NVDA misses topline

Investment Analysis Clubs Saul’s Investing Discussions
21

I don’t think they’re missing topline estimates next Q by $700m because Crypto is lower than 3 months ago.

I am in NVDA for DC and AV and see gaming as a long-term lower-sloping uptrend that will be a cash cow to fuel R&D for the other lines of business that should overtake gaming as the top 2 segments in a few years. DC sooner than AV though.

DC is already a $3b/yr business growing 50%+y/y, so much bigger than a NTNX or PSTG, as an example, and growing faster.

The miss, as I understood it, is this:
Crypto craze causes price to go up, sucks supply out of market.
The “channel” probably consists of wholesalers and retailers and they are each marking up the heck out of these things due to crypto demand on top of strong gaming demand.

Then crypto demand plummets. But many retailers (my guess here) either had too high of a cost, or were too stubborn, to quickly lower prices back to normal levels to get inventory moving.

NVDA mgmt, in a rare misstep (although in fairness, no one ever dealt with a crypto craze before) continued making midrange GTX cards and added more supply to the market, as they figured the supply/demand would work itself out, but it appears prices weren’t lowered quick as they thought they would, so too much supply.

Hardcore gamers aren’t stupid, and know a new release (RTX) is liking coming down the pipe, and so rather than upgrade from their 980GTX or whatever, and pay the crypto-created higher-prices, they decide to wait, further exacerbating the supply issue.

So NVDA mgmt has decided they won’t introduce new GTX supply in Q4, to allow inventory to move, and to focus efforts on establishing the new RTX cards (which will creating a self-serving loop of more ray-tracing games being made by developers) in the marketplace.

So they appear to be writing off, essentially, the Q4 midrange desktop gaming number.
CEO Huang reiterated on the call numerous times that the DC market is strong and he seems no change coming there. DC and AV just hit ATH’s in revenue.

Crypto was a desktop-specific GPU issue, and notebook sales are not affected, and Huang called out 50%y/y growth in notebooks in China, as an example.

I just don’t see the thesis broken. All I see is a bunch of millenials (probably) in their parents’ basements lamenting the demise of their get-rich-quick crypto plans.

Holding onto NVDA for now.

Dreamer

27 Likes
22

When GPU prices are 30% lower than January this year, you have to sell 40% more units just to keep the same revenue amount, not to mention lower GP. I am surprised it took this long to disappoint when prices peaked in January.

You may be right, but I think for the wrong reason.
The issue appears to be supply/inventory.

I do not think NVIDIA raised prices. They have a list or perhaps a negotiated discount even for large resellers in the channel. It is the ultimate store or website point of sale retailer that is/was jacking up the prices, which NVIDIA spoke out against. Example here:
https://www.digitaltrends.com/computing/cryptocurrency-minin…

Gamers stopped buying inflated GPU prices and instead decided to wait. Once crypto miners stopped buying, NVDA mgmt expected the prices would drop, but they didn’t drop fast enough to move the excess inventory that flooded the market. (you also have to account for AMD inventory here too, which NVDA can’t measure as well, as they called out on CC today)

To compound the problem, NVDA announces new line RTX…so more gamers wait.

To fix this short-term problem, NVIDIA won’t produce more midrange GTX cards to allow channel inventory to flush out and so they can also just concentrate on building up RTX user base.

I didn’t buy into NVDA in 2016 because they were killing it in gaming…that was just a nice side bonus. It is the IOT/Jetson, AI/ML for data analytics, cloud DC solutions and on-prem DGX DC solutions, and the entire AV market that is still not yet material and ahead of them.

I actually never invested with crypto in mind, so this is annoying, but I hope a good dip-buying-price at $168.

Dreamer

5 Likes
23

If you really wanted to know you can go back and check the ASP for every quarter I work in the electronics semiconductor industry and I can tell you it is common practice for manufacturers to raise prices or lower them via their price book that is why you see the wild swings and earnings for micron because they do in fact raise prices during times of constraints not sure about Nvidia though but I’m sure the same applies to them

24

It would also explain why they plan on missing on every single category next quarter and the mist of this quarter on data centers ASP’s being lower across-the-board including data centers do two more competitive pricing in the marketpla

25

12x,

It was not the gaming miss that disturbed me. In fact I am not paying any mind to that. Channel problems are transitory, crypto problem is a thing of the past, Nvidia still has a virtual world wide monopoly moving forward. That is at least the narrative, and will probably play out more in the coming year as the Turig architecture takes over.

My issue was with the miss on data center. The growth slow down was not awful, but it was a material slow down nonetheless, and miss.

Are you saying that the crypto issue was also related to the data center miss this quarter? Since Nvidia has specific data center chips that are not used for gaming, and since crypto used either another non-gaming GPU or gaming GPUs, why would data center be impacted by the channel issues in gaming?

Turig has “record” quick adoption in servers, as just announced. Nvidia has overwhelming dominance in the datacenter, with AMD having near nothing, with the primary competition coming from FPGA in the data center. Yet you are saying there was pricing pressure on data center chips as well? At least speculating about this.

I would be interested to know your thought process in regard to this as it relates to data center numbers this quarter, that also missed. This is the miss that I found concerning. Data center is not related to gaming, except that some gaming chips can be used in data centers, although Nvidia would classify these as gaming because they have no way of knowing when they are used in a data center (smaller ones, anyway) capacity.

Thanks.

Tinker

3 Likes
26

The issue is a glut of GPUs on the market caused by people who bought them for crypto mining during the boom and who are now selling them. Pretty sure, that’s why demand is down and Nvidia is sitting on an excess of investory.

I think it could be also a glut of AMD GPU’s on the market and people think they are “Good enough”. If so this might take a year to work through.

Andy

2 Likes
27

The rule holds, no exceptions to textbook cheap. I think most here, and my thinking was Nvidia was getting lumped in w AMD and other semi ‘s. As we see now the rule of cheap even applies to a Nvidia. Will not make that mistake again.

I didn’t expect Gaming to come down this far. Who would have thought that gamers would be buying AMD GPU’s only because they were cheap. But these guys have to be on the top of the game or they lose so how long can they run these “cheap” GPU’s? I think this was a cyclone that caught most/not all unaware. Just didn’t expect people to go for the cheap stuff. Another year and they will be falling all over NVDA.

Andy

2 Likes
28

I’m wondering how many of you could be in denial and not reading below(what the master has said) several times over…Yes next year, yada yada…but to me, it’s right in my face…and for the time being money needed elsewhere for growth stocks…

“Overstock, working down inventories, expecting demand to grow as inflated prices fell, pricing stabilized”

That’s an entire vocabulary of problems you don’t ever see if you are not in hardware companies. It’s all the difference in the world!

Saul

8 Likes
29

Tinker,

My issue was with the miss on data center. The growth slow down was not awful, but it was a material slow down nonetheless, and miss.

It’s possible that datacenter customers are anticipating the release of the successor to Volta (which was introduced in 2016). The next-gen successor should be released in 2019 and will help re-accelerate datacenter growth.

“Volta-next” was teased recently in the US Dept of Energy’s new supercomputer:

Coming to the GPU side of things, the NVIDIA based Volta-Next GPUs will be offering the bulk of compute power on the supercomputer. Each node will contain 4 Volta-Next GPUs. The Volta-Next naming convention means that the GPU will be the successor to Volta and will offer greater than 7.0 TFLOPs of compute (GV100 currently does 7.5 TFLOPs), more than 32 GB of HBM2 VRAM and next-gen NVLINK for fast interconnect between the GPUs. Details are not known regarding the new NVIDIA HPC GPUs but we can expect them next year at GTC ’19. The GPUs are said to perform up to 4 times faster than the current Cori system.

https://wccftech.com/amd-epyc-milan-and-nvidia-volta-next-pe…

30

My issue was with the miss on data center. The growth slow down was not awful, but it was a material slow down nonetheless, and miss.

DC grew (YOY) at 71%, 83% and 58% in Q1, Q2, and Q3 respectively, so looks like a slow down. During the call management said DC is doing well for Q4 and non gaming rev should grow sequentially Q4 on Q3. So, this mean Nvidia expects DC in Q4 to be >$800B. I think they would have to beat this handily, more like >$900B which would put DC growth for Q4 at over 50% (YOY). Anything below that will be very hard to explain and not good for the stock. In other words most of the $700B shortfall for Q4 should come from gaming slowdown based on the explanation we got on the CC.

BTW Cramer called this one right.

31

I’m wondering how many of you could be in denial and not reading below(what the master has said) several times over…Yes next year, yada yada…but to me, it’s right in my face…and for the time being money needed elsewhere for growth stocks…

Branmin,
When it comes to my money the only master is me. I will listen to other people’s thoughts but I am not willing to follow any herd. Nvda is a great company and I am willing to wait and let the thesis play out. I can afford to be a little more patient.

Andy

25 Likes
32

I think this is interesting.

Leaving aside the fact that Andy or any of us are our own masters, (Saul has always said don’t blindly follow my trades but take responsibility for your own decisions and Buffet has always said “put as few people between you and your money as possible”; there are some points of debate here for sure - especially with Nvidia.

First there is the point that this is hardware as opposed to software or SAAS or anything else.
Second there’s the point about chasing growth all the way to the bleeding edge and ruthlessly allocating resources to the fastest/tiniest rocket ship vs larger and potentially slowing plays and making tradeoffs in terms of risk considerations.

Conceptually without wanting to break away from the purpose of this board concerning Saul method investing I have voiced a few thoughts over the years about both these points.

1) Hardware vs Software/SAAS etc
Ok there’s quite a lot going on within this to maintain a rule around.
Firstly as THE master (Gates) has always said hardware is a depreciating asset and software is an appreciating asset. I think the depreciation of hardware was always known but the appreciation of software was disguised in the perpetual license business model but now in the subscription era it is obvious.
Secondly there is the durability and substitutability of hardware. Mobile phone and mobile phone processors become old and past it within a couple of years. Some hardware players in this space have to maintain continued R&D and bring out the highest performing chip or phone otherwise they are out of the game. Other hardware plays can extend duration for much longer but have learnt that this can eventually be a bad thing and resort to built in obsolescence that occurs the day after your warranty runs out, (Sony Walkmans anyone). (The reason why Crocs hit their speed bump was that their plastic shoes just never wore out so no-one ever replaced them; on the other hand Skechers’ soft rubber components are worn out in months). Then you have substitutability issue. Unless you have an IP moat or monopoly like Intel x86 to protect you, players face the danger of instant substitutability which was the case for Skyworks and many chip component players etc.

In both these situations hardware can be a good play or a very exposed play.
I would not want to be in Skyworks or Cisco which are more exposed than advantaged but I am prepared to be in Micron and Pure Storage who I feel are more advantaged than exposed and I have advocated as much.

Back in the 90s/2000s Nvidia was in a running battle with its Tegra vs Qualcomm’s Snapdragon constantly outdoing each other with the next release. The durability prospects and substitutability prospects were terrible. This seemed to change with the continued lead in performance in the graphics space, the introduction of CUDA and proprietary protection in its code as well as the crypto boom. Still without crypto I would say Nvidia is still exposed to the durability issue and with a revitalised AMD and increasing variety in server chip set designs, substitutability looks a risk as well but for CUDA in AI - but AI is still tomorrow’s world in terms of volumes. Then there is still the issue of depreciating hardware as has been seen in prices although I would say semi pricing is more cyclical than constantly downward depreciating like automobile components.

2) Growth vs risk tradeoff
Another topic I have raised is the point about whether there is room in a Saul like method to chase growth at all costs vs build in some lower risk tradeoffs. This arises for me in in particular in 2 areas - profitability and size.
Firstly - profitability: As we know many of our software or SAAS Saul method choices are very high growth rocket ships, however many are not profitable and potentially there’s a risk that they won’t be and growth might curtail before breakeven. I raised this on Shopify as well as Arista and you could also make the case for Nvidia. Would you be willing to tradeoff 60-70% growth in an unprofitable venture for a 30-50% growth profitable venture? Is this risk reduction worth the tradeoff.
Secondly - size: Again whilst we are all wanting to be in the next Microsoft and Amazon and in positions that could 10x or 100x, one of the benefits of larger corporation investing whilst generally facing the law of big numbers (Ali Baba and Facebook have been exceptions), is that they are supposed to be more transparent, stable and less likely to face sudden systemic shocks. This isn’t always the case, (GE anyone?).

Actually I learnt the importance of this tradeoff ironically when I was running a high yield portfolio investing approach in the UK. High yield is great up until the point when you are chasing yield but actually also increasing risk. Sometimes there is a reason why even blue chips appear to be offering high yields. Constantly chasing the highest yield actually increased risk and even reduced overall returns because some of the high risk plays were not sustainable. Same for growth.

For Nvidia - they were supposed to be large and profitable and much lower in risk. They were supposed to have durability and substitutability protection and benefiting from secular tail winds. I even bought into this risk reward picture. I don’t think it is wrong to still consider hardware selectively and still consider tradeoffs in the growth profile, but I think it is worth being honest and re-evaluating what we are seeing in Nvidia.

I don’t have the answers so, apologies if you feel you’ve just waded through a long post with more questions than answers but hopefully the questions are good ones to be asking ourselves to get to the right answer on this or at least learn from.

Ant

36 Likes
33

I was reminded this morning of a paper I heard at a professional meeting many years ago on pricing. The core point of the talk was that the best pricing was based on the units of whatever the customer valued. In any transaction oriented business, the value lies in the transactions so if you charge by transaction the customer is likely to be happiest. E.g., if the business is a highly seasonal one, like Christmas trees, for example, charging by transaction means high costs during the high revenue months like November and December and little or no cost in other months when there is no revenue. SQ is a great example of this and is one of the reason they can be a good fit even for something like a vegetable vendor who sells through farmer’s markets because any costs are directly related to revenue and there are no costs during the idle winter months.

Hardware is very hard to sell this way, except for PAAS. You either have it or you don’t. PSTG probably approaches this pattern except that the charge is not for the volume of actual usage.

SAAS is obviously the context in which to implement this principle for software, but one has to be aware of the pricing structure, i.e., a fixed monthly cost is not a great fit for a highly seasonable customer.

2 Likes
34

posted this on NPI…I think I was hit so hard yesterday I was in denial a bit that the narrative may have changed. Slept on it a bit, and here are my thoughts this morning:

Ok. Had a restless night to mull this over. Tinker gave me an ear worm (brain worm?) by mentioning something like “market won’t give them same multiple anymore”. Gave this a lot of thought and ran numbers in my head. BTW, you can’t sleep well if you lay in the dark running numbers in your head.

The problem with my original post is I didn’t have a plan if a post-ER drop WAS warranted, and it was in this case. Just completely unexpected.

What I didn’t care that much about:
Crypto affected channel supply.

What bugged me:
Huang seemed to lack crisp/prepared replies for what was surely the biggest ER/CC in recent memory. Hard to judge body language via phone conf call, but a couple times he says “what was that other question?” and maybe was too casual with “we thought prices would go down”. Then he kind of robotically threw out the “Moore’s Law has ended” line 2 or maybe 3 times, as if robotically repeating things will make the pain all go away.
I was also disturbed that they could be in the gaming business for this long and not have a better handle on channel inventory.
DC growth was down, but not much. If gaming had not been crushed, I may have glossed this over more. But since everything was on a spotlight, even I have to admit that 58% is great but still a slowing number.

What should have happened on CC:
Strong, serious tones. Open with “we as mgmt take full responsibility for this inventory misjudgement. It was never our focus or intention to be in the crypto business, but we aren’t new to the gaming channel and should have better predicted inventory impact, especially in light of the RTX series announcement, which served as a secondary reason for gamers to not buy existing channel inventory. Make no mistake, this is a one-off occurrence and will be fully behind us by X date. Further, our other lines of business are firing on all cylinders and are not impacted by crypto. Laptops are also not impacted by crypto.”

Instead they were kind of half-exasperated and “hey…just inventory…ummm…we thought there would be less.”

I expected more from Huang, who is great on stage, but maybe I haven’t listened to a CC for NVDA in quite a while and have been spoiled by Jeff Green at TTD or Twilio’s mgmt.

All that is just talk, but the “multiple” ear worm via Tinker got me running numbers.

If gaming is this utterly unpredictable, and there is no quick path to cloud-online/streaming gaming that NVDA can slide their existing gaming business over to (removing pesky channel inventory issues and making gaming a “cloud” play), then that whole segment goes from being a still-growing cash cow that fuels innovation in other segments to just a commodity business that deserves a lower multiple.

making up rough numbers here, but they will do about $12b in rev this year. Gaming about 50% of that.

Their $6b in gaming may only repeat for the next 4 Q’s, and may have little growth or even negative growth on next TTM compared to last TTM. What multiple does that deserve? Maybe what the company looked like in 2015 before DC/AV and other markets started transforming this gaming company.

Let’s call it a $30b market cap, or P/S of 5 for the gaming segment.
DC growing 58% y/y latest Q, and if next year only grows a bit over 40% and then only 30% year after that, along with a sequentially up Q4 in 3 months, they will be, at end of calendar 2020, a $6b DC segment.

Let’s be generous and give DC segment a 12 P/S or $72b
AV is further behind, but let’s assume one day it can/will match DC. What year is that…2022/23?

Let’s be generous and say AV will be a 12 P/S and at $72b at end of calendar 2023.

Then you have a $170b market cap, if you wait until 2023, or about 4 years from now.
Yesterday they were $120b market cap, and today they may be back close to $100b.

That is 70% growth, which sounds pretty good, but it is over 4 years, and do I believe TWLO, TTD, or even MDB would do better than 70% growth over the next 4 years?

I hope so. Only reason they don’t is factoring in a recession I guess.

NVDA mgmt let me down. I can forgive crypto. I can’t forgive a lack of candor or preparedness, and in investing I can’t forgive a broken thesis that impacts 50% of the company.

You could then treat DC/AV as “hidden growth” ala TWLO, but that narrative doesn’t hold up when you first have to greatly compress the multiple on 50% of the revenue (gaming segment).

Sold out of NVDA pre-market, and slowly adding into TTD and TWLO. Maybe MDB over time.

Dreamer

26 Likes
35

Ok. Had a restless night to mull this over. Tinker gave me an ear worm (brain worm?) by mentioning something like “market won’t give them same multiple anymore”. Gave this a lot of thought and ran numbers in my head. BTW, you can’t sleep well if you lay in the dark running numbers in your head.

Dreamer, we have all been there but not sleeping well due to stock picking, well it’s time to re-evaluate or it will churn you up inside…not worth it, your health is more important.fwiw.

2 Likes
36

Apple owner here for many, many years. This company obviously trying to get away from being known purely as a hardware Company and into service and software related fields. Understand though what you are saying.

I’m now re-thinking my whole “NVDA is also a SW company” thing, and Apple is an excellent reason for this. Here is the problem they both share: they are trying to (or can only) sell their SW and services to owners of their HW. (maybe not 100% accurate, but more than close enough). In other words, if iPhone sales don’t accelerate there is only so much more iCloud growth they can get, or iTunes move rentals, or Apple Pay customers, etc. Apple Music might be the exception to that, but honestly I think Spotify probably is the go-to music service for non-Apple people.

NVDA has CUDA, yes, and lots of other software, and that IS important. But, as far as I know, it must run, or runs best by a large margin, on NVDA hardware. So… if NVDA doesn’t sell more GPUs they aren’t going to grow the SW.

Just my take. I was currently revising my AAPL holdings for over a month now, and has added my NVDA holdings to that list.

Bill Jurasz

1 Like
37

Tinker,

It is true, I am speculating, and I cannot find hard evidence to support my theory.

So my theory is as follows:
If there is a shortage of GPUs in the market, GPU manufacturers are in a position to raise prices, regardless if it’s for gaming or datacenters, especially if they are built from the same wafer/die (which there has been an increase on wafer prices by the way, and there has been an increase on memory prices used on the video cards. I would be surprised NVDA would absorb these price increases instead of passing them on). Due to this shortage, datacenter GPU prices could go up, especially when AMD will do little or nothing to support lower GPU prices during this time. So there would be price increases all across the board due to the wafer being the scarce object. Then when the shortage ends a drop in price occurs across the board, including datacenter, causing a shortfall in revenue growth directly tied to unit prices but not volume.

I did not listen to the con call so they could have a better explanation.

I also looked at historical MSRP information for video cards and it’s consistent with what they are today, when you compare pricing for the exact same model. However, the newer models are reaching higher than ever before seen pricing for GPUs, and it’s quite possible that NVDA and the rest stuffed the channels with what they preferred to make (high $ product rather than low margin… because hey, people will take what they can get when they need to bitcoin mine). But now that things are normalized NVDA may be seeing more competition from Intel or AMD in the datacenter thus helping lower prices for GPUs there.

As I said I have no evidence to support this, other than NVDA’s gross profit margin history:

https://ycharts.com/companies/NVDA/gross_profit_margin

There you can see that NVDA’s gross profit margin peaked in April 2018, the exact same quarter that the bitcoin craze would have ended. Since then gross margin has fallen for 2 quarters in a row. This supports my theory that NVDA has been generating (possibly unsustainable) high margins for it’s product.

This is the only explanation that I could have for an across the board shortfall in NVDA’s sales channels. You can see that the gross margin has been growing over the years but around the mid fifties since 2013, then started really escalating in 2016.

2 Likes