A perfect score and an empty chair


Hi Reader,

NVIDIA reports its Q3 tomorrow, and I'd argue it's the strangest setup I can remember for a company of this size. The business scores 100 out of 100 on MonkScore (it ranks first among the 67 US semiconductor companies we cover), the stock is around 36% below its 52-week high, and Jensen Huang, who founded the company in 1993, is no longer there. All three things are true at the same time, which is something that doesn't happen very often.

I'll go over four things in this email:

  1. What the numbers say (and why the score is what it is)
  2. What the numbers don't say
  3. What I'll be looking for in tomorrow's call
  4. What I am doing with my shares

Without further ado, let's get to it.

The numbers

Let's start with the good news, because there's quite a bit of it. NVIDIA's gross margin over the last twelve months sits at 74.7% against a peer median of 50%, and its return on invested capital is 96.3% against a peer median of 12.7%. Revenue has grown at 110% a year over the past three years, whereas the median peer has grown at 6.7%. I don't think one needs to be a semiconductor expert to understand that these are not the numbers of a normal company, and this is why the Profitability and Growth pillars both sit at 100.

Last quarter was no different. Revenue came in at $96 billion (more than double the prior year) and management guided to $108 billion for Q3 and to around 70% growth for the next fiscal year. They also claimed that the limit to growth is supply and not demand, which is a pretty good problem to have (if true).

What the numbers don't say

Now, a score of 100 doesn't mean that there's nothing to worry about, and I believe NVIDIA is a great example of this. The fifth pillar, Market Conviction, sits at 65, which is 30 points below its three-year trail. In other words, the business kept improving while the market became way more skeptical. The market can be wrong (it wouldn't be the first time), but I think there are reasons behind the skepticism that are worth understanding.

The first is cash. Operating cash flow was 0.70 times net income over the last twelve months, against a peer median of 1.06, and customers now take 60 days to pay. This is why MonkStreet flags accruals as elevated. It doesn't mean that the profits aren't real, but it does mean that a growing portion of them has not yet arrived as cash.

The second is where part of the demand comes from. Management said on the last call that NVIDIA had invested nearly $50 billion in AI labs and that the labs it supports should be roughly a quarter of next year's business. An analyst added up around $500 billion of commitments and asked for confirmation, which he didn't get. Colette Kress was pretty candid about how this looks from the outside:

"We recognize the scale of this support, and we know some will call this circular financing. We see it differently."
— Colette Kress, CFO (Q2 FY2027 earnings call)

Of the 8 questions analysts asked, management fully answered 2 (I'll let you guess if those were the easy ones).

The third is evidently the founder, and this is the one that I think no number can capture. The board accepted Jensen's resignation on November 11 while an independent investigation was open, and named Colette Kress interim CEO (she continues as CFO). Now, Jensen was not a CEO in the conventional sense. He had 36 direct reports and an opinion (almost always the right one) on pretty much every detail. He could go deep into the technical design of a chip and, the next moment, come up with a customer financing structure that nobody had seen before. The management team is everything in a company like this one, and I believe Jensen is irreplaceable.

There are also plenty of rumors flying around. The board has reportedly approached outside candidates, quite a few people are asking for Jensen to come back, and there's even talk of him joining Elon Musk to run Terafab. I have no way of knowing which of these (if any) is true, but I do know that the last one would be terrible for NVIDIA. Losing an irreplaceable founder is one thing, but seeing him build chips for somebody else is quite another.

In the meantime, the person who signed off on the customer financing now runs the company and its finances while the board searches for a successor. This doesn't have to end badly, but it's the main reason why our Qualitative View reads "Mixed" despite the perfect score. The moat (CUDA, 7.5 million developers, customers buying the full rack) hasn't gone anywhere in a week. What has changed is the risk, and I'd be lying if I said otherwise: I liked NVIDIA considerably more a week ago than I do today. There's now more risk, more volatility and more uncertainty for the same (or a lower) expected return.

What I'll be looking for tomorrow

Analysts expect EPS of $9.30 this fiscal year and $15.75 the next. Some napkin math: at $156 this would be around 17x and 10x earnings, for a company growing at these rates. Either the estimates are too high or the price is too low, and I'd be lying if I told you I know which one it is. What I do know is that 60 of 75 analysts still rate it a buy with a median target of $300, even after six of them downgraded the stock in a single day last week.

Revenue will most likely not be what matters tomorrow. I'll be looking at three other things: whether days to get paid rise above 60, whether gross margin holds the 74% guide before dropping to 71%-72% next quarter, and whether Vera Rubin gets to the 20% of data center revenue that management promised. A fourth would be for somebody to finally clear out all the succession drama, but I wouldn't hold my breath.

What I am doing with my shares

I should be transparent here. I've owned NVIDIA for 9 years and I have not sold a single share in all that time, and that includes last week (I added more). This evidently means that I am biased (nobody holds a stock for 9 years without getting somewhat attached to it), so take everything above with a pinch of salt.

Now, 9 years is long enough to have seen this stock drop quite a few times, and what I've learned is that the drops were never the difficult part. The difficult part was telling apart the drops where only the price had changed from the ones where the business had changed. I believe this one is different to the others I've lived through for a simple reason: for the first time, one of the things I was relying on (the founder) is gone, and no number can tell me what that's worth.

This is ultimately why I'm not reading the 100 as a reason to relax, and why I'm not reading the "Mixed" as a reason to sell. Tomorrow I'll look at the three numbers I mentioned above and I'll go from there. Time will tell if holding was the right call.

MonkStreet doesn't give investment advice, and what I do with my shares shouldn't dictate what you do with yours. If you want to go through the four views (quantitative, qualitative, management and analysts) before the call, you can do so with the free 15-day trial (no card needed) here.

Happy investing,

Alberto

The Investing Monk Weekly

Gain access to data-driven insights, investing resources, and science-backed stock picks, delivered straight to your inbox.

Read more from The Investing Monk Weekly

Earnings are an opinion. Cash is a fact. In a market that moves trillions every quarter on one number, almost nobody asks whether the number is real. In 1957, an economist named Joseph Berliner published a book about Soviet factory managers. He’d spent the previous decade interviewing emigrants who had run plants under the central planning system. What he found was a system addicted to a single number. Factories were graded on output (tons of steel, square meters of glass, units of...