Thread regarding Intel Corp. layoffs

DeepSeek is disrupting AI the way ARM disrupted x86.

Open source eventually gets good enough to overcome the first mover advantages of proprietary tech.

This maybe, just possibly makes the Intel and AMD open source AI efforts more valuable, as they are both focused on lowering cost.

At the very least, if this is a reckoning of what appears to be trillions in AI capital misallocation, that won't mean too much to Intel.

Remember, pioneers are the ones who end up with arrows in their back. Sometimes it is better to be a fast follower.

by
| 2251 views | | 20 replies (last January 28, 2025) | Reply
Post ID: @OP+1jjm5yxm1

20 replies (most recent on top)

Fake news to steal your money.

by
| | Reply
Post ID: @cf+1jjm5yxm1

Deepseek runs in ancient H800!!!
If we were to just combine A850 and HBM, we have a winner.

by
| | Reply
Post ID: @c9+1jjm5yxm1

What is DeepS hit AI you are talking about?

by
| | Reply
Post ID: @c1+1jjm5yxm1

AI is inevitable, resistance is futile, get assimilated.

by
| | Reply
Post ID: @bn+1jjm5yxm1

It seems like Blackwell just got Blackballed.
Now on the flip, side if we can keep our cheap GPU momentum in place and can combine it with cheap CPU option, we may come out ahead in the game very soon.

by
| | Reply
Post ID: @bm+1jjm5yxm1

A couple of things.
If better optimized modeling reduces modeling cost by 40 to 50x, then that makes at least that particular modeling able to scale much faster. It probably also reduces the competitive power of Cuda.

If what happens next is a dot Com style crash in Ai data center build out, as it shifts away from Blackwell or just puts projects on pause for a while, the Ai providers, which did OK today, will have improved profitability and customer growth due to the lowered costs.

Hard to see that this anything but positive for adoption, and that growth will be good for more than just nvda.

It really does seem like a tech shift similar to smartphones taking compute away from pcs, combined with a potential for severe price discovery in the first movers, mag7 and the like.

If it revolves into a crash then Intel gets hit along with the rest of tech, but that may not affect the broader economy.

Coming out of that, the service providers such as pltr, crm or snow, should big winners.

Oh, and have little doubt that ccp will sneak in backdoors. They may have even come up with this as a way to take down the mag7.

by
| | Reply
Post ID: @bd+1jjm5yxm1

To put things into perspective, Nvidia lost about SIX Intels today.
Still the top dog though valued at 29!! Intels.

Another way to say that Intel is just irrelevant.

by
| | Reply
Post ID: @b2+1jjm5yxm1

Where this might hurt Intel is if a bursting of the AI investment bubble causes global investors to rebalance from the US to global markets.

Most of the MAG7 return has been due to the rest of the world buying in. That will reverse at some point and this is the sort of thing that could trigger the shift.

In that case Intel eventually follows the market. It just has less to lose, all around.

Fewer AI datacenters also means fewer Xeons

by
| | Reply
Post ID: @ar+1jjm5yxm1

@ah+1jjm5yxm1 Do some goggling.

Aside from what it means for NVDA to fall back to earth, the massive and growing AI datacenter buildout looks like a deer in the headlights at the moment.

The crypto miners getting crushed, as they were just starting to get into AI hosting. TSM also taking a big hit, because a lot of their future volume was for AI.

The reason why the Chinese were not using Gaudi was because they could not buy it. They could buy H100 and so that is what they worked with.

NVDA was trying to build a stronger moat around CUDA with Blackwell. If all that is needed is H100, then switching to Intel or AMD is relatively easy.

The situation reminds me of Spring 2000 for the Dot-Coms and also kinda like what it meant when the first iPhone was released. Once people realized they could do most of their work with a phone, they stopped upgrading laptops so frequently.

All this may not mean much to Intel but it doesn't seem like any kind of negative either.

by
| | Reply
Post ID: @aq+1jjm5yxm1

How did this DeepSeek seem to suddenly appear from Nowhere? Isn’t it generally available to anyone? Might this pop company’s like Nvidia as overblown hype? Do Intel gpus have a similar concept CUDA cores and parallel compute? Do Intel gpus provide a lower acquisition cost alternative compared to Nvidia? So many questions. TIA!

by
| | Reply
Post ID: @ah+1jjm5yxm1

Well the layoffs are coming!!!

Your dreams of finding a job at Nvidia are gone lol

by
| | Reply
Post ID: @af+1jjm5yxm1

The Chinese government is clearly involved and supporting this. They want everyone running on their apps so they can control and spy on the world.

by
| | Reply
Post ID: @ab+1jjm5yxm1

AI’s Sputnik moment?

by
| | Reply
Post ID: @aa+1jjm5yxm1

From Microsoft
https://x.com/satyanadella/status/1883753899255046301

by
| | Reply
Post ID: @a9+1jjm5yxm1

@a7 why aren’t the Chinese using Gaudi or AMD? They’re even cheaper!

They’re using Nvidia for a reason.
It’s the dominant platform - from software to hardware and networking all up and down the stack.

by
| | Reply
Post ID: @a8+1jjm5yxm1

DeepSeek is able to run on the H100 NVDA chip, of which Gaudi is competitive.

Not saying this is a win for Intel but if this open source approach removes the competitive moat that NVDA has with CUDA, then that does create an opportunity for Intel (if they bother to stick with GPU development).

The Chinese company says it developed the software for $5m, which is nothing.

DeepSeek is said to beat some of the existing models and not others. There are many different benchmarks and any model will have some areas of relative outperformance.

by
| | Reply
Post ID: @a7+1jjm5yxm1

@a4 exactly.
Now more firms developing AI algorithms can buy Nvidia GPUs instead of the top 7 mega cap companies. It democratizes AI.

2K GPUs at $5 million opens the door for thousands of companies to develop their own AIs. Instead of billions of dollars of investment from a small number of firms.

by
| | Reply
Post ID: @a6+1jjm5yxm1

Disagree with previous. DeepSeek apparently shows the ability to perform computational AI with far fewer Nvidia cores

by
| | Reply
Post ID: @a4+1jjm5yxm1

Making AI computationally cheaper will expand the demand for Nvidia GPUs.
Now the game is open for even small AI companies.

by
| | Reply
Post ID: @a3+1jjm5yxm1

The leader of the pack may change, yet the lagger does not.

by
| | Reply
Post ID: @a2+1jjm5yxm1

Post a reply

: