Nvidia’s stock plunges 5%, wiping out $250 billion in market value: Why did “round financing” concerns pierce the AI chip faith?

On July 28, 2026 (Beijing time), U.S. stocks’ semiconductor sector was hit by an unexpected sell-off. NVIDIA’s share price closed at $196.51, down 4.99% on the day, marking the biggest single-day drop since June 5. The market value evaporated by about $250 billion in a single day, falling to $4.76 trillion. During the trading session, the stock briefly dipped as low as $195.44, down more than $11 from the prior trading day’s close of $206.84.

The reason behind this sell-off is precisely the “good news” announced by NVIDIA itself.

Over the past week, NVIDIA has rolled out three major collaborations in succession: an AI infrastructure partnership worth more than $500 billion with South Korea’s SK Group; discussions to provide up to $250 billion in financing guarantees for OpenAI to help it rent a 10 GW super data center developed in Ohio by SoftBank; and additionally, NVIDIA is also separately negotiating with OpenAI a chip procurement financing deal worth up to $350 billion. The total across the three deals exceeds $750 billion—about 3.5 times NVIDIA’s revenue for fiscal year 2026.

However, the market’s response was a full-blown sell-off. Not only NVIDIA: the Philadelphia Semiconductor Index fell 2.23% on the day, and at one point during trading it dropped nearly 5%. SanDisk plunged 11.02%, SK Hynix fell 7.47%, AMD dropped 5.17%, and ASML fell 5.8%. Apple, by contrast, rose 1.17% to close at $336.91, lifting its market value to about $4.95 trillion—marking the first time since April 2025 that its closing market cap surpassed NVIDIA’s, reclaiming the top spot globally.

The essence of this sell-off is not that the market suddenly stopped believing in AI, but that investors began questioning a more fundamental issue: why would a chip-selling company use its own credit to guarantee purchases of its own chips for customers?

A crack in the AI capital expenditure logic

To understand this sell-off, you need to first look back at the market’s valuation logic for NVIDIA over the past few years.

Between late 2022 and 2025, driven by a surge in GPU demand, NVIDIA’s stock price soared by more than 1,100%. The logic chain was clear and concise: AI demand grows → cloud providers expand data centers → buy NVIDIA GPUs → NVIDIA revenue rises → the stock price increases. Contract balances held by ultra-large-scale cloud service providers were as high as $1.45 trillion, making it seem like the situation of compute supply lagging demand had not changed.

However, since 2026, that logic chain has started to show cracks. NVIDIA’s year-to-date gains have narrowed to about 4% to 5.4%, significantly underperforming the S&P 500, which is up about 8%.

The market’s questions are shifting from “is there demand for AI?” to “where does the money behind AI demand come from?” While chip demand remains strong, customers’ procurement is increasingly relying on financing support from NVIDIA itself—meaning the order size the market sees may not fully reflect customers’ true ability to pay. Whether AI revenue growth is sustainable depends on whether these money-burning companies can ultimately generate positive cash flow, not merely whether they can keep securing financing.

This is the core of the “circular financing” controversy.

Every dollar in the loop

The structure of “circular financing” can be broken down into three layers:

First layer: NVIDIA provides $250 billion in financing guarantees to OpenAI. OpenAI currently does not have an investment-grade credit rating—it is still burning cash. With NVIDIA’s credit backing, lenders are willing to fund data center projects at a lower cost.

Second layer: The data centers are developed by SB Energy, a unit of SoftBank, in Picton, Ohio, with a planned capacity of 10 GW—enough electricity for 8 million U.S. households. Total investment is expected to exceed $500 billion.

Third layer: NVIDIA is also separately discussing up to $350 billion in chip procurement financing with OpenAI. Combined with the $250 billion guarantee, NVIDIA’s exposure to this single customer could reach $600 billion—while NVIDIA’s own annual revenue is about $216 billion.

The path of the closed loop is: NVIDIA guarantees financing → SoftBank builds data centers → OpenAI rents compute → OpenAI uses the guaranteed funds to buy NVIDIA chips → NVIDIA recognizes revenue → then guarantees more projects again.

The market’s concern is that every dollar of AI investment is circling within the same closed loop. If OpenAI fails to burn capital into products that generate sufficient cash flow, who ultimately bears the credit risk? Is it NVIDIA, the guarantor, or SoftBank, the lender to OpenAI, or financial institutions holding related bonds?

Michael Burry, a famous investor for profiting from subprime mortgages, wrote only one sentence on social platform X: “Turn it, turn it, turn it. NVIDIA needs to provide a guarantee for $47.6k that it will spend on ChatGPT using NVIDIA chips.” Known short-seller Jim Chanos was even more direct: “We’ve reached a stage in this cycle—does NVIDIA really have to provide financing guarantees covering two-thirds of the cost for the chips it sells to data centers?!”

Gary Tan, a portfolio manager at Allspring Global Investments, was more restrained: “Although NVIDIA’s investments and collaborations increase investors’ confidence in long-term AI infrastructure buildout, investors remain concerned about circular financing. An increasing amount of capital is being used to finance future AI customers and infrastructure deployments.”

Credit markets sounding the alarm

Stock markets can chase growth stories, but bond markets only care about cash flow and debt repayment ability.

On July 27 (Beijing time), NVIDIA’s five-year credit default swap (CDS) spread surged by 14 basis points in a single day to 82 basis points—its largest single-day increase since the contract began trading in November 2025. To buy five-year default protection on $10 million worth of NVIDIA debt, you would have to pay about $82k per year.

In essence, CDS is like purchasing insurance for corporate bond debt. A higher spread means the market believes the risk of holding the related bonds is greater. NVIDIA’s rapid CDS jump reflects investors starting to assess the potential pressure that NVIDIA’s large-scale support for AI customers and data center construction could impose on its own balance sheet.

More importantly, this repricing of credit risk is not unique to NVIDIA. According to LSEG data, CDS prices for companies including Oracle, SpaceX, Alphabet, Amazon, Meta, and Broadcom have all risen to record highs in recent days. This highlights growing unease in the market about large technology companies’ massive spending on data centers, chips, and computer memory.

John Flood, a trader at Goldman Sachs, noted that AI capital expenditure had previously been a tailwind for the market, but investors are now beginning to question it—credit risk is the focus right now. A view from an analyst at Société Générale is being validated by the market: “Now you look at CDS, not EPS.” The AI narrative is shifting from revenue growth to credit risk.

What the market is repricing

The sell-off on July 27 was not just simple panic about an AI bubble—it was a systematic repricing of the financing structure.

Several structural factors are accelerating this process.

NVIDIA’s own debt expansion. In June 2026, NVIDIA issued $25 billion of corporate bonds again after five years to fund the above investment and guarantee plans. This means NVIDIA is not only backing customers’ procurement with equity value—it is also concretely increasing its own leverage.

Industry-wide ballooning of capital expenditures. S&P estimates that the combined 2026 capital expenditures of Alphabet, Amazon, Meta, Microsoft, and Oracle could reach $737 billion, far above $261 billion in 2024. When the entire industry burns cash at unprecedented speed, the question of when cash flow will turn positive becomes unavoidable.

A shift in relative valuation. The Nasdaq 100 Index’s forward price-to-earnings ratio is currently about 21.8x, versus its past 10-year average forward P/E of about 23.6x—discounted by nearly 10%, and at the lowest valuation level since the beginning of the AI boom cycle in early 2023. NVIDIA’s current P/E is about 30x. The market is reassessing: behind these numbers, how much is truly demand, and how much is inflated orders driven by financing?

Conclusion

On July 27, NVIDIA erased $250 billion in market value, drawing a clear line for the market.

On one side of the line is the old logic: “AI demand growth → buying GPUs → revenue growth → stock price rises.” On the other side is the new reality: “Chip demand depends on financing support → customer cash flows are questionable → revenue quality faces a major reassessment.”

NVIDIA’s fundamentals remain strong. Second-quarter net profit was $26.42 billion, up 59.2% year over year. Among the 12-month target prices for the future held by 64 analysts, the median is $300. But these numbers are no longer the only variables the market is focused on.

When a company is not only a supplier, but also the customer’s financing provider, guarantor, and infrastructure investor, the market asks with two sets of language at the same time: the stock market asks “how fast is growth?”, while the bond market asks “how big is the risk?” On July 27, the latter’s voice for the first time drowned out the former.

This is not the end of the AI story. It is the beginning of the AI story moving from a “compute narrative” into a “ledger narrative.”

FAQ

Q1: What is “circular financing”? Why is the market worried about it?

“Circular financing” refers to a structure where a supplier also acts as the customer’s financing party or guarantor, forming a funds-closure transaction model. Taking NVIDIA as an example: it provides a financing guarantee to OpenAI, OpenAI uses that money to buy NVIDIA chips, and after NVIDIA recognizes revenue, it then provides guarantees for more projects. The market is concerned that this model artificially boosts demand and valuations; if AI revenue falls short of expectations, credit risk will be transmitted and amplified along the loop.

Q2: What does the surge in NVIDIA’s CDS imply?

CDS (credit default swap) is effectively a “default insurance” purchased for corporate debt. When the CDS spread rises, it means the market believes the risk of holding that company’s bonds has increased. NVIDIA’s five-year CDS surged 14 basis points in a day to 82 basis points, the largest increase since the contract began trading. This reflects investors beginning to worry that NVIDIA’s large-scale customer financing could damage its own balance sheet.

Q3: How do analysts view NVIDIA’s outlook?

Based on data covering 64 analysts, NVIDIA’s median 12-month target price is $300. Bank of America maintains a “Buy” rating with a target price of $350; Goldman Sachs has a target price of $210. The consensus rating is “Strong Buy”: 36 recommend buying, and 1 recommends holding. But in the short term, the stock price is still affected by uncertainty around the circular financing controversy and the pace of AI capital expenditures.