Nvidia - Semiconductors and AI

I find myself returning constantly to the topic of sovereign states getting into the ai/datacenter business. To me, and I emphasize I am a complete rube and know little about the semiconductor or datacenter industries, if the leading nations of the world all feel it is important to their security to have their own ai infrastructure, then it takes on the flavor of an arms race, no? If each new generation of chip is 2x as powerful, and uses 1/2x the power, nobody can afford to be left behind. Which means there are now 5, 10, 20? very deep pockets that have been added to the customer base (much better resourced than google or amazon, btw), and none of those new customers will want to see their neighbor getting a leg up on them.

And has been pointed out elsewhere, right now we have a short list of areas where we envision ai being useful. That is certain to expand into very unexpected areas as people (and governments, alas) begin to use it.

This is getting curiouser and curiouser…

17 Likes

I alluded to sovereign AI data centers in my previous post. As for applications, it’s hard to think of any governmental function or agency that could not benefit from AI. If nothing else, governments, all governments have enormous record keeping functions. In many cases the data is manually captured and input to databases, often redundant in nature, highly error prone as implemented on outdated technology. As mundane as it seems, AI could greatly improve efficiency, accuracy and privacy.

Then there’s applications related to defense, investigations, tax collection, fraud detection and many other functions that could be greatly enhanced via AI.

I’ll provide a non-US example. I travel to China on an annual basis because my wife is Chinese. When you enter the country as a foreigner your photo is taken. If you’re a Chinese citizen, your photo is recorded as part of you hukou record (sort of like our social security system).

China has a population of about 1.3 billion, I’m not sure how much foreign visitors contribute to the photo database, but I’m sure it’s significant. All these facial photos have been digitized with unique identifying keys registered. Any person’s name and address can be retrieved within minutes if not seconds based of their photo.

Here in the US we would consider this an outrageous invasion of privacy by the government. However, in China, first of all no one thinks that there is any sort of right to privacy, it’s just not a thing. Secondly, most people view this as a very effective deterrent of crime because their are video cameras quite literally everywhere. At least everywhere in the cities and even in the countryside, just not so many. The identification and information retrieval wouldn’t be possible without AI.

16 Likes

Here here! Good post monika and I share your level of optimism.

For now, the build out continues in an AI arms race amongst the tech giants. (And this should continue as NVDA iterates over the next few years, as Gaucho said).

Count me amongst those disappointed that this AI wave is not moving deeper into the data layers yet like SNOW and MDB.

But AI-driven solutions are and will emerge from here to give us opportunities. Keep watching.

-muji

39 Likes

While I do not disagree at all with the actual take on AI, nobody was being impatient.

I was curious if anyone else saw the point in investing in the likes of META, AMZN, GOOG, MSFT, even AAPL from here, beyond Tesla and Nvidia. Apparently not, and that’s fine. But in between OT macro reasons and Big Tech rush to AI, I think market caps north of 10T for Big Tech are a matter of several years, but not decades.

So while one waits for winners among the less gargantuan market caps to emerge, one can invest in Big Tech (and cloud infrastructure that everyone needs).

And the benefit of that is likely getting a very high win rate (say 5 out of 7) with practically no risk (safer than bonds, as a group, IMO), and very solid returns (above QQQ).

My 2c.

18 Likes