they may do another split. It will then be cheap again
I love Nvidia but questioned its P/E ratio yesterday. Today it hits the mark big time and the Nasā drops 2%. I give up!
2 Likes
Even with a split the stock stays the same. I feel there will be a big sell off soon but thatās judt a hunch
We never know itās the stock market after all 
I know you donāt actually think this but there are probably people reading these posts that might so for the avoidance of doubtā¦ a stock split doesnāt change the value of the company, just the number of shares issued.
3 Likes
If NVDA hadnt split last time they would be trading at $1800
psychology who wanst to pay $1800 a share 
.
I imagine a large percentage of that would have been profit taking, the larger investors feeling the frothiness of the stock and taking some off the table.
Some people are questiong why theyāre doing a share buy back as well.
Share buybacks make more when the stock price is undervalued in the eyes of the CEO/Board. Itās a good problem to have when you canāt think of a better use of $25bn.
Apropos of nothing I did find this from Nvidia, their employee stock purchase plan looks amazing. If you joined just 1 year ago and signed up to 10% of your salary, which could easily be $0k, you can spend that on Nvidia stock at $180!
2 Likes
But theyāre saying why would they try bump up a already overvalued stock?
Spending $25bn of the companyās own money boosts the managementās own stock values / makes their options more valuable.
They have very low debt levels and clearly donāt think they can use the cash so returning it to shareholders does make sense. I donāt know if they publish a schedule of the purchase or have stated if theyāll keep them or cancel them.
3 Likes
I think most large publicly traded companies have fairly similar schemes, Nvidiaās just looks particularly good because of the meteoric rise.
1 Like
What do you make of Nvidia @Cameron Youāre one of the few technically knowledge people in these parts. Can they continue to charge $40,000 for these things? Can anyone close any gaps?
I have a share scheme at work which comes with a 20% discount on the closing price. 1.5 years later my option price is 7p lower than share price - might be a bit salty 
Iām not Cameron but hereās my thoughts.
Do you believe AI has a strong future? If so, NVIDIA will continue to grow.
The chip youāre referring to is Hopper. Officially called the H100 tensor core GPU and is sought after by pretty much everyone in the tech sector. H100 is the powerhouse behind generative AI, giving language models like ChatGPT the power needed to process billions of parameters for their output.
Without H100, advancement of AI grinds to a halt. Because of this, theres been a rush to buy the chips. Gulf countries are also scrambling to get their hands on the chips as they ramp up their AI ambitions. In start up hubs, the shortage of these chips has made bosses so concerned about start ups falling behind without them, that they have started buying units on the future companies behalf.
Elon Musk reportedly purchased 10,000 H100 chips as part of his plan to transform X into a super app. And is buying them up as he launches his own verson of ChatGPT called xAI.
These extraordinary maneuvers to purchase chips for generative AI make it hard to overstate just how valuable they are.
Nvidia has also unveiled a next-gen version of the H100 called the GH200, which is in for launch in the second quarter of next year.
The up shot: pay the $40,000 or fall behind your competitors.
4 Likes
What was it they state now a days for computers about deminishibg returns like it should to be the speeds where massive and each computer was outdated within 2 weeks but as we have progressed itās now moved on but with each new realise the speed gets smaller and smaller yet we pay the same or more, I forgot the saying its some law anyways I read it in 2012 so have forgotten.
1 Like
Yeah it was pension craft who done a snippet on the reason why it was weird for them to do a share buy back on a company with a p/e ratio of x y z already, again forgot the exact terms of what he says Iām currently productifly dying with a hangover right now hehe
1 Like
Yes this is almost certainly temporary from a hardware perspective, thereās simply no reason why the design part of the value chain should capture this much value vs say materials, lithography, fab, packaging, validation, deployment or any the other key steps.
That said there is a huge short term spend overhang and for lack of a more efficient way to deploy capital Nvidia is capturing a huge amount of the current compute bottleneck.
But thereās no way this lasts, Googleās TPUv5 is allegedly going to ramp to huge volumes for cloud compute and other hyperscalers are taping our their own ASICs for large volume training and inference so they wonāt pay such a large Nvidia tax forever.
Itās also been suggested that when we get to the truly gargantuan training runs labs might start taping out specialized ASICs for a particular model/training run. This is plausible because even $10m in costs to design & tape out a specialized chip could be worth it on a $1bn run if itās slightly more efficient, although thatās pretty speculative.
I think the bet with Nvidia at the moment is their cloud offering itself - if they can rent out the picks and shovels to AI labs then they could continue to capture a lot of the ongoing spend even when hardware design margins drop off.
2 Likes
Good point. Generally startups (perhaps most companies) will not be buying this hardware - they canāt. Originally, I couldnāt even get a quote on a 8-GPU HGX server. So I took the route of going via a private cloud (which itself worked with a data centre). Still our lead time was many months and even then we only got what we wanted in chunks.
1 Like