3 months ago, in the article “CHEF’S TABLE: NVIDIA,” I already highlighted how Nvidia’s sophistication in cooking its books reached a remarkably high level. At that time, thanks to fresh disclosure in its report, not only did we manage to figure out the whole structure Nvidia, with the support of Dell and SMCI, is using to circumvent US export restrictions via Singapore. But we also managed to figure out how much Nvidia, by its own implicit admission in the risk disclosures, is dependent on OpenAI to generate a significant portion of its revenues (both directly and indirectly).
When the company reported its latest quarterly results after US market close on Wednesday, I admit that initially, reading the whole 65-page 10-Q, I could not spot any new elements or missing elements within the numbers or disclosures worth noticing beyond all the red flags and shenanigans I had already analyzed and presented in the long series of articles I wrote on Nvidia (full archive here).
Reading Nvidia’s earnings report always felt like walking through a thick fog where all you could easily observe was intentionally placed by the company under your nose and, of course, it made the company look great. Uncomfortable data or disclosures were always placed “far into the fog” (for example, using intricate attorney terminology) so the company could claim they did not hide information from investors while practically making it hard to see. However, reading Nvidia’s report this last time felt truly smooth, with the company (surprisingly) clearly describing all the regulatory and geopolitical issues it is going through and even clearly warning about the “remote” risks of the demand for its GPUs coming to an abrupt halt. The numbers, as well, were reported exactly where they were expected to come out. Too much of a Wall Street analysts’ expectations being beaten, when even Sam Altman is now openly warning about a Bubble in AI (”OpenAI’s Sam Altman says AI market is in a bubble”), would have surely raised eyebrows and prompted a lot of scrutiny. A Wall Street analysts’ expectations miss, even if it could be easily justified considering all the mounting headwinds against Nvidia, would have jolted markets and sent Nvidia’s stock price through the floor along with the stock prices and valuations of all its partners (in crime). As I write, Nvidia’s share price is trading just roughly 1% lower on the day, and it feels like markets are taking a sigh of relief thanks to the release being smoothly digested without causing any volatility.
This whole perfection and openness by the company felt too good to be true, similar to when a kid you are used to seeing misbehave suddenly becomes quiet and gentle. In those cases, more often than not, the kid is in reality hoping to hide his mischief by behaving well. As a consequence, I started to screen through the report, paying even more attention than usual to any detail that I might have missed until I noticed something very subtle but interesting in this table.
What am I exactly referring to? Allow me to guide you here:
- First of all, what is the definition of “deferred revenue”? Deferred revenue (also known as “unearned revenue”) is a liability on a company’s balance sheet that represents payment received for goods or services that have not yet been delivered or performed. In simpler terms, it’s money collected in advance for a product or service the company still owes to the customer. Because the company has an obligation to provide something in the future, it is recorded as a liability, not as earned revenue.
- Here is the very interesting thing: the production and delivery cycle of Nvidia GPUs is fairly long, surely longer than a quarter, especially when the company continues claiming “overwhelming demand” while they have to deal with limited production capacity. If this is the case, how is it possible for Nvidia to collect very large cash advances from customers and fulfill those obligations in full within the same quarter?
- Now, please take a second look at the table. While in the past six months, Nvidia’s deferred revenues account has only increased by ~200 million USD, within the 6 months Nvidia added ~8 billion USD of deferred revenues and then immediately recognized a similar amount in the same time period.
Wait a second, was this just an exceptional case? Please take a look at the second table here from the previous quarter.
Clearly, there is something wrong going on here, so I went back many quarters to check this out.
How to interpret this chart?
- First of all, you can notice how Nvidia didn’t collect a significant amount of cash advances in the past, but this amount started to grow significantly quarter after quarter (bear in mind the figure in the chart is on a trailing basis during the year). Is the company worried about the ability of its clients to fulfill their obligations, so it requires a cash advance that, if things go south, can at least cover its production costs? On the other hand, though, isn’t it unusual for a company that isn’t directly the producer but externalizes its production to third parties to ask for significant cash advances when its manufacturing partners don’t, and the company even has very flexible contracts to cancel production with minimal penalties? More on this soon.
- Secondly, how is it possible that Nvidia can fulfill its obligations so quickly despite all production delays and bottlenecks? As you can observe in the chart, the company always managed to recognize deferred revenues right away. However, while there is a possibility this can be done for relatively small orders (although it’s hard to imagine how a small order can jump the queue and be in front of other massive Nvidia clients), it’s objectively very difficult to do it for very large orders unless Nvidia has a massive ready-to-ship inventory and the delivery can be completed within the quarter (which is not the case by far).
Here is what I suspect is happening here: Nvidia is experiencing a rise in cancellations of future orders (surely from Chinese customers, but likely from US customers too, now more and more deploying their internally developed GPUs in their own infrastructure) and is filling the growing revenue shortfall by anticipating the recognition of cash advances as revenue. The big question, though, is how it was possible for Nvidia to collect 6.2 billion USD in cash advances in Q1-25? It’s hard to imagine they could come from the likes of Microsoft or Meta since these large clients did not pay big cash advances in the past despite their very large purchases. Any ideas here? Frankly, I am not sure which company could deliver that chunk of money for advance payments to Nvidia in the last quarter, but I am sure that those revenues have been incredibly convenient to cover the 5.5 billion USD impairment the company had to record for being restrained from selling H20s GPUs to China-based customers.
Where that chunk of cash advance came from and how Nvidia managed to recognize it as revenues within the quarter is a good question someone should ask Jensen Huang. In the current quarter, ~1.5 billion USD of revenues have been recorded using this method, and this rather small amount was enough to “beat” Wall Street revenue expectations set for 46.06 billion USD while Nvidia reported 46.74 billion USD.
While digging into Nvidia’s deferred revenues and wondering how the company could be so fast in fulfilling its obligations despite overwhelming demand and a long queue of customers waiting for their orders to be delivered, I took a look at Nvidia’s prepaid supply and capacity agreements.
Here is what the company reported in the last quarter
And here is the chart combining all the info on supply and capacity agreements accounted between long-term and short-term assets for all the most recent quarters
Personally, I find it incredibly fascinating how Nvidia could grow its revenues significantly for many quarters while the contractual amount of prepaid supply and capacity with its suppliers has been in a downtrend since Q3-2024. Logically speaking, if a company has such an “overwhelming demand” for its products, shouldn’t it be seeking to increase the amount of capacity with its production partners? Yes, this would be the case, but hey! Logic and common sense are something Nvidia investors and analysts forgot to apply long ago, which is why very few are paying attention to all the red flags and warnings that continue to pile up all around Nvidia.
JustDario on X | JustDario on Instagram | JustDario on YouTube
SYNNAX SUMMER PROMO
Get exclusive intelligence on stocks you can’t find anywhere else – insights that could give you a serious edge. Claim your free Synnax trial month now: https://synnax.app/PROMO/JUSTDARIO
If you love what we built at Synnax (and I think you will), this advanced market intelligence will cost you less than a cup of coffee a month for the next 12 months. After the 30th of September annual subscription price will be 154.99$ – redeem your free access now and give it a try.