Replies to post #12896 on NVIDIA Corp (NVDA)

[ShawnP123]

You have obviously misinterpreted my post. Nvidia is coming out with the Rubin next but you're missing the point.

Google, Meta, Microsoft, and Amazon are also coming out with Chips which will be adequate for a number of applications thereby reducing their dependence on Nvidia's total product line. Yes, they will still need the Blackwell but certainly not the H100. China has already forced a lower price on the H20 series due to the competition from Huwai. You have also not considered the following:

There is limited production in the ability to produce advanced chips without huge investments in increasing production. TMS isn't going to put all their eggs into one basket. BTW, Apple has currently tied up TMSC's production on their most advanced line which utilizes 2nm technology. There is also huge investment needed to to increase production of the wafers needed to make the chips. There are only about 5 companies located in 3 countries which can produce the highest purity and minimal defect wafers. Don't forget the last chip shortage problem was based on the availability of wafer supply which went out as long as 58 weeks or more.

My post was more about looking at other ways to achieve more efficient ways to achieve AI success. Collecting data is the most common but then there are the royalties which are involved in a number of applications.

Overall, I think you should do more investigating and it's not as simple as you think.There's a saying- "with enough money and time one can do pretty much anything. Well time my be unlimited but money isn't.

[ShawnP123]

Intermediate to long term Chips aren't going to be the answer. There are limitations based on physics and availability of the highest and defect pure wafers. I think you should look at how and what's involved in making those and who makes them. I've been through two semiconductor plants (one in the US and one in Korea) but unfortunately not any plants which grow the wafers. If one is looking to the next step, it will most likely be Quantum mechanics and not silicone which generates data at quantum speeds. All that is needed for that to happen is improving error correction by 100 x's. I find it hard to believe there is a lack of understanding as to why Nvidia's biggest customers are having their own chips built and that it won't affect Nvidia. Not everyone or every application requires the latest. Yes, Nvidia is looking more at systems and software which is good but not everyone will be tied to that. So again competition forcing lower margins will again be an issue. Google is a good example.
BTW Google is involved in Quantum computing and is doing better in error correction.