AI đã hết bữa trưa miễn phí?
- Quan điểm cốt lõi: Thị trường tín dụng, thông qua sự gia tăng chóng mặt của CDS, đã phủ quyết giả định "vốn cơ sở hạ tầng AI luôn rẻ", tiết lộ rằng các ông lớn AI do Nvidia dẫn đầu đang xây dựng bảng cân đối kế toán thông qua tái cấp vốn (vay tiền mua chip), tiềm ẩn rủi ro hệ thống, khiến dòng vốn thị trường chuyển từ các tài sản liên quan đến cơ sở hạ tầng AI sang các công ty có bảng cân đối vững chắc (ví dụ Apple).
- Các yếu tố chính:
- CDS kỳ hạn 5 năm của Nvidia tăng vọt 14 điểm cơ bản trong một ngày, liên quan đến thỏa thuận tài trợ 600 tỷ USD với OpenAI đã gây ra sự định giá lại trên thị trường tín dụng, quy mô này vượt quá GDP cả năm của Thái Lan.
- Mô hình tái cấp vốn tồn tại "thuật giả kim tài chính": Nvidia cung cấp tín dụng giúp khách hàng mua chip của chính mình, tăng trưởng doanh thu phụ thuộc vào tín dụng tự tạo ra, chứ không phải nhu cầu thực tế từ người dùng cuối.
- Oracle bị S&P hạ xếp hạng xuống BBB- do đốt tiền AI và phụ thuộc vào một số khách hàng lớn; Alphabet lần đầu tiên có dòng tiền tự do âm kể từ khi niêm yết 20 năm, đánh dấu rủi ro tín dụng lan truyền dọc theo chuỗi tài trợ.
- Với chiến lược chi tiêu vốn thận trọng (capex Q2 giảm 36% so với cùng kỳ), Apple đã giành lại ngôi vương vốn hóa toàn cầu, cho thấy thị trường chuyển từ khen thưởng capex quy mô lớn sang khen thưởng bảng cân đối kế toán vững chắc.
- KOSPI của Hàn Quốc đã có lần ngừng giao dịch thứ 8 trong năm, cơ chế thanh lý bắt buộc của các quỹ ETF đòn bẩy khuếch đại tín hiệu định giá từ thị trường tín dụng thành thảm họa cho tài khoản nhà đầu tư cá nhân, với 62% người chịu lỗ là nhà đầu tư trẻ trong độ tuổi 20-30.
Original source: Wall Street Journal
In the past two years, the market defaulted to the assumption that someone would always pay the bill for AI infrastructure. That assumption was rejected by the credit market this Monday.
On July 28, South Korea's KOSPI fell 10.84%, triggering its 8th circuit breaker of the year. Nvidia fell 5%, SK hynix fell 7.47%, and Japan's Kioxia fell 18%.
On the same day, Apple rose 1.2%, reclaiming the title of the world's most valuable company. The Dow Jones Industrial Average rose 0.51%, and all three major European indices closed higher.
The same day, the same news cycle, completely opposite price movements. The market wasn't panicking — it was voting. The assets being sold off were all companies deeply tied to AI infrastructure, while the assets being bought were those not dependent on the AI infrastructure narrative. The signal triggering this vote didn't come from the stock market, but from the bond market.
Who is Saying No
On Monday, Nvidia's five-year CDS surged 14 basis points to 82 basis points in a single day, the largest increase for the contract. Oracle's CDS rose to 215 basis points, Alphabet's CDS rose to 67 basis points, and Broadcom, Amazon, and Meta all hit new records. SpaceX's new bonds issued in June broke below their issue price in the secondary market.
The pricing logic of CDS is completely different from stocks. Stock pricing looks at growth potential, narratives, and the future. CDS buyers are pension funds, insurance companies, and hedge funds. They don't read stories; they only read cash flow — the cash you generate now, whether it can cover the principal and interest due at maturity.
For the past two years, the pricing power of AI assets has been entirely in the hands of the equity market. The equity investor's assumption is: Capital expenditure equals a moat. Whoever has the larger capex wins. Every GPU is a down payment on future market capitalization. Under this narrative, the combined capital expenditure of the top five tech giants in 2026 is projected to be $737 billion, up from $261 billion two years ago.
Buyers in the credit market have remained silent because they don't trade "stories." They were just waiting.
The direct trigger for Monday's trading was Nvidia's July 25 announcement of a $500 billion AI infrastructure cooperation with SK Group, along with OpenAI's $250 billion financing guarantee and a $350 billion chip procurement financing arrangement. The potential exposure from OpenAI alone exceeds $600 billion, more than Thailand's annual GDP. The market took the weekend to digest the implications of this number — when markets opened on Monday, CDS jumped directly.
Before this, signals had been accumulating for three weeks. Oracle was downgraded by S&P from BBB to BBB- on July 9, just one notch above junk status. Alphabet's free cash flow turned negative by $5.855 billion in its most recent quarterly earnings report, the first time in its twenty years as a public company. When these events occurred individually, the market chose to digest rather than panic — Oracle's issue could be attributed to overly concentrated clients, Alphabet's to being still in an investment phase, each problem seen as isolated.
Nvidia's $600 billion revolving financing scheme connected the dots.Oracle's biggest client is OpenAI. OpenAI is the core beneficiary of Nvidia's revolving financing scheme. Alphabet, Meta, and Amazon all operate on the same "borrow to build infrastructure" financing model. These are not independent events. This is the same balance sheet, being re-hypothecated under different borrower names.
The credit market took three weeks to digest the signals, make connections, and reach a judgment. On Monday, the judgment landed.
How Credit Entered a Closed Loop
The traditional model of the semiconductor industry was unidirectional: Nvidia sells chips to cloud providers, cloud providers rent computing power to AI companies. Funds flow from end-users upwards layer by layer, and each layer bears its own credit risk.
Nvidia's current model turns this into a cycle.Nvidia provides credit to its clients → Clients 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, calls such arrangements "financial alchemy." John Flood, head of trading at Goldman Sachs, wrote in a client memo that many traders feel "uncomfortable" about this, pointing out that the credit market has become a key window for judging the current AI narrative.
This is not a problem unique to Nvidia. The entire financing model for AI infrastructure is built on an assumption — that the cost of capital approaches zero and the market is eternally willing to pay for the "future." When CDS starts pricing, this assumption is being tested for the first time.
Chain Reaction
The credit market's skepticism doesn't need to verify each company individually. It propagates down the financing chain by itself.
Oracle was the first to hit the line. S&P downgraded it from BBB to BBB- on July 9, citing three reasons directly: AI business cash burn exceeding expectations (FY2026 capex up 162% YoY to $56 billion, free cash flow negative $23.7 billion, projected to rise further to $95 billion in FY2027); overly high client concentration (OpenAI accounts for about half of its remaining performance obligations); and insufficient internal load capacity 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 insuring its bonds has doubled.
This risk points directly to OpenAI itself. According to Fortune and the Financial Times, in 2025, OpenAI had revenue of $13.07 billion but an operating loss of $21 billion. After including one-time expenses, its net loss was approximately $39 billion. A company losing nearly $40 billion a year is Oracle's biggest client and the core beneficiary of Nvidia's revolving financing scheme.
The chain continues. Alphabet's free cash flow turned negative. Meta's $12.55 billion bond for its Texas data center carried a 7.534% interest rate — a coupon typically seen only in the junk bond realm. The subscription multiple was only 1.6 times, less than half the average subscription multiple for US corporate bonds this year. The average borrowing cost for AA/A-rated companies in the entire tech sector rose from 3.9% a year ago to 5.1%, for BBB-rated companies from 4.7% to 7.0%, and unsecured loan rates for lower-rated AI-related companies have touched 10.5%-12%. Meanwhile, the capital expenditure guidance for the top five tech giants continues to be revised upwards.
This is not a problem of one single company. It is the entire financing model switching from "infinite cheapness" to "market pricing." Every level of spread widening transmits to the next level, the next borrower, the next market.
The Bill from 2000
This script isn't new.
In March 2000, Cisco became the world's largest company with a market cap of $555 billion. Its growth engine was "vendor financing" — lending money to unprofitable internet companies to buy Cisco equipment. At its peak, the exposure was around $4 billion, covering over 700 clients. John Chambers called it a "once-in-a-century revolution."
After the Nasdaq peaked, clients went bankrupt in droves.In May 2001, Cisco wrote off $2.2 billion in inventory and set aside $800 million for bad debts. Its stock price fell from $79 to $9.5. In December 2025, Cisco's stock price finally surpassed its 2000 highs, taking 25 years. However, due to continuous buybacks reducing share count, its current market cap is still over 40% lower than its 2000 peak — the stock price returned, but the company's size did not.
From 1996 to 2000, vendor financing was priced as "ecosystem empowerment" and "demand catalyst." After the Nasdaq crash in 2000, clients died en masse, and Cisco's bad debts surfaced — the same string of numbers was re-read as "when will you actually pay the money?"
The shift in logic is abrupt: before the shift, the equity market provided narrative legitimacy for vendor financing; after the shift, the narrative disappears, and the bill is just numbers. Jim Bianco, President of Bianco Research, concludes: before the credit market holds pricing power, equity valuations can always be debated; once the credit market starts pricing, the debate is over.
Cisco's exposure was $4 billion. Nvidia's potential arrangements related to OpenAI alone exceed $600 billion. The CDS market and debt buyers reacted much faster this time than in the previous cycle — it took Cisco 14 months from peak to write-off, while Nvidia's CDS began repricing just one trading day after the news was disclosed.
The Rewarded "Opt-Out"
Apple reclaiming the title of the world's most valuable company today is no coincidence.
Apple's capital expenditure for its latest quarter was only $1.97 billion, down 36% year-over-year. During the same period, Amazon set its full-year capex guidance at $200 billion, and Alphabet raised its to $205 billion. Apple's choice was to return cash to shareholders — it authorized an additional $100 billion in buybacks in the second quarter while also raising its dividend. Tim Cook's explanation was "maintaining a cautious and thoughtful approach to capital expenditure."
It's not that Apple isn't investing in AI. The A19 chip features a built-in neural accelerator, Apple Intelligence is embedded across its entire product line, and the new Siri is rebuilt based on Gemini. The difference lies in the path: on-device inference combined with external model access, and a refusal to build large-scale training clusters in-house.
From 2024 to 2025, Wall Street consistently criticized Apple for under-investing. The prevailing judgment at that time was that building in-house computing power equaled a moat.By the summer of 2026, the credit market reversed that same judgment — massive capex went from being a competitive advantage to a debt burden, and a disciplined balance sheet went from conservative to flexible. Jay Woods, Chief Global Strategist at Freedom Capital Markets, said: "Apple was previously criticized for insufficient AI investment, but looking at it now, it successfully avoided the capital expenditure trap."
This reversal was confirmed by prices on July 28. The market is not just punishing the over-leveraged side; it is also rewarding the side with a healthy balance sheet. These two events are two sides of the same coin.
The 8th Circuit Breaker of the Year
The revaluation by the credit market on Monday took only one day to reach the accounts of South Korean retail investors.
The 8th circuit breaker of the year for the KOSPI. Samsung and SK hynix account for 60% of the index weight, and these two stocks contributed approximately 70% of the KOSPI's gains over the past two years.
In May, regulators approved 16 2x leveraged ETFs linked to them. These funds ballooned from 4.9 trillion won to over $9 billion in two months, with retail investors holding 92%. On the way up, leverage amplified gains, attracted incremental capital, and pushed up the underlying stocks. On the way down, the same mechanism operates in reverse: underlying stocks fall → ETFs are forced to deleverage → underlying stocks fall further → margin calls → forced liquidations → continued decline. In the first two weeks of July, 1.2 million leveraged accounts received margin calls, and between 320,000 and 460,000 were fully liquidated. Retail investors aged 20-30 account for 62% of those suffering losses.
This group couldn't possibly have known what happened before the market opened on Monday. They didn't know about Nvidia's $600 billion revolving financing structure, the reasons behind Oracle's downgrade, or that the credit market was re-pricing AI assets using a framework completely different from their own. They only knew that Samsung and SK hynix were falling and that their leveraged accounts were receiving margin calls.
This is precisely the structural characteristic of the entire transmission chain: Information flows from professional credit market buyers 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 absorbs a greater shock than the one before. A 14 basis point jump in CDS on Monday translated into a circuit breaker on the KOSPI. In leveraged accounts, it translated into being wiped out.
What the credit market repriced on Monday wasn't just a single stock. It was an entire pricing framework that had taken "eternally cheap capital" as a default assumption. It recalculates from the root, and information travels down the leverage chain step by step, amplifying at each stage. Those who knew first took their sides; those who found out last paid the price.
