Jensen dong

[Purgatory]

Trump probably had something to do with it lol

It would not surprise me that he was setup to fail and later expose the other affairs.

He does want Tim Cook of Apple to be CEO.. Maybe because he can control him..

 

[opethdisciple]

To be fair Kimberly Powell is hot.

 

[Purgatory]

Reddit

www.reddit.com

Reddit

www.reddit.com

If Tim Cook ran Nvidia...

 

[TNA]

Some of these videos crack me up. Lol

 

[scubes]

 

[scubes]

how many has he pulled now ive lost count

 

[Kelt]

how many has he pulled now ive lost count

Jensen's huang knows no limits or boundaries.

 

[KompuKare]

Ooof, if this stuff is true I think we're just getting started with the Jensen revelations.

https://bsky.app/profile/cdemerjian.bsky.social%2Fpost%2F3n27ohomkwc2f

Charlie should be glad that back when he exposes the solder defect "bumbgate", Nvidia was not yet too big to fail and considered a national security thing as a similar expose now would probably have three-letter agencies knocking at his door!

A lot of it seems iffy, A guy who is retired CIA commented on this saying "Keep an eye out on the people replacing him in the interim and permanently, Then look at their connections to external groups, That's why Huang was ousted" will be interesting to start noticing patterns.

This is possible. National goverments, political and security players (either American or non-domestic) are very much into playing dirty.

NVIDIA Allegedly Engineered Scarcity By Funding AI Startups To Buy Its GPUs

Following the departure of now-former NVIDIA CEO, Jensen Huang, a new report by The Wall Street Journal (via Wccftech) has emerged alleging that NVIDIA artificially inflated the ongoing GPU and memory shortage by investing in AI startups that would go on to buy NVIDIA GPUs and build data centers....

www.techpowerup.com

Nvidia has been playing the world's biggest game of channel stuffing. Invest in AI companies (and not just startups) on the condition that money is used to buy GPUs. Even if they don't need the GPUs at that time it's still attractive because news of an NVidia investment attracts other investors.

Companies that don't need investment get offered priority access to GPUs by Nvidia, on the condition they buy now and commit to buying more over the next few years. Again, that means customers buying GPUs the may not need.

Result: artificial scarcity, rising prices and NVidia's revenue skyrocketing.

If the AI industry had kept expanding and finally figured out how to make money, this may well have worked. But now there are going to be warehouses full of GPUs and AI servers that can't be deployed, owned by companies that may have to sell them to bring in some cash in the near future. As soon as that happens the market falls off a cliff.

And for those who though cartel behaviour required shady meetings on gulf clubs... Well for DRAM with only 3 players we have known that was no longer the case as long as they could each accurately read what the other would do.

For GPUs, it turns out (where reading was easier as there was only NV and AMD - and AMD are really predictable) you don't even need any competitors to play cartel: seems with some very creative accounting you can create your own "shortage" all by yourself!

I honestly believe they will exit gaming, They are actively losing money with gaming, With a "gaming" GPU they can slap extra memory onto, Call it "Pro" and sell it for 5 times the amount to the AI crowd. Gaming right now as a revenue source makes no sense IMO.

"Loosing money" is going a bit far IMO. They are just not making as much as elsewhere. So 60% margins rather over 90%+, I guess?

Strategically, aside from maximise sales by any means (corporate 101), I think a big part of selling to China is that if they did not then the CUDA moat would breach. If China can no longer "obtain" Nvidia GPUs then the efforts China start to wean AI researches away from CUDA would become a flood. PRC has pretty much made AI one of their primary goals.

Ironically the huge strides they have made to de-carbonise - mainly for energy security and to clean up their local environment - will be held back by not being able to use TSMC because 7nm class DUV-made AI chips will consume more power. As they are so good at big project and planning, I expect they may shift all inference workload to DUV chips and then repurpose all the GPUs they currently have for training workloads. Eventually CXMT will make HBM and along with HiSilicon (Huawei) will probably results in some designs for training but will surely take longer.

If you think you're getting a new generation of cards anytime soon, think again. Nvidia is notorious for not releasing new cards until all the old cards sitting in stock are all sold off, famously ending production runs months before the cards officially become EOL to maintain artificial scarcity and sky high prices. With hundreds of billions of stock sitting unused in warehouses, even Amazon has them, there won't be any new cards for the foreseeable future as Nvidia does not dump old stock onto the market at reduced prices thats not the Nvidia way. Besides all those data centres have got to see at least several years of usage to recover some of their cost before they think about upgrades and thats Nvidia's market now gaming revenue doesn't even register on their official sales figures anymore.

Well, data centre GPUs are not really re-usable so slightly different to their scarcity game with consumer GPUs. For AI data centres, the same power inefficiencies which made DUV-made chips unviable will eventually mean they will have to upgrade though.

Some of the Jensen hints about having a surplus is about wafer commits though. While their AI GPUs - or any of their designs - are not modular (like Ryzen and EPYC) and hence they cannot repurpose them after they are made, for future wafers they could do some "low-margin" stuff GB102/GB103 etc. Well, if they can find any memory for those. Oh, and admit the bubble has, if not burst at least deflated somewhat.