pricing looks at growth potential, narratives, and the future. CDS buyers—pension funds, insurance companies, and hedge funds—do not read stories; they read cash flows. They ask: Can the cash you are generating now cover the principal and interest due?

For the past two years, the pricing power for AI assets rested entirely with the equity market. Equity investors operated on the assumption that capital expenditure equals a moat; whoever spends more on capex wins, and every GPU is a down payment on future market value. Under this narrative, the combined capital expenditure of the five major tech giants is projected to reach $737 billion in 2026, up from $261 billion two years ago.

Buyers in the credit market remained silent because they do not trade "stories." They were simply waiting.

The direct event that ignited Monday's trading was NVIDIA's July 25 announcement of a $500 billion AI infrastructure partnership with the SK Group, along with OpenAI's $250 billion financing guarantee and $350 billion chip procurement financing arrangement. The potential exposure from OpenAI alone exceeded $600 billion, surpassing Thailand's entire annual GDP. The market spent the weekend digesting the implications of this figure—CDS spreads jumped directly at Monday's open.

Prior to this, signals had been accumulating for three weeks. On July 9, S&P downgraded Oracle from BBB to BBB-, just one notch above junk status. Last week, Alphabet reported negative free cash flow of $5.855 billion for the first time in its twenty years as a public company. When these events occurred individually, the market chose to digest them rather than panic—Oracle's issue could be attributed to customer concentration, and Alphabet's to its investment phase. Each was seen as an isolated problem.

NVIDIA's $600 billion circular financing scheme connected the dots. Oracle's largest customer is OpenAI, which is the core beneficiary of NVIDIA's circular financing scheme. Alphabet, Meta, and Amazon all operate on the same "borrow to build infrastructure" financing model. These are not several independent events. This is the same balance sheet, with repeated collateralization under different borrowers' names.

The credit market took three weeks to digest the signals, establish connections, and make a judgment. On Monday, that judgment landed.

How Credit Entered the Closed Loop

The traditional model of the semiconductor industry was unidirectional: NVIDIA sold chips to cloud providers, who rented computing power to AI companies, with funds flowing upstream from end users, each layer bearing its own credit risk.

NVIDIA's current model has turned it into a cycle. NVIDIA provides credit to customers → Customers use this credit to purchase chips → Chip sales confirm NVIDIA's revenue growth → Revenue growth supports its rating and valuation → A stronger balance sheet supports larger guarantees → The cycle continues.

In this closed loop, part of NVIDIA's revenue growth does not come from end-user demand, but from the credit it injects itself. Sal Naro, Chief Investment Officer at Coherence Credit Strategies, called such arrangements "financial alchemy." John Flood, head of trading at Goldman Sachs, wrote in a client memo that many traders felt "uneasy" about this, noting that the credit market has become a key window for judging the AI narrative.

This is not a problem unique to NVIDIA. The entire financing model for AI infrastructure is built on one assumption—that the cost of capital approaches zero, and the market is always willing to pay for the "future." As CDS begins to price this risk, this assumption is being tested for the first time.

Chain Reaction

Skepticism in the credit market does not need to verify each company individually; it transmits itself down the financing chain.

Oracle was the first to cross the line. On July 9, S&P downgraded its rating from BBB to BBB-, citing three reasons: AI business burning cash beyond expectations (FY2026 capex increasing 162% year-over-year to $56 billion, free cash flow negative $23.7 billion, expected to rise further to $95 billion in FY2027); excessive customer concentration (OpenAI accounts for about half of its remaining performance obligations); and insufficient internal load to absorb excess computing power. Oracle's CDS rose from 144 basis points at the start of the year to 215 basis points. For a company with annual revenue of $67.4 billion and cloud infrastructure revenue surging 77%, the cost of bond insurance doubled.

This risk points directly to OpenAI itself. According to Fortune and the Financial Times, OpenAI's revenue in 2025 was $13.07 billion, with an operating loss of $21 billion. After including one-time expenses, the net loss was approximately $39 billion. A company losing nearly $40 billion a year is Oracle's largest customer and the core beneficiary of NVIDIA's circular financing scheme.

The chain continued to transmit. Alphabet's free cash flow turned negative. Meta's Texas data center issued $12.55 billion in bonds at an interest rate of 7.534%—a coupon rate typically seen only in the junk bond range. The subscription multiple was only 1.6x, less than half the average for US corporate bonds this year. The average borrowing cost for AA/A-rated companies in the tech sector rose from 3.9% a year ago to 5.1%, while BBB-rated costs rose from 4.7% to 7.0%. Unsecured loan rates for lower-rated AI-related companies have touched 10.5%-12%. During the same period, capital expenditure guidance from the five major tech giants continued to rise.

This is not a problem for one company. It is the process of the entire financing model switching from "infinitely cheap" to "market-priced." Every widening spread transmits to the next level, the next borrower, and the next market.

The Bill from 2000

This script is not new.

In March 2000, Cisco became the world's largest company with a market capitalization of $555 billion. Its growth engine was called "vendor financing"—lending money to unprofitable internet companies to buy Cisco equipment. At its peak, the exposure was about $4 billion, covering over 700 customers. CEO John Chambers called it a "revolution once in a century."

After the Nasdaq peaked, customers went bankrupt in droves. In May 2001, Cisco wrote off $2.2 billion in inventory and provisioned $800 million for bad debts. Its stock price fell from 79 to 9.5. In December 2025, Cisco's stock price surpassed its 2000 high, taking 25 years. However, due to continuous share buybacks reducing the share count, its current market capitalization remains more than 40% below its 2000 peak—the stock price returned, but the company's scale did not.

From 1996 to 2000, vendor financing was priced as "ecosystem enablement" and a "demand catalyst." After the Nasdaq crash in 2000, customers died in batches, and Cisco's bad debts surfaced—the same string of numbers was reinterpreted as "when exactly will you pay us back?"

The shift in logic was abrupt: before the break, the equity market provided narrative legitimacy for vendor financing; after the break, the narrative vanished, leaving only the numbers on the bill. Jim Bianco, president of Bianco Research, judged that: before the credit market holds pricing power, equity valuations can always be debated; once the credit market starts pricing, the debate ends.

Cisco's exposure was $4 billion. NVIDIA's potential arrangements related to OpenAI alone exceed $600 billion. The reaction of the CDS market and debt buyers this time was much faster than in the previous round—it took Cisco 14 months from peak to write-off, while NVIDIA's CDS began repricing within one trading day after the news disclosure.

Rewarding Those Who "Did Not Follow"

Apple reclaiming the top spot in global market capitalization today is no accident.

Apple's capital expenditure in the latest quarter was only $1.97 billion, a 36% year-over-year decrease. In the same period, Amazon set its full-year capex guidance at $200 billion, and Alphabet raised theirs to $205 billion. Apple's choice was to return cash to shareholders—adding $100 billion in buybacks in the second quarter while raising dividends. Tim Cook explained this as maintaining a "cautious and thoughtful approach to capital expenditure."

Apple is not ignoring AI. The A19 chip includes a neural accelerator, Apple Intelligence is embedded across all systems, and the new Siri is rebuilt based on Gemini. The difference lies in the path: on-device inference plus external model access, refusing to build large-scale training clusters.

From 2024 to 2025, Wall Street constantly criticized Apple for insufficient investment, with the prevailing judgment being that building proprietary computing power equated to a moat. By the summer of 2026, the credit market reversed this same judgment—large-scale capex shifted from a competitive advantage to a debt burden, and a restrained balance sheet shifted from conservative to flexible. Jay Woods, Chief Strategist at Freedom Capital Markets, said: "Apple was criticized for insufficient AI investment, but now it appears it successfully avoided the capital expenditure trap."

This reversal completed its price confirmation on July 28. The market is not only punishing the side with excessive leverage but also rewarding those with intact balance sheets. These two things are two sides of the same multiple-choice question.

Eighth Circuit Breaker of the Year

The revaluation made by the credit market on Monday took only one day to transmit to the accounts of South Korean retail investors.

The KOSPI triggered its circuit breaker for the eighth time this year. Samsung and SK Hynix account for 60% of the index weight, and these two stocks contributed about 70% of the KOSPI's gains over the past two years.

In May, regulators approved 16 leveraged ETFs linked to them with 2x leverage. In two months, assets under management swelled from 4.9 trillion won to over $9 billion, with 92% held by individual investors. On the way up, leverage amplified gains, attracted incremental funds, and pushed up the underlying stocks. On the way down, the same mechanism operated in reverse: underlying stocks fell → ETFs passively reduced positions → underlying stocks fell further → margin calls → compulsory liquidation → continued fall. In the first two weeks of July, 1.2 million leveraged accounts received margin calls, and 320,000 to 460,000 were subject to compulsory liquidation. Retail investors aged 20 to 30 accounted for 62% of those suffering losses.

This group could not have known what happened before Monday's open. They did not know about NVIDIA's $600 billion circular financing structure, the reasons for Oracle's downgrade, or that the credit market was repricing AI assets using a framework completely different from theirs. They only knew that Samsung and SK Hynix were falling, and their leveraged accounts received margin calls.

This is precisely the structural characteristic of the entire transmission chain: Information flows from professional buyers in the credit market to the equity market, from the equity market to the global semiconductor supply chain, from the supply chain to the index, from the index to leveraged ETFs, and finally to retail investor accounts. Each layer is more lagged, each layer adds its own leverage, and each layer bears a greater shock than the one before. On Monday, CDS spreads jumped 14 basis points, which transmitted to the KOSPI as a circuit breaker. Transmitted to leveraged accounts, it became a wipeout.

The credit market is not just repricing a single stock, but an entire pricing framework that made "capital is always cheap" the default assumption. It is recalculating from the roots, with information transmitting and amplifying level by level along the leverage chain. Those who knew first set the direction, and those who knew last paid the price.