Zero → Billion $ Business opportunities
Jensen Huang 10/16/23 interview by Aquired (where he positions Nvidia is at the 47 minute mark).
My notes:
CUDA (compute unified device architecture) was UDA (The Network is the platform), from Day 1. After 30 yrs, all their chips are fully compatible and can be unified with infiniband (Mellinox was their greatest aquisition of all time.). As we become more productive, with AI, Enterprises will want to expand (never ending ambition) and therefore hire more people to get there. By elevating chip design to a high level language Nvidia succeeds. (He goes,on to say that LLMs are assisting in this and sounded to me like , perhaps obviously, Nvidia is advancing composability through the use of AI in house.) Not a chip company, Nvidia’s market cap is unlimited. Huang’s habit of always going back to first principles of physics, Jensen Huang reminds me of Elon Musk. Because Jen Huang prefaces his opinions with principled morality, I just like Huang much much more.
With 20% of Nvidia sales in China, and possible export restrictions ‘around National defense issues’, maybe there’s less ‘on allocation’ going forward.
I’m looking to add to my 7% position,
Best
Jason
18 Likes
And this morning, the administration announced it plans to halt shipments of “more advanced AI chips designed by Nvidia and others” … Reuters
4 Likes
In reference to the Above interview, in my notes I wrote:
“He goes on to say that LLMs are assisting in this and sounded to me like , perhaps obviously, Nvidia is advancing their architectural designed in composability through the use of AI in house.”
In reference to a recent research publication by Nvidia, Nvidia tests chatbots in chip design process in bid to use more AI
Here is some clarification on some of the ways Nvidia is fine tuning LLMs to advance their business.
Designing the millions of transistors into an improvement on the past, requires huge amounts of capital. The key finding, in Nvidia’s recently published research into how best to use AI within Nvidia, was that a chatbot, fine tuned by Nvidia’s 30 years of experience, could become more accurate and save more money than a Senior chip designer could, in answering questions from Junior chip designers. The article goes on to list other ways Nvidia is using these chat bots to ‘significantly helping control the cost of future systems’.
I see this as at least a few data points, pointing into the direction of how Nvidia will only grow their lead going forward.
Nvidia now 14% of portfolio,
Best
Jason
21 Likes