Jlagreen
- have emotions - have self-awareness - express creativity
Why? Because that would be the end of humanity. In such a case, AI will feel "superior" like humans do to other animals. And we know how that will end.
And intelligent entity without the crap above won't ever "feel" superior while it will still be able to built a beaming device on humand command.
Nvidia's margin is so high because they deliver way more than just a GPU.
Do people seriously think that data centers and AI compute is only built in US? WTF?
Hyperscalers reach their limit in CapEx while Nvidia continues to grow. Nvidia has no limits since their TAM is every industry and every enterprise as well as every person in the world who is interested in using AI compute.
AI compute and AI usage is a global thing.
Look at South Korea in the 70s and today. Technology helped South Korea to leapfrog many Western companies in a few decades.
It will happen again. Many 3rd world countries have better mobile phone networks than Europe while they basically have ZERO copper in the ground.
While in Western countries everyone is skeptical about AI, Asian countries are building robot armies to replace Western collar workers. So we think about regulation and bubbles while others think of how leapfrogging us. Invest and cost is 2 sides of the same coin. It's obvious what we see here and what others see around the globe.
And which company profits the most from exponential growth in token generation?
Here we go, I just drew a "circle" for these naysayers.
In the smartphone market, Apple isn't #1 in units but generates 75% of the profit of the whole market.
Would you rather be #1 in units or profit?
There is a reason why Apple can ask for prices only few competitors can keep up with and the reason isn't the HW for sure. Similarly with Nvidia, the margins come from the ecosystem, not the HW.
What is the TAM of AI and therefore the future AI compute demand?
Answer this question and you might understand what is going on.
But Murphy's law is dead. No future chip will leapfrog easily current chips because we have reached hard phyical limits in chip density and downsizing. Huang's law by Jensen Huang focuses on something else and that is token performance per Watt at scale.
Blackwell needs double TDP than Hopper and Rubin again needs almost double TDP on a rack but in the end Rubin will be like 100x token performance per watt on a scaled data center. This means you have more energy need but you get multiples of token performance because you start scaling in the data center.
The local chip will never be able to keep up with the data center scaling economics. This is why everyone is so crazy about building data centers because they can see the economocs behind it.
What people don't seem to understand if tokens become more available and cheaper then not only more people can use them but a single person can use more as well. Why should you be limited to 1 AI agent? Why can't have you have multiple agents running on multiple devices daily for you?
This is why demand will grow exponentially with the growth of token economics. We have seen it for the last few years and much more is yet to come.
To get an idea, if you go gaming via cloud then fast internet + office PC or Laptop is enough. So you save way more than the GPU only in a proper comparison.
This is why I play consoles only. I can play games for years without ever changing HW and save tons of money compared to my PC gaming times.
Nvidia having 4 customers at 60% is good news because it was 90% 2 years ago.
Tech companies have the following options:
1. Buy Nvdia high performance stack with stable and fast support and deploy your SW developers to quickly get started
2. Buy AMD whatever stack and deploy your SW developers to make a lot of ground work
3. Develop and deploy your own chips and use your SW developers to make your SW from scratch but exactly as you need it
The only reason they might go with No. 2 is if AMD gives them the HW more or less for free but maybe even then will mix 2 + 3 if they don't want 1. And AMD deal is showing exactly this, AMD has to give free quity to get a customer for their HW.
What people don't get, who in their right mind would switch from one vendor lock-in to another with the difference on investing SW development into the 2nd???? The resources spend on AMD will bind customers to AMD. It doesn't matter if RoCm is open source as long as it runs only on AMD. If RoCm would run on any AI chip (including Nvidia) then we would have a case of an interesting switch but then the question comes up, why buy AMD if RoCm doesn't require it?
Imagine you have 1 million GPUs and you have 99% utilization of theoretical performance in the system with inferencing. That would mean 10k of GPUs are basically idle and draw power. You could now try to identify which ones are idle but you won't find them because utilization is a dynamic process so while all GPUs are under load not all are running 100% performance beause of interconnects and networking not providing data fast enough so your whole network becomes a bottleneck.
So what you need is a very smart routing process of computation requirements on the whole cluster. This is pure SW issue and not HW issue. This is the SW Nvidia has been working on for years and where AMD is years behing.
This is also why Jensen is absolutely right to say that competitors can offer their chips for free because Nvidia's key in TCO performance is the idea of one giant GPU so SW and networking allowing for highest utilization of a data center. You can't build a GPU the size of 1 million GPUs so you have to think of the utilization problem of a network of GPUs.
In the real world utilization rates are way below 100% so every % better of utilization is way more worth than the price of single GPUs. The idea here is that the company providing 2-3x higher utilization can easily ask for like 5x higher pricing per chip and will still deliver a better TCO.
The issue here is now, that every large customer of AMD will now probably ask for equity. AMD has put itself into a pit hole with that deal.
If I were Hyperscaler CEO, I would basically ask for the a similiar deal as OpenAI or no business. Sorry Lisa Su but as a CEO giving equity to a customer is an absolute red flag because it starts a negative spirale you can't stop.
It seems that no matter the discount, OpenAI wasn't ready to make deal without equity. This tells you exactly how AMD is seen in the AI world.
OpenAI will take the compute for free and help AMD to rise stock value but it won't help AMD one bit because if AMD remains in the current position then OpenAI and Hyperscalers can get great deals with equity from AMD. The incentive isn't now to improve AMD to be competitive but to squeeze everything out of a company being desperate enough to give equity to customers.
And AMD will feel this. Nvidia will remain dominant because of ecosystem and supply. AMD can't easily replace Nvidia in supply chain and Nvidia is already strongly entrenched in many AI compute operations. And on the other side Hyperscalers are focused on their own chips (even OpenAI LOL) so they will tell AMD "Give us equity or no deal". This deal might be really the worst AMD deal yet because AMD is telling the world "here, you can get free AI compute from us financed by our equity". And while it might push AMD share price the very share price will drop 80-90% like any other one in case of an AI bubble pop.
Nvidia is more worth than $1 per share on gaming alone.
And now imagine what will happen when OpenAI makes deals with Nvidia and AMD. Do you think Hyperscalers will just watch?
I expect Musk to make a $1 trillion deal soon. I guess, that's why he wants to get the $1 trillion from Tesla.
And do you think Meta, Amazon, Microsoft and Google will stand by while Altman and Musk are buying future supply from Nvidia and AMD?
I love that. As an investor in Nvidia, I hope that these future promises will push the stock 4-5x quickly in Cisco fashion because then I can sell and retire in my 40s with a huge pile of money watching the bubble explosion on some beach on an island :)
Or some interesting news here: https://nvidianews.nvidia.com/news/nvidia-powers-worlds-larg...
At least, they aren't waiting until CUDA quantum becomes as large as CUDA for GPUs.
And just like with CUDA no one else has currently the resources to go there because at the moment you have to invest heavily without any return.
AMD however has to design new special APUs for Xbox and PS. Why do they do that? They could just decide to step away from the tender but they won't because they seem to be desperate for any business. Jensen was like that 20 years ago but he has learned that some business you simply step away from.
Will you buy an AI GPU coming from China on which you will train your sensitive data?
People, really create some facts without checking some simple basic things. Huwai was banned for modems in the Western world but sure we will buy their AI GPUs LOL.
And at the same time how can it be that Chinese companies want the far inferior H20 or soon B20 instead of anything coming from China itself??? Nvidia has warned us that they can't provide the demand for H20 chips due to supply but yes sure Chinese chips will kill Nvidia lol.
Also Nvidia could even afford to burn 100s of billions of dollars since they are becoming the most profitable company in the world with probably passing Apple this year.
BUT then Nvidia would compete with TSMC and that is a problem because Nvidia is building up a cash and supply moat. Why should Nvidia build TSMC competition if they can simply buy 95% of the packaging supply at TSMC so competitors like AMD and others struggle to get supply? By booking everything at TSMC, Nvidia can easily keep their market share even if competitors improve their products.
This is a huge differences to many industries where most companies have their own production. Imagine there would be only 1 car manufacturer and all other car companies would only design. And now 1 car company would book 100% of supply from the manufacturer. What would the other car design companies do? They would get out of business even if they design better cars.
Nvidia has done this before and that was in the 90s. By speeding up design and releasing new products and more products 2-3x faster than any competitor. The result was that from 90 competitors in 1993, there was only 1 left in 2003 for Nvidia in gaming GPUs.
And Nvidia is doing that again by speeding up their roadmap cadence and as well as booking all supply. Nvidia is crashing competition not only with a great product but by removing their competitors' option to place their product in the market.
Estimates for next year are $300b. Sources are rumors about the supply of Blackwell and Rubin GPUs for 2025 and 2026 with estimated ASPs.
Nvidia is selling everything they produce so insiders at supply chains can easily tell how much revenue Nvidia will roughly make.
Nvidia is going to sell >5 million Blackwells this year and will do $200b in revenue with that alone.
Nvidia has a high net profit margin of >50%. If Nvidia would make $4 trillion in revenue then they would have >$2 trillion in net profit. Then the market cap would easily be 5-10x higher than today because otherwise Nvidia would be the cheapest stock in history of all time.
Market cap is also a very bad indicator as it doesn't really tell how much money was really invested into the stock. Market cap is just a product of shares * prices. For example, I bought Nvidia shares in 2016 for a certain amount. These shares are >100x more valuable today but I didn't put any extra money into them. So 99% of "my" market cap was simply created by traders pushing up the stock price.
If tomorrow, the majority of Nvidia stock holders decide to sell and all stocks are sold then I guarantee you that never ever will $4 trillion be traded because if there is a strong sell move then the stock price will drop like a rock and the last sellers will get a fraction of money as they have based on todays market cap. We might be lucky to see $500b of trading volume.
If you build large eCommerce like Amazon then of course you need huge IT infrastructure for it. Amazon was kind of forced to build the IT infrastructure and in that process Bezos saw another business opportunity because not every smaller company can easily build large IT infrastructure but many companies need it. And just as AWS built "SW services" for Amazon eCommerce it also became the foundation of the cloud business.
That's what people don't get. We're still primarily in AI research mode. The race in LLM training isn't about making money, it's about a R&D race and whoever gets a better product faster than competition.
Therefore Nvidia won't fail, because we're far far away from any AI production mode since the computing for that would need decades to install. Imagine how much compute power you would need for 24/7 assistant inferencing in real time for every person on earth. We're are just scratching the surface. For Nvidia to fail at this time would be the same as that the world would stop on AI research lol.
Imagine, you have an industry where 80% of all companies' / private investors R&D money becomes revenue of a single company. That's basically Nvidia's position in a nutshell.
Basically, Nvidia could do $1 trillion revenue on world wide R&D budgets alone, no need for their customers to make money yet.
And that gives customers the most flexibility. Nvidia dominates training and is highly competitive in inferencing. At the same time, SW improvements speed up single node and networking performance. H100 released 3 years ago is today several times faster than it was on release with constant SW updates.
Customers who buy Nvidia for training today can use the older GPUs from Nvidia for inferencing later. And Nvidia supports even V100 still in SW updates and speed improvements. And since all is based on the same SW ecosystem, it allows for more seamless operations for customers. You can mix different Nvidia GPU clusters but you can't easily mix Nvidia solutions with other solutions.
That is also why Nvidia has always been dominant and that's flexibility. NVFP4 is a good example of what they do to stay ahead. And it is even supported by Hopper so any old customer can use Nvidia's new format to further improve model training performance. Suddenly old Hopper clusters become more valuable with some SW releases by Nvidia.
Nvidia has a track record which no competitor can match. Going with Nvidia is no mistake today while going with any competitor is a risky bet. If you spend billions, you think twice about making bets.
The issue is that the press wants to show how a 8GB new card is slower than a 12GB older card in putting both in higher resolution comparison.
The press doesn't care about thinking or reasoning about product placement and instead wants to always uncover some scandal.
Nvidia clearly has communicated if you want the card for free then these are the conditions. You say yes or no. Nvidia HASN'T forbidden GN and others to buy the card and test independently, right?
There is a reason why Nvidia dominates and that's because when you're paranoid and desperate is when you innovate the most.