Datacenter Processors for AI is already 78% of Nvidia's Revenue | Hacker News Reader

Datacenter Processors for AI is already 78% of Nvidia's Revenue

visualcapitalist.com

57 pointsbratao20 comments

When I was a financial auditor, one of the tests on revenue we did was a concentration test, since having a large portion of your revenue coming from a few customers is a risk to business continuity. Wonder how Nvidia folks are thinking of mitigations for that.

If it wasn’t apparent already that nvidia is overpriced. This is the nail in the coffin.

As soon as the AI bubble pops (ie, current generation AI fail to meet unreasonably high expectations), nvidia will come crashing down fast.

OpenAI ceo pursued a $7 trillion dollar ask for investment in semiconductors.

https://www.wsj.com/tech/ai/sam-altman-seeks-trillions-of-do...

VR, crypto, real estate. All of the hallmarks of a bubble exist with the current phase of AI

There seems to be a non zero chance that AI is a transformative tech a la the internet. Then the bear case becomes is Nvidia the same as Cisco in 1999? Did Cisco have a moat like CUDA? Will the AI boom be like 1999? Do you compare p/e, forward p/e or p/s which all paint a different picture of how similar the stocks are.

I would suggest it’s not totally apparent what will happen with the Nvidia stock price, especially as the ramifications for the world could be so huge if AI is as transformative as speculated. It threatens the stability of many other markets / stocks.

Time will tell!

While time will tell, if "AI" brings us a beneficial application, you can safely bet against a lot of industries, since the destructive uses are already proven. So far AI excels only in human mimicry, which strongly enables abuse and exploitation, but has little legitimate value.

Spam and fishing will break basically everything without authentication. Youtube and Google search are already half broken because of AI spam. Dating Apps will break, comment sections will break, the web is gonna break for the most part... Not looking good for classic telecommunication either, since you can't trust messages and calls anymore.

Education, data analysis and coding are applications showing real benefits, especially as supplementaries to the existing practices.

I think spam is a problem that will be fixed due to it's importance. There are real potential uses of blockchain technology here. I'm super interested in how the banking system KYC stack could be used in conjunction with NFTs to provide proof of factual claims.

Alot of dating apps, especially Tinder, are already broken. The number of legit profiles vs. the number of fake ones is probably like 10-1. Zoosk at least tries.

Cisco was a very similar situation actually. Lots of "routers" out there (way more than other brands making GPUs), but the cisco stuff just "worked" in corporate environments and had the best interop with the rest of the network.

Ultimately though people figured out the other stuff and while cisco is still very popular for core routers they are no longer the only reasonable option.

The stock has only just now recovered 25 years later from the .com crash.

Nah he didn’t really do that, did he? Hehe

I wonder if this years crypto bull run will even register.

How high was reduction of money stream to GPU miners after ETH went proof of stake?

4 years ago even Intel wanted to release chip dedicated to mining. But it missed the peak and abandoned the project, I believe. This time AI swamps everything.

This highlights an interesting thing: while AI is in the news constantly these days, we should really be concerned about what is NOT in the news regarding AI...

Can you elaborate what you mean by that?

My point is that AI is an exceptionally powerful tool that can shape society by creating a huge amount of information that may be detrimental to society. And, much of this content creation will happen behind closed doors without much fanfare at all. Only, we will be exposed to it and modified by it. While the news and media will advertise the popular stories, what really concerns me is how much of this is being done without our knowledge and how much it can change society without us knowing a thing about it.

Does anyone understand what "data center processors _for analytics_" mean?

Numerical crunchers that can take in vast amounts of multi channel data .. non-AI analytics is lilely a small percentage of the total volume now .. but it's a similar type of operation, form high dimensional data representations and dimensionally reduce to find patterns, form predictions, or create "running kernals" that can sharpen features of future incoming streams.

Theoretically it means you wouldn't need much manual data engineering. "Just throw AI at our data problem and give me interesting/actionable info"

I'd even simply this a bit - can you have the AI translate natural language into SQL queries?

ChatGPT can already do this at least for certain fairly simple cases.

The idea that AI will get this right (for interesting / realistic use cases) sufficiently often to take the human out of the loop seems optimistic pre-AGI. Most humans I’ve tried to help with data analysis don’t know what they want, don’t know how to get it, don’t know how to interpret it, and don’t know how to check if their analysis is correct.

(Fortunately, mostly they only want to use the data to affirm their preexisting decisions, so it’s actually fine).

GeForce vs everything else

They should sell their gaming division to Valve.

Or buy valve and have them lead the gaming division

Lol, yeah, my math was a bit off.

"According to Bloomberg, Valve is valued at $6.9 billion in March 2024."

"Nvidia market cap = $2.23 trillion"

100% - 78% = 22%

$2.23 trillion * 22% = $490.6 billion

And, $490.6 billion > $6.9 billion.

So yeah, kudos to my math.

Reply on news.ycombinator.com