Tech Stocks Brace for a DeepSeek Haircut, NVIDIA Down 12% in Pre-market Trading

[freeagent]

It would be cool if the AI world as well as gamers turned their back on Nvidia as users migrate to AMD and Intel.

Ooo the irony of it all.

 

[TumbleGeorge]

Please what is percentage fall between 24 and 27 January?

 

[MacZ]

Real points of interest (and not the chaos of the stock market) :

- I was informed that chinese people can only copy superior western tech, but somehow they made something better.
- This while being sanctionned for the (express but unacknowledged) purpose of hindering their developpement (especially in AI).
- Free better chinese model or $200/month 'OpenAI' model ? One has to wonder how 'OpenAI' is expecting to make money now. Borrowing money is not free anymore.

 

[R-T-B]

This isn't going to "bring nvidia back to gaming," guys. NVIDIA has always made more in just enterprise then gaming even before the AI era. I hate to break it to you, but this is more likely to make then leave completely rather than come back.

 

[nguyen]

This isn't going to "bring nvidia back to gaming," guys. NVIDIA has always made more in just enterprise then gaming even before the AI era. I hate to break it to you, but this is more likely to make then leave completely rather than come back.

Hate to break it to you but Nvidia still pays more attention to gaming sector than their competitors do.

I mean who introduced new GPUs and new gaming related features during CES?

 

[TumbleGeorge]

new gaming related features

New features? Or new versions of already existing?
Ps. In fact, is it even correct to call them features?

 

[nguyen]

New features? Or new versions of already existing?

Mega geometry and neural rendering are entirely new

 

[kondamin]

The AI powered CoPilot O365 growth?

If CoPilot only costs 50cents/month to replicate instead of $50/month to run, will Microsofts O365 really be able to get the growth they want out of AI enabled Office?

If it only costs 50cents/month to run CoPilot then that's also 100x less Azure needed to run everyone's O365.

And if it costs 100x less to make OpenAI like LLMs, then Microsofts rather sizable investment into nuclear power plants (https://www.npr.org/2024/09/20/nx-s1-5120581/three-mile-island-nuclear-power-plant-microsoft-ai) is looking like a waste.

--------

Or to put it differently: if AI is going to be 1/100th the cost than expected, how much is Microsofts $13 Billion bet on OpenAI worth right now? Maybe only $0.13 Billion if the competitions software really is this much better.

Or maybe these DeepSeek people are lying. But either way, people are asking questions today and reevaluating 365 and Azure and how it relates to AI.

On the contrary, I think a lot of users have or are planning to dump their office subscription because the recent price hike.
that 30% price hike isn’t as easy to absorb in current economic conditions so Microsoft might do well reversing the prices changes and retain their user base For something they do not use.

Also of note is that it’s not that unlikely that even the more efficient way of doing things is going to balloon in to using up what ever is being thrown at it and offer a better experience on bigger hardware.

imdont know if that is going to calm the market, as that is irrational and mostly done by bots

 

[Tartaros]

Real points of interest (and not the chaos of the stock market) :

- I was informed that chinese people can only copy superior western tech, but somehow they made something better.
- This while being sanctionned for the (express but unacknowledged) purpose of hindering their developpement (especially in AI).
- Free better chinese model or $200/month 'OpenAI' model ? One has to wonder how 'OpenAI' is expecting to make money now. Borrowing money is not free anymore.

The best thing I read about this is this twitch message, it's chef's kiss:

 

[dragontamer5788]

On the contrary, I think a lot of users have or are planning to dump their office subscription because the recent price hike.

Well yeah. Microsoft spent $Billions buying OpenAI and investing into nuclear power plants to support more training. The money needs to come from somewhere, and price hikes for Office are the natural spot. Its not like people are buying the CoPilot subscriptions.

Microsoft is deep into AI investments. EVEN in its traditional Office365 product line.

Now what happens economically from here on out is a mystery. But if people are feeling squeemish about current economic conditions, they should have dropped MSFT as much as any of the other AI stocks.

 

[kondamin]

Well yeah. Microsoft spent $Billions buying OpenAI and investing into nuclear power plants to support more training. The money needs to come from somewhere, and price hikes for Office are the natural spot. Its not like people are buying the CoPilot subscriptions.

Microsoft is deep into AI investments. EVEN in its traditional Office365 product line.

Now what happens economically from here on out is a mystery. But if people are feeling squeemish about current economic conditions, they should have dropped MSFT as much as any of the other AI stocks.

Well that’s where it’s not people that are trading it’s bots that are trading on behalf of them.

indicators mark Msft as safe, as it is safe they have a good revenue and their capex has been reasonable so the bots leave them be.

 

[Rightness_1]

I really hope this is as good as is being implied! Nv needs its bubble burst!

 

[Vayra86]

The AI powered CoPilot O365 growth?

If CoPilot only costs 50cents/month to replicate instead of $50/month to run, will Microsofts O365 really be able to get the growth they want out of AI enabled Office?

If it only costs 50cents/month to run CoPilot then that's also 100x less Azure needed to run everyone's O365.

And if it costs 100x less to make OpenAI like LLMs, then Microsofts rather sizable investment into nuclear power plants (https://www.npr.org/2024/09/20/nx-s1-5120581/three-mile-island-nuclear-power-plant-microsoft-ai) is looking like a waste.

--------

Or to put it differently: if AI is going to be 1/100th the cost than expected, how much is Microsofts $13 Billion bet on OpenAI worth right now? Maybe only $0.13 Billion if the competitions software really is this much better.

Or maybe these DeepSeek people are lying. But either way, people are asking questions today and reevaluating 365 and Azure and how it relates to AI.

Still doesn't change the fact MS has a core business that will float on a non-AI world just fine, and there is no conclusive evidence CoPilot is worth anything. Its just a gamble like all AI in the consumer space. Throwing shit at walls. Basically, either everyone has AI and its not a USP, or nobody does and its also not a USP

 

[LittleBro]

It's not a loss unless you cash out.
But the big problem with many absurdly overvalued companies(Nvidia, Apple, Amazon, MS, Alphabet, Meta, Tesla, etc.) is that if a tiny fraction of the total shares where put on the market the market value would plummet.
(That's why you should probably not invest directly in any such company, but rather through funds…)

This. Go for funds. Still risky, but much more secure than pouring all the money into 2-3 stocks.

Real points of interest (and not the chaos of the stock market) :

- I was informed that chinese people can only copy superior western tech, but somehow they made something better.
- This while being sanctionned for the (express but unacknowledged) purpose of hindering their developpement (especially in AI).
- Free better chinese model or $200/month 'OpenAI' model ? One has to wonder how 'OpenAI' is expecting to make money now. Borrowing money is not free anymore.

When USA restricts China from access to neural accelerators stuff, it's not gonna prevent China's technological advancements, it will just slow it down.
At one time point China's advancements will start to have edge over US technologies. It's a matter of time. China can do everything faster and cheaper. Maybe this is the time.

Gardening analogy - when you grow two same plants and you keep putting water with fertilizer into first one and water only into second one, the latter will grow slowly but the plant will have much better resistance to sudden weather changes, droughts and other external impacts. You may even cut the 80% of the plant off and it will regrow. Slowly, but surely. This means the 2nd plant can manage it's resources much more better/wisely and can even prepare for sudden weather changes, should you give the plant enough time.

Hate to break it to you but Nvidia still pays more attention to gaming sector than their competitors do.

I mean who introduced new GPUs and new gaming related features during CES?

And who already has stocks of upcoming GPUs in stores prior to launch and who did paper launch only? Huh?

 

[Hecate91]

Hate to break it to you but Nvidia still pays more attention to gaming sector than their competitors do.

I mean who introduced new GPUs and new gaming related features during CES?

Except less than 10% of their gaming segment is actual gaming GPU sales, the rest is AI.
And you mean those new GPU's which aren't much of an improvement in gaming?

 

[TumbleGeorge]

Next strike from DeepSeek. Janus Pro is released.

 

[R0H1T]

Basically, either everyone has AI and its not a USP, or nobody does and its also not a USP

Well, there's always a third option: What if they're pretending it's a "USP" or whatever & stashing the real "AI" for doomsday?

Spoiler

 

[AnotherReader]

This isn't going to "bring nvidia back to gaming," guys. NVIDIA has always made more in just enterprise then gaming even before the AI era. I hate to break it to you, but this is more likely to make then leave completely rather than come back.

No, they only started making more money in enterprise than gaming when AI took off. Even at the time of Ampere's release, datacenter revenue was slightly less than gaming revenue. If we go further back to the end of the Maxwell era, datacenter revenue was a measly fraction of gaming revenue.

 

[phanbuey]

So - for sure this is an open model... I just asked it all about 1989, and why this is a restricted topic in China and here was the response:

This is the 14B param model running locally.

No way Chinese officials would publish this or allow distribution if they see/realize this. The bureaucracy hasn't caught up to the tech.

 

[TumbleGeorge]

So - for sure this is an open model... I just asked it all about 1989, and why this is a restricted topic in China and here was the response:
View attachment 382206

This is the 14B param model running locally.

No way Chinese officials would publish this or allow distribution if they see/realize this. The bureaucracy hasn't caught up to the tech.

In your opinion, something that is a fact, that is, has already happened, cannot possibly happen? Send me some of what you smoke, it's very good, it breaks all logic.

 

[R-T-B]

Hate to break it to you but Nvidia still pays more attention to gaming sector than their competitors do.

They do, but if money gets tight, don't expect that to continue. This is mostly a "gimme" granted by the CEO to gamers as thanks for jump starting them, not a smart economic move.

No, they only started making more money in enterprise than gaming when AI took off. Even at the time of Ampere's release, datacenter revenue was slightly less than gaming revenue. If we go further back to the end of the Maxwell era, datacenter revenue was a measly fraction of gaming revenue.

View attachment 382143

Those aren't just before AI tookoff, but also before all of of the crypto changes which is where enterprise took over. I'd need newer figures to even begin to comment. I was commenting on the market now, not yesteryear.

 

[phanbuey]

In your opinion, something that is a fact, that is, has already happened, cannot possibly happen? Send me some of what you smoke, it's very good, it breaks all logic.

What? You're clearly unfamiliar with how Chinese censors work. No way it would happen IF the bureacracy was aware of what they were publishing.

This was not on purpose/not home grown.

 

[AnotherReader]

They do, but if money gets tight, don't expect that to continue. This is mostly a "gimme" granted by the CEO to gamers as thanks for jump starting them, not a smart economic move.


Those aren't just before AI tookoff, but also before all of of the crypto changes which is where enterprise took over. I'd need newer figures to even begin to comment. I was commenting on the market now, not yesteryear.

The figures for FY 2021, the year that saw the launch of Ampere, are well after the Crypto boom of 2017. Now, of course, datacenter is in the driving seat and that is almost entirely due to AI.

Note that the datacenter revenue in the table above also includes Mellanox.

according to NVIDIA, Mellanox contributed about 10% to the company’s FY2021 revenues

 

[R-T-B]

are well after the Crypto boom of 2017.

That's right in the heart of the second crypto boom, the ethereum one.

AI doesn't really become a factor until a year or so post COVID. So no, it's not all blamable on AI.

 

[efikkan]

This. Go for funds. Still risky, but much more secure than pouring all the money into 2-3 stocks.

It's always the first thing people bring up with investing in funds; the emphasis on risk, like it's disproportionally more risky than other types of investments when it's actually the other way around. (Even your money in the banks have risk, hint: inflation) The biggest risk by far with funds is your own stupidity (panic selling). But if you understand the immense power of compound interest, pick funds with a solid track record, and know how to withdraw correctly, then nothing but of a total collapse in the world economy, governmental confiscation or fraud from the fund itself should ruin your success. Even if the market suddenly drops 30% or more, it will recover, and even if you have to withdraw at the worst possible time, the chances of you losing more than your initial investment is shrinking rapidly for each year. Compared to the alternatives, incl. investing in properties (and don't get me started on precious metals), the risk of solid funds are very low.

The tabloid coverage of economical topics unfortunately leads to lots of amateur investors dumping their investments and the professional ones swooping in to buy at a bargain price. Keep in mind the news are creating havoc for a earth-shattering -3.07% for Nasdaq Composite and -1.46% for S&P 500 indices, more than half of which recovered in a single day, and this is only about half the gains this month. And do you remember the "tech collapse" back in August last year? Well even the indices have gained about 10% on top of that recovery (and some managed funds even better than that). Unfortunately the only ones who lost were the ones who panicked. Meanwhile the youth is pouring their money into worthless crypto scams and piling up generational debt.
This is what we get for not educating people on personal economics.

-----

As for Nvidia, I wouldn't dare to speculate when their dominance in the "AI" market will end, but it's coming, and I just hope they course correct in time to leave a relevant player in the area of graphics. (They should probably split their "AI" products into a separate company at some point.)