Citron's next target: NVDA

Investment Analysis Clubs Saul’s Investing Discussions
21

Oh…and Google is using it to make searching better.

Mike, you’ll have to define “better,” IMO Google has become much worse. The other day I was looking driving around looking for a particular store. I remembered they had a store in a specific location, but when I got there - no store. OK, I Googled the store by name. The name contained the word “appliance.” I received maybe six returns before I scrolled down to the one I was looking for. Google apparently prioritized paid advertising above the specifically named company I was searching for.

Kinda pissed me off. About ready to switch to Bing, although I’m not sure it would be any better.

1 Like
22

Tinker,
Couldn’t agree more. It makes sense to be aware of emerging technologies, but don’t let that greatly influence your investment decisions. There are now prototype photonic transistors in the lab. Far superior to electronic transistors, orders of magnitude faster with lower power requirements. But we’re talking prototype transistors. When, or maybe if they ever become part of an integrated circuit product is the time to reassess investments in chip makers. Until then, it’s mostly noise.

1 Like
23

Agreed on Google search. It feels more that they’re emphasizing what THEY want to show me instead of what they infer I want to see. Ads. No matter how specific the search terms.

I’ve defaulted all my devices to Bing. It’s better in terms of putting what I want closer to the top but the other features aren’t as good. Nothing beats Google Maps

Google services like search and map use their in house developed TPU(ASIC) to run. Presumably those networks were trained on Nvidia GPUs before they were put to work on the TPUs. The TPU would not have been capable of training them but would be well suited to run an image through google translate to tell you what it means after it has been trained to do so. Google owns lots and lots of Nvidia GPUs in the data center.

2 Likes
24

And a product that is just slightly better won’t do it. It has to be much better, and it has to meet a moving goal post as what GPUs can do, at the power they can do it as, continues to improve.

This is such an important point. NVDA’s technology improvements are out pacing Moore’s Law. It’s a double exponential. Let’s look at example in another industry: DNA sequencing. Illumina has 90% market share. Other companies starting developing technologies to compete with ILMN. To get the big dollars it mostly comes down to sequencing cost. PacBio started develop a promising technology but the several years of development were required for commercialization. In that time period ILMN advanced its technology and drove the cost of DNA sequencing down significantly. The goal posts had moved which left PacBio unable to compete in the parts of the market where the big dollars, the most important applications, were. PacBio was left to be able to go after some niche applications. Look at the market cap changes of ILMN and PacBio over the past 12 years. Look at the stock prices. Today, ILMN still commands 90%+ market share.

NVDA is moving the goal posts at a blistering pace. They have set the SW standard. They have solved manufacturing issues. They have solved the scale issue. They have locked the distribution channels. They have set the standard. They have the partners networks…320 partners in the automotive space. They have the complete product solution. Yes, the valuation of the stock is high. No question about that. But no one is anywhere near to catching them, not for several years. If this were a 10k race on a track, NVDA is several laps ahead of anyone and it will see any contender approaching far in advance. And as Tinker pointed out, this visibility will allow NVDA to take action whether it’s a buyout or something else such as $2b into the better next gen version. Xavier took 2 years and 8000 man-years of engineering development.

Chris

20 Likes
25

I think one he brought on stage at CES had about 50,000 engineeeing years or some such (might have even been more).

Whenever I had classmates bragging about how much time they spent studying or how many times they read a given textbook, my question was always why it took them so long or so many tries…

I also wonder what kind of counting they use, since NVDA apparently has 10,000 total employees. Are all of them engineers who worked 5 straight years? Or half of them who worked 10 straight years? Maybe they learned how to count from their law or accounting firms’ billing departments. (just kidding, Tinker)

4 Likes
26

Chris,

Nice ILMN plug! I’m happy to say they are a descent part of my port. You pretty much nailed why. In the earnings call last weak they reported that clinical consumables (the recurring revenue part of gene sequencing) is growing like crazy. From 39% to 45% of total. A lot of research and treatment is being done using the various gene therapies and it shows in ILMN performance. 90% share of that market is just freaking fantastic. Another great company doing something really cool.

27

Of Nvidia’s 10,000 or so employees they claim about 5,000 software engineers

28

50,000 man years of labor? Hmm, I haven’t heard that before. First thing, let me suggest you check the number, seems a tad on the high side to me.

As a manager in the s/w development realm I had to estimate man hours for tasks and and projects. It was a necessary input in order to make staffing and scheduling decisions.

Consider, each FTE (full time equivalent head) works 40 hours a week. But, you never use 40 hr/week per FTE. You need to account for vacations, sick leave, staff meetings, water cooler, etc. Generally, where I worked we estimated each FTE at 50% - 60% time on task. The 60% was only used when we already had pre authorised overtime.

50,000 man years is a lot of man years.

29

The shortage of top end Nvidia chips is well documented. Most on MF consider that means excess demand. But suppose Nvidia has the Tesla 3 Problem, can’t make very many of them, problems somewhere in supply chain or fab. I’m not saying that’s likely because I think Nvidia would have had to disclose if that were the case. However I might make a WAG and say demand has surprised Nvidia, they may have underestimated it in drawing up chip production plans

In another 9 months or so all those chips will be obsolescent , newer better Nvidia gaming and AI chips will be available. If present users want to stay top dog they will need the newest chip.

Right now and for the next few years it is gaming that mainly drives Nvidia earnings. AI may be disproportionately profitable since I think few AI developers want second-rate ,slow chips, talented programmers are too expensive. Thus two fold increases in number of AI chips sold may yield 3 or 4 fold profit increases.

30

I also wonder what kind of counting they use, since NVDA apparently has 10,000 total employees

Watch CEO Huang’s keynote address at 2018 CES. He said Xavier took 4000 engineers 2 year to complete it costing about $2B.

Chris

31

Nice ILMN plug!

ILMN has really advanced DNA sequencing technology, my post was not intended to be a plug for ILMN stock. I owned the shares at one point a few years ago (think I bought at around $50 or $70 (can’t recall) and sold at around $130). I won’t buy it today. Why? It’s too expensive, IMO. Their future is dependent on adoption of sequencing into clinical diagnostics (research market won’t provide the needed growth). I also work for a competitor so wouldn’t rather not see ILMN succeed too much.

Chris

1 Like
32

But suppose Nvidia has the Tesla 3 Problem, can’t make very many of them, problems somewhere in supply chain or fab.

Rumor mill says that access to and supply of GDDR5 VRAM is the biggest challenge, and since NVIDIA sources from the higher-end sources (Samsung and Micron), that’s where you would pin the blame. They might be sourcing from SK Hynix, but that would /only/ be for the lower-end video cards, since Hynix quality is definitely inferior overall to the primaries and I’m sure NVIDIA is smart enough to ensure their higher-end cards “just work” near 100% of the time. (I love the @#&*^ out of my GTX 1060, bought on release day…)

I’m also not sure they would HAVE to disclose a memory supply shortage, as long as there were other factors in play, such as crypto-miner demand.

1 Like
33

https://seekingalpha.com/article/4142794-intels-secret-indus…

Could this be the answer to which company bests NVDA?

1 Like
34

<<<AMD (AMD) and Nvidia (NVDA) discrete GPUs are also accommodated by the Intel photonic fabric. The recent Intel-AMD collaboration in mobile might be a clever way for them to disguise this bigger project.>>>

Seems the GPU is part of this vision from Intel. Thus not an issue with Nvidia. From what I read Intel has given up on its own iGPU or something like that when it teamed up with AMD. Thus it is still Nvidia vs. AMD in regard. From everything reported Nvidia has a larger lead on AMD in GPUs (at least for any highly profitable and desirable space for them) than Intel has an advantage in CPUs over AMD.

The one current technology that may be material to Nvidia is the Google Tensor.

There is nothing proprietary about Google’s tensor, as they made the specs public (albeit, they have not made their next generation tensor public).

What makes it material is that Google has said that the next generation of tensor should be able to do machine learning as well. That would be a challenge to Nvidia’s core of machine learning, but also Nvidia is making moves to dominate the executing of the machine learning programs, that tensor does for Google (at least in some circumstances).

That is the one technology worth watching at the moment I think in regard to Nvidia. Will the tensor be both materially less expensive, less energy intensive, and greater performance than the Nvidia GPU? Google also has an AI infrastructure in place that could expand and challenge Nvidia.

However, Google is one of the prime players in AI and will be competing against many of its potential customers from IBM to Baidu and many companies in-between. Will these companies really want to buy tensors that benefit Google?

Further, to date, although Google gave away the specs, I have not heard of any news that companies are using the Google specs to in-source their own chips. We don’t know what Amazon and Microsoft and IBM are doing necessarily with their data centers as they could in-source in a fabless method. However, given how important scale is in semiconductor production, I doubt even Amazon could productively in-source in a fabless manner and compete with the scale of Nvidia GPUs.

Further, Amazon customers want to use CUDA, or use Google’s AI language, or the open source language that Nvidia maintains. Using an in-source chip may alienate part of their customer base.

Anyways, replacing Nvidia GPUs is not a simple thing, which is why the replacement technology needs to be materially better and then some.

Tinker

7 Likes
35

Sounds a whole lot like the Cramer, Greenberg, Cahodes and Rocker Partners handling of Novastar, Cree and Overstock.com via TheStreet.com years ago.

https://www.sec.gov/rules/other/265-23/rwyork100405.pdf

http://investors.overstock.com/phoenix.zhtml?c=131091&p=…

36

That bozo is at it again.

I thought there was a post about no personal attacks. I guess that applies only to bulls I guess. If you are bearish on a stock for whatever, reason, then it is a open season, I guess.

2 Likes
37

Look, if the facts can be shown and proven to be false (or at best, disingenuous), then I think the label fits. Further to the point, Andrew Left’s past behavior is its own best predictor of what we should expect from him now and into the future. I stand by the label, which wasn’t intended as a personal attack. He may be a great human in real life, but the purposeful spreading of FUD that hurts others is not the most honest attribute a human could exhibit. Feel free to disagree.

Bozo is the name of a popular clown. Thus, it wasn’t name-calling but equating him to a clown. Don’t think that qualifies as a personal attack, but I’m always open to being wrong. And if the consensus is that it falls under a personal attack, I’ll refrain from drawing such analogies going forward. But I still think he’s a clown and likes making circuses…

8 Likes
38

The choice of the words are yours and yours alone.

Someone calling a valuation at 15x sales and the recent move is too quick, the stock could trade down is begin interpreted as engaging in FUD, shows your intolerance for alternate views and the words you choose to describe is the manifestation of that intolerance.

I understand we are all humans. It is no big deal. People call each other much worse.

2 Likes