NVIDIA's earnings report from a bearish perspective: Is it an AI bubble or an AI revolution?

Author: danny; Source: X, @agintender

NVIDIA announced its Q3 financial report on November 19. While it can't be said that the results are outstanding, they can be considered better than expected. The problem is that the market did not respond positively to such results, starting to plummet after a 5% rise. Many in the crypto community are left bewildered. This article attempts to summarize, interpret, and analyze what hidden issues may lie within this seemingly “too good to be true” financial report from the perspective of the bears.

In addition, there are too many bullish articles, so I won't elaborate on that here.

If you are too lazy to read a long article, here are the core short-selling viewpoints, feel free to take them:

1. Cyclical Financing for Revenue Generation: Nvidia has built a capital recycling closed loop by investing in clients like xAI, turning investment funds into its own accounting income, lacking substantial cash delivery.
2. Abnormal Surge in Accounts Receivable: The accounts receivable balance reached $33.4 billion, with a growth rate far exceeding revenue, and the calculation of turnover days raises suspicions of concealment, indicating serious “channel stuffing” and backend loading phenomena.
3. Inventory and Narrative Divergence: Under the narrative of “supply not meeting demand,” the finished goods inventory unexpectedly doubled, indicating potential risks of customers delaying pickup or products being unsold.
4. Cash Flow Inversion: Operating cash flow is significantly lower than net profit, proving that the company's profits primarily remain on paper and have not been converted into actual cash.

This article does not constitute any investment advice. This article is merely a collection of opinions.

1. Recurring Revenue and Vendor Financing Model

1.1 Closed-loop mechanism of capital flow

Background: In November 2025, Elon Musk's xAI completed a $20 billion funding round, with Nvidia participating in about $2 billion of equity investment. However, this was not just a simple “investment behavior.” Let's follow the logic step by step:

Capital outflow (investment side): Nvidia allocated cash (approximately $2 billion) from its balance sheet, recorded as “Purchases of non-marketable equity securities,” as equity investment in xAI or related SPVs. This outflow of funds is reflected in the cash flow statement under the “investing activities” section.

Capital Conversion (Client): xAI receives the funds and uses them as a down payment or capital expenditure budget for the purchase of a GPU cluster (i.e., Colossus 2 project, involving 100,000 H100/H200 and Blackwell chips).

Capital Repatriation (Revenue Side): xAI subsequently issued a purchase order to Nvidia. Nvidia shipped and confirmed “data center revenue.”

Financial results: Nvidia has actually converted the “cash” assets on its balance sheet into “revenue” and “net profit” on its income statement through xAI as an intermediary.

Although this operation is generally allowable under accounting standards (GAAP) (as long as it is assessed), it is actually a form of “Low-Quality Revenue” (IFRS here indicates discontent, only seeking a battle?)

This is also criticized by bears like Michael Burry, as this model of “almost all customers being funded by their suppliers” is a typical characteristic of the late stage of a bubble. When a company's revenue growth relies on its own balance sheet expansion, once it stops investing externally, its revenue growth will also dry up. (Does it feel a bit like a crypto circle?)

The Leverage Effect and Risk Isolation of 1.2 SPV

If you find the circular income model a bit stunning, then the special purpose vehicle (SPV) structure involved in trading might really open your eyes even wider.

According to news reports, xAI's financing includes equity and debt, with the debt portion constructed through an SPV, which is primarily used to purchase Nvidia processors and lease them to xAI.

The operating logic of SPV: SPV, as a legally independent entity, holds GPU assets. Nvidia is not only a seller of GPUs but also an equity investor in SPV (First-loss capital provider). This means that Nvidia plays a dual role in the transaction: as a supplier and as an underwriter.

Revenue recognition in the circular arbitrage model: By selling hardware to the SPV, Nvidia can immediately recognize the full sales revenue from the hardware. However, for the end user xAI, this is essentially a long-term lease (Operating Lease), with cash outflows occurring in installments (e.g., over a 5-year period).

Risk Concealment: This structure converts long-term credit risk (whether xAI can pay rent in the future) into immediate revenue recognition. If the price of AI computing power crashes in the future, or if xAI cannot generate sufficient cash flow to pay rent, the SPV will face default, and Nvidia, as the equity holder of the SPV, will face asset write-down risk. However, in the current earnings report season, all of this appears as the glamorous “Genesis Revenue.”

Shadow of vendor financing during the Internet bubble period 1.3

The current business model is somewhat similar to the internet bubble of 2000. At that time, Lucent lent billions of dollars to customers to purchase its own equipment. When internet traffic growth fell short of expectations, these startups defaulted, and Lucent was forced to write off massive bad debts, with its stock price crashing by 99%.

Nvidia's current risk exposure (direct investments + SPV debt support) is estimated to have exceeded $110 billion, representing a significant proportion of its annual revenue. Although Nvidia does not currently list “customer loans” directly on its balance sheet, its substantive risk exposure is consistent through holding customer equity and SPV interests.

2. Accounts Receivable Suspicion

2.1 Accounts receivable ratio growth “swift”

According to the Q3 FY2026 financial report, Nvidia's accounts receivable balance reached 33.4 billion USD.

The year-on-year growth rate of accounts receivable (224%) is 3.6 times that of the revenue growth rate (62%). In normal business logic, accounts receivable should grow in sync with revenue, especially since Nvidia is so “strong”. When the growth rate of accounts receivable far exceeds that of revenue, it usually indicates two possibilities:

a. Decline in income quality: The company relaxed credit terms, allowing customers to defer payments to stimulate sales.

b. Channel stuffing: The company rushes to ship products to distributors at the end of the quarter to recognize revenue, but this portion of the product has not truly been absorbed by the end market. (This will be discussed further later)

2.2 DSO (Days Sales Outstanding) algorithm

The DSO for this quarter is 53 days, a slight decrease from 54 days in the previous quarter. So what is the actual situation?

First, the standard DSO calculation formula: DSO = ( accounts receivable / total credit sales ) x number of days in the period.

Beginning AR ( Q2 end ): 23.065 billion USD

Final AR ( Q3 end ): 33.391 billion USD

Average AR: $28.228 billion ( (Q2+Q3)/2)

Quarterly revenue: 57.006 billion USD

Days: 90 days

The standard DSO is approximately 282.28 / 570.06 *90 = 44.566 ( days)

However, the reported DSO is 53 days. Generally speaking, from the perspective of “window dressing” the reports, one would usually report a more “aggressive” number, yet here it is conservative? This suggests that Nvidia may be using the accounts receivable at the end of the period as the numerator, or its calculation logic is more inclined to reflect the end-period capital occupation situation.

If calculated using the final balance:

333.91 / 570.06 *90 = 52.717 (days)

The numbers here align with the report. But what does this mean? It means that the accounts receivable balance at the end of the quarter is extremely high relative to the total sales for the entire quarter. This suggests the phenomenon of Back-end Loading, where a large amount of sales activity occurs in the last month or even the last week of the quarter.

If sales are evenly distributed, the accounts receivable at the end of the period should only include the sales of the last month (approximately $19 billion). However, the current balance is $33.4 billion, indicating that nearly 58% of the quarterly revenue has not been received in cash.

In the narrative of a so-called “seller's market” and “supply not meeting demand,” Nvidia should have strong bargaining power, even demanding advance payments. However, the reality is that Nvidia not only did not receive advance payments but instead offered customers payment terms of nearly two months?! This seems to be inconsistent with the narrative of “panic buying”?!

3. Inventory Puzzle: The Paradox of Supply and Demand vs. Inventory Backlog

While Jensen shouts “Blackwell demand is off the charts,” Nvidia's inventory data seems to tell a different story.

The reason for inventory doubling 3.1

The total inventory for Q3 FY2026 reached $19.8 billion, nearly doubling from $10 billion at the beginning of the year and increasing by 32% from $15 billion in the previous quarter.

More importantly, the composition of the inventory:

Raw Materials (: 4.2 billion USD

Work in Process, WIP: $8.7 billion

Finished Goods ): 6.8 billion USD

In a word, finished goods inventory has surged. At the beginning of 2025, finished goods inventory was only $3.2 billion. Now it has skyrocketed to $6.8 billion. Especially when Lao Huang is shouting that demand is crazy to the point of explosion, under the assumption of chip shortages and customers queuing for goods, finished goods should be “produced and shipped immediately”, and inventory levels should be kept at extremely low levels.

Why is that? Are you going to collect the debt after the New Year?

( 3.2 $50 billion procurement commitment

In addition to the inventory on the balance sheet, Nvidia also disclosed supply-related commitments of up to $50.3 billion. This is the amount Nvidia has committed to future purchases from suppliers such as TSMC and Micron.

This is a huge “hidden danger.” If AI demand slows down or weakens in any form over the next few quarters, Nvidia will face a double blow:

1. Inventory Impairment: The existing $19.8 billion inventory may depreciate.
2. Breach of Contract or Forced Procurement: A $50 billion procurement contract, leading to more inventory backlog, or payment of hefty penalties for breach of contract.

The emergence of this “heavy asset” characteristic signifies that Nvidia is no longer the lightweight chip designer it once was, but is increasingly resembling a hardware manufacturer burdened with a heavy supply chain.

Production capacity is ahead, while inventory is lagging behind, is this a disconnection from reality?

4. Profits increased, but cash flow actually decreased?

) 4.1 The inversion of operating cash flow (OCF) and net profit

In general, a healthy tech company's operating cash flow should be higher than its net profit (as depreciation, amortization, and equity incentives are non-cash expenses that are added back). However, Nvidia's data shows the opposite trend.

Q3 Net Income (: 31.9 billion USD

Working Capital Changes ):

Accounts receivable increase leads to cash outflow: - $5.58 billion

Inventory increase leads to cash outflow: - 4.82 billion USD

Q3 Operating Cash Flow ###OCF###: Approximately 23.75 billion USD

Conclusion: The operating cash flow in Q3 is significantly lower than the net profit. For every dollar of profit, only about 0.74 dollars has actually been converted into cash inflow, while the rest has turned into chips in the warehouse (inventory) and receivables from customers (accounts receivable).

Of course, the phenomenon of OCF < Net Income depends on how you interpret it. It could mean that the company's profits are recognized by accounting standards rather than being supported by real cash in the bank account; it could also mean that the company is experiencing rapid growth.

( 4.2 Cash Outflow in Investment Activities

Purchases of non-marketable equity securities ) —— commonly known as investments: This quarter saw an outflow of 3.7 billion USD.

This 3.7 billion dollars is flowing to ecosystem partners such as xAI, CoreWeave, and Hugging Face. In comparison, this figure was only 473 million dollars during the same period last year. Nvidia is ramping up its buyouts of the ecosystem at a SpaceX-like speed.

From the perspective of xAI's model, the investment process may be as follows:

1. Nvidia accumulates cash through bond issuance or previous profits.
2. Invest cash in startups (cash outflow)
3. The startup uses this money to buy chips (confirmed as revenue)
4. Nvidia's book profits increase, stock prices rise, attracting talent through equity incentives, and then financing through bond issuance or additional share offerings (Doesn't it feel a bit like the crypto space's nested structures?)

If this really is the mode, it feels a bit like musical chairs. Of course, as long as the music doesn't stop, this game can go on forever. But once the financing environment tightens (such as rising interest rates or a collapse of the AI bubble), this game could come to a sudden stop.

5. Nvidia's dominance is not sacred and inviolable

In the 10-Q filing, Nvidia disclosed a very high customer concentration, with “Customer A” accounting for 22%. Although not named, there are only a few wealthy companies on this planet, and it can almost be certain, definitely, 100% that it is Microsoft.

There is another layer of risk related to “related party transactions” hidden here. Microsoft is OpenAI's largest financier, and Nvidia has also invested in OpenAI. A significant portion of Microsoft’s purchase of Nvidia chips should be for OpenAI's use. Both are shareholders, but the profit-sharing details here are unknown. Who knows if there are any special clauses?

What if Microsoft decides not to buy today?

In addition, the presence of Customer B (15%), C (13%), and D (11%) means that the top four customers control Nvidia's lifeblood. This level of concentration means that Nvidia does not have the absolute dominance in pricing negotiations that the outside world imagines. On the contrary, these giants are using their massive purchasing power to force Nvidia to make concessions in areas such as supply chain allocation and custom chip design, and they are even accelerating the research and development of self-developed chips (such as Google TPU, AWS Trainium, Meta MTIA) to reduce their dependence on Nvidia. This is also reflected in the increase in accounts receivable.

The following diagram gives you an intuitive display of the complex structure of the OPEN AI cluster. Let me ask you, can you make sense of the account?

Afterword

Why write this article? First of all, it has been a long time since I wrote this type of report research long article, and I want to revisit the feeling of reading reports.

Secondly, there is a sentiment or logic in the market now where the crypto market looks at the U.S. stock market, the U.S. stock market looks at the AI revolution, and AI looks at Nvidia's performance. Although Nvidia's earnings report exceeded expectations, there are still many bearish viewpoints.

Compared to other seemingly plausible bearish views, honestly looking for clues in financial reports has empirical discussion significance.

It's been a long time since I wrote a report analysis like this, consider it another perspective for everyone to look at the current market environment.