The software related optimizations are expected, not a surprise. Folks are reading too much because they are coming from China. Today $GOOGL announced they can run the 27 B model in a single H100 GPU. Like I said $NVDA, AI may undergo some digestion phase. But overall compute requirements are going up. Whether that is coming from $NVDA or others we will see. The cost vs benefit of using cheaper chips needs to be seen. Not all announcements become products.
1 Like
Let me give you a different view. The below is Micron, the memory manufacturer. If you see the last 10 years, the growth is uneven, the margin story is bad, so as a stock unless you picked the cycles correctly you would not have made money.
But the revenue growth tells you while the cost per GB of memory has come down drastically, they ended up selling more memory is why the revenue went up.
What happens to NVDA price is going to be driven by AI cycle, but overall compute requirement is just going to go up in the next decade. So, if you could get $NVDA who is the market leader for the GPU chips, you will do okay over a decade. Competition in technology is nothing new, but technology and medicine are two fields only the best solution has longevity and staying power.
On this we certainly agree.
This is a definite maybe. Remember SyQuest and Iomega? Great products, obsoleted almost overnight by cheaper tech. I typed on a Wang word processor in the 80ās and they had a nice run, until our IT guys washed them away with a sweep of the hand. If you had them 10 years earlier you āhad a decadeā, if you bought them at the end of the cycle you had a year or two.
Like I said, this is a ādunnoā, and I donāt follow things daily or weekly; at this point of my life I need more stability. I see that you trade, use options, puts and calls, so youāre quite active. Iām on the other side of that mountain. I hope Nvidia works it for you, as I say it probably will.
Right now, I donāt have any stakes. I have converted my position to covered call around $140. I am waiting for an entry price. This will not be a short-term trade rather catching an intermediate term wave.
But you donāt have to be tech to make money. It all depends on oneās risk tolerance, return requirement, time commitment etc. Someday, I will be able to stop all this, and just travelā¦
Yep. The hard part is determining WHICH decade! Is it 2015-2025? Is if 2020-2030? Or is it 2025-2035?
1 Like
Let us say, you are not a tech person and completely missed AI Wave so far, in my view we are still in the early innings. Let me further say, there is going to be an AI digestion phase, and post that a growth phase. So you still have time and I hope we can get $NVDA below $100 or even $75. But if you want to play AI, you donāt even need to be in $NVDA , look at software names like $SNOW, $MDB, $PLTR (not at current price), etc.
Very early innings, Iād say. Thatās why itās hard to tell where the market will go. Think: AOL, Apple (first iteration), Atari, Commodore, MySpace, RIM, etc.
Thereās an interesting piece in the new issue of Bloomberg (no link)
For most of the world DeepSeek seemed to explode out of nowhere in January with open-source artificial intelligence software that rivaled models from OpenAI and Google - all of it built, purportedly, at a fraction of the cost of competitorsā models. For fans of a website called Chatbot Arena, however, it was bit of an eye-roll moment: theyāre been watching - and rating - the progression of models from the Chinese company behind DeepSeek for months.
The chatbot arena offers users a ātaste testā, you ask a question and get two different AI answers, you rate which one is best, and only then is the name of the AI revealed. Meanwhile the rantings are totaled up, and DeepSeek is consistently rated at or near the top.
Of course the ātestā may be flawed in several ways, just as the Pepsi challenge test didnāt really get at the true āloveā of Coke. For instance, it may not reflect the way in which a company or people will actually use their AI bot - but still itās an indication that the most expensive chips may not be the only way to pursue AI - which would undercut the forward Nvidia sales predictions somewhat.
As I say, too foggy for me. Interesting sector to watch, tho.
Do not underestimate US technology power, hegemony. From resources, to talent, US tech companies are miles and decades ahead. For ex: Google has been contributing to the AI field through open source for 20 years, from machine learning to the ground breaking ātransformerā architecture on which modern LLMās are build and everything in between.
What $NVDA has done in the chip field is incredible, actually words do not adequately explain.
| Model | transistors | year |
|---|---|---|
| 8086 | 29000 | 1978 |
| Pentium | 3000000 | 1993 |
| Intel - Core i9-14900K 14th Gen 24 | 4.2 B | |
| blackwell geforce | 208 billion | 2025 |
The reason I am presenting the above is NVDA dislodged such a powerful technology company, which dominated computer chips with almost 90% market share for 4 decades.
So someone may in future dislodge NVDA, but that day is not today .
Separately, NVDA CUDA platform is entrenched for 2 decades with millions of developers. So they have moat.
It is a separate conversation whether that moat is worth $3 Trillion or $1 Trillion. But $NVDA will be around at least for another decade plus and will dominate the AI field.
AOL failed because they failed to innovate. OTOH, netflix while their origin may be delivering CDās using us postal service, evolved and innovated. Not only they innovated, they inspired their partner AWS to innovate. Not only Amazon retail business, but Netflix streaming is another pillior that drove the AWS infrastructure growth, especially Netflix is the primary reason behind AWS content delivery network buildout.
The reason I am mentioning this is, one Deepseek is not enough, you need the ecosystem. Thatās the reason people come to valley.
3 Likes
You misunderstand. Iām not saying DeepSeek will dethrone Nvidia, Iām saying any cheap chipset might . You keep arguing āNvidia is greatā. OK, it is. It appears that a product with 75% margins can be closely simulated, perhaps even outperformed by chips which are 50-70% cheaper. Thatās never a good sign; margins tends to come down, with only a few exceptions.
Microsoft OS and iPhones manage the trick, but there is a slew of supporting middlewares that keep their product out of reach for other contenders. If a cheap chipset will produce a desirable result, I donāt see why people needing an AI interface from customer support or whatever will pay 100% more for the privilege.
For example, the Apple M3 Ultra chipset has 184 billion transistors. It has the same order of magnitude of power as the latest Blackwell version. But it is 90+% less expensive AND it uses quite a bit less energy to run. I wonder if Apple is building racks and racks with that new chip in them?
1 Like
Just to be clear, DeepSeek is not a chip, but LLM model. Now, Intel for almost 3 decades had 60%+ gross margin. So donāt assume automatically the high margin can be replaced by others.
1 Like
Yes, and maybe Nvidia will hold on to those margins for that long. Intel prospered because it was an essential part of a combo-duopoly in a business for which there was no substitute. Their margins these days are not the same, tho, are they?
(But sure, Intel was a good investment for many years. I had a slug back on the aughts. Gave it up when I cleaned house just before the housing crash.) Maybe Nvidia will last that long, if they do, good for the owners. I just see a potential moat breaker on the horizon, but acknowledge I could be totally wrong. Either way, itās in the ādunnoā pile for me.
1 Like
It looks like today is going to be 90% of the volume on the upside. Yet another sign the bottom is close.
If you are someone with cash, you may want to consider nibbling on your favorite names.
1 Like
This is an hourly chart month to date comparing QQQ, PLTR, APP, NVDA SNOW, comparing the index against growth names. The vertical line is Monday 3/10 at 13:30 hours. The point is while the index moved sideways, and bottomed on Thursday, many growth names actually bottomed on Monday and started rallying, moving 10%+
1 Like
