Intel on the Brink of Death

[XYang2023]

Do you agree with his analysis, sell Intel Products to finance IFS?

I disagree:

1. It is very risky. The main issue is that TSMC holds a strong monopoly position in the market.
2. The usefulness of generative AI is still up for debate. Based on my daily experience working with it, the current approach lacks reliable reasoning capabilities, making it difficult to meet people's expectations. For most everyday uses, AI-enabled PCs are more than sufficient.
3. Although the market share of x86 for client architecture might be shrinking in the future, other segments can compensate for the numbers. The goal should be to achieve a profitable margin.

Intel on the Brink of Death | Culture Rot, Product Focus Flawed, Foundry Must Survive

Board Short Circuits, Cultural Rot, x86 Has No Moat, Sell PC Business, Nvidia's PC CPU Push, Roadmap Review

semianalysis.com

 

[Rumenovic]

Wonder why they think Clearwater Forest/Diamond Rapids is uncompetitive considering it would compete against the current Turin Dense as AMD probably has no answer until 2026.


Also Granite Rapids and Sierra Forest might not cut it but they are certainly in a much better position now than they were before.
These current and future performance ratios between AMD and Intel in the datacenter space were known for quite a while, why make this point now?
Wouldn't it have made more sense to make those points back when Intel was WAY less competitive in both mobile and datacenter ?

 

[XYang2023]

Wonder why they think Clearwater Forest/Diamond Rapids is uncompetitive considering it would compete against the current Turin Dense as AMD probably has no answer until 2026.


Also Granite Rapids and Sierra Forest might not cut it but they are certainly in a much better position now than they were before.
These current and future performance ratios between AMD and Intel in the datacenter space were known for quite a while, why make this point now?
Wouldn't it have made more sense to make those points back when Intel was WAY less competitive in both mobile and datacenter ?

I think Intel needs to keep up with the competition and remain profitable to fund future products.

 

[siliconbruh999]

Wonder why they think Clearwater Forest/Diamond Rapids is uncompetitive considering it would compete against the current Turin Dense as AMD probably has no answer until 2026.

Clearwater forest and Diamond Rappids haven't launched so how can he stipulate they are not good

Also Granite Rapids and Sierra Forest might not cut it but they are certainly in a much better position now than they were before.
These current and future performance ratios between AMD and Intel in the datacenter space were known for quite a while, why make this point now?
Wouldn't it have made more sense to make those points back when Intel was WAY less competitive in both mobile and datacenter ?

Exactly

 

[Xebec]

Do you agree with his analysis, sell Intel Products to finance IFS?

I disagree:

1. It is very risky. The main issue is that TSMC holds a strong monopoly position in the market.
2. The usefulness of generative AI is still up for debate. Based on my daily experience working with it, the current approach lacks reliable reasoning capabilities, making it difficult to meet people's expectations. For most everyday uses, AI-enabled PCs are more than sufficient.
3. Although the market share of x86 for client architecture might be shrinking in the future, other segments can compensate for the numbers. The goal should be to achieve a profitable margin.

Intel on the Brink of Death | Culture Rot, Product Focus Flawed, Foundry Must Survive

Board Short Circuits, Cultural Rot, x86 Has No Moat, Sell PC Business, Nvidia's PC CPU Push, Roadmap Review

semianalysis.com

Agree in whole, minor item to note on GenAI -- there may be a tipping point where it's more important to have AI capability on the PC (and phone) than it is in the data center. This has happened before (Mainframes -> PC, Large scale compute challenges --> onboard GPU, etc).

This isn't likely at all in the next 3-5 years IMO but later down the road, but after a few AI boom/bust cycles, the local AI strategy might be good in the long term. Note this favors Intel products over foundry too.

 

[soAsian]

Do you agree with his analysis, sell Intel Products to finance IFS?

I disagree:

1. It is very risky. The main issue is that TSMC holds a strong monopoly position in the market.
2. The usefulness of generative AI is still up for debate. Based on my daily experience working with it, the current approach lacks reliable reasoning capabilities, making it difficult to meet people's expectations. For most everyday uses, AI-enabled PCs are more than sufficient.
3. Although the market share of x86 for client architecture might be shrinking in the future, other segments can compensate for the numbers. The goal should be to achieve a profitable margin.

Intel on the Brink of Death | Culture Rot, Product Focus Flawed, Foundry Must Survive

Board Short Circuits, Cultural Rot, x86 Has No Moat, Sell PC Business, Nvidia's PC CPU Push, Roadmap Review

semianalysis.com

1. some say Intel is a monopoly until they f'ck themselves. whose to say that TSMC is the king, forever?

2. I'm excited for AI but not for all the buzz. the fact to the matter is GPU is really good at processing data and more money goes into making/developing "AI chips" is good for the industry(?)

 

[Daniel Nenni]

A little over dramatic as usual. Here is my 2 cents:

What is Wrong with Intel? - Semiwiki

One of the most popular topics on the SemiWiki forum is Intel, which I understand. Many of us grew up with Intel, some of us have worked there, and I can say that the vast majority of us want Intel to succeed. The latest Intel PR debacle is the abrupt departure of CEO Pat Gelsinger.…

semiwiki.com

 

[siliconbruh999]

Clearwater Forest/Diamond Rapids的竞争更多是分析当下市场需求与产品路线图的契合度，而Granite Rapids和Sierra Forest的改进则是Intel重建市场信任的关键一步。市场对性能比的关注也体现了技术竞争的本质，如果Intel能够持续提升性能、效率和生态，仍然有机会重新夺回数据中心市场的主动权。

Translation for our understanding
The Clearwater Forest/Diamond Rapids competition is more about analyzing the fit between current market demand and the product roadmap, while the improvements in Granite Rapids and Sierra Forest are a key step for Intel to rebuild market trust. The market's focus on performance ratio also reflects the nature of technological competition. If Intel can continue to improve performance, efficiency and ecology, it still has the opportunity to regain the initiative in the data center market.

 

[count]

Intel needs to split. Whether that means sell the design side or sell the foundry side is going to be a hard call. I think sell the design side is probably more viable because there are more potential buyers, but the buyer would need to make a 5-10 foundry commitment like AMD made to GloFlo.

 

[siliconbruh999]

Intel needs to split. Whether that means sell the design side or sell the foundry side is going to be a hard call. I think sell the design side is probably more viable because there are more potential buyers, but the buyer would need to make a 5-10 foundry commitment like AMD made to GloFlo.

It will kill both the design and product and they will loose the scale

 

[hist78]

Intel needs to split. Whether that means sell the design side or sell the foundry side is going to be a hard call. I think sell the design side is probably more viable because there are more potential buyers, but the buyer would need to make a 5-10 foundry commitment like AMD made to GloFlo.

It will kill both the design and product and they will loose the scale

If Intel split into two independent companies, foundry and product design, I believe at least one of them will survive. If they don't do it quickly now, Intel product design and manufacturing divisions may go down together and to be sold in pieces.

Sorry, it may hurt many people's feelings, mine too.

 

[count]

It will kill both the design and product and they will loose the scale

Yet somehow AMD is doing well with even less scale, and GlobalFoundries doing ok as well.

This is monopolist thinking, "If we aren't the biggest what is our advantage?" the answer is they will have to be scrappy like a start up and not just use their size as a bludgeon.

 

[tedc]

Yet somehow AMD is doing well with even less scale, and GlobalFoundries doing ok as well.

This is monopolist thinking, "If we aren't the biggest what is our advantage?" the answer is they will have to be scrappy like a start up and not just use their size as a bludgeon.

It's being hopeful that Intel can still keep what has defined them for decades (i.e., being an IDM and using that scale to set industry standards). AMD/GF never had that.

 

[siliconbruh999]

Yet somehow AMD is doing well with even less scale, and GlobalFoundries doing ok as well.

This is monopolist thinking, "If we aren't the biggest what is our advantage?" the answer is they will have to be scrappy like a start up and not just use their size as a bludgeon.

I think you don't realise how much their own fabs give them the advantage even with inferior products

 

[MKWVentures]

I normally agree with Dylan but I think the question is whether you think Intel has a better competitive position in Products or Foundry. I think Intel is better competitively in product markets as the leader in PC and DC CPUs. Intel needs to solve a lot of cost issues in manufacturing even thought they have tremendous skill and knowledge. Financing that will not be easy and it is hard to see how this ends well. I don't think the Intel Product guys think Intel Fabs give them an advantage.... I think the Intel Product guys think TSMC gives Apple, Nvidia, and AMD an advantage. Just a thought....

 

[tooLongInEDA]

I think you don't realise how much their own fabs give them the advantage even with inferior products

I certainly don't accept that as a given. Surely only if they have at least cost and technology parity with TSMC. Which must be increasingly difficult to achieve as Intel's chip volume relative to TSMC drops.

Consider the car (auto) industry. When this started, huge car plants like Ford Rouge River had integrated on-site power stations and steelmills. At some point the vertical integration became a cost and was phased out.

Besides which, the real solution to inferior products is to design better products !

An argument was made yesterday that Intel's design team had some critical advantange from working closely with its own fabs that would be lost by working with TSMC. Working with TSMC, it was argued, would leave Intel at a disadvantage to its competitors (alongside the statement to the effect that Intel still has world class design capabilities). I don't buy that argument either. If AMD can do it, why can't Intel ?

The fact that something worked well in the past is no guarantee that it will continue to do so in the future.

 

[BruceA]

The BoD should be fired!!! The whole Paul, BK, Bob, Pat is going to go down as the worst board governance in history. What is lost is Andy Bryant is missing as one of those who failed!

Who has the DNA and chops to want this and do it with the passion that Pat did ? No question his audacity of behavior on many things leaves distaste in many and can’t be corralled it is Pat. Almost every name on the list of possible CEOs likely will destroy the company.

To fund leading edge foundry requires a whale customer0.

One can argue but CCG and DCAI are big whales with volume and need for leading edge silicon. If IFS is within 6 months of reliable ramp of leading edge HPC for then using IFS or TSMC is a wash for them. The big question can IFS deliver and does MJ believe they can?
Put all the next three generations on Intel silicon and bring it all back to US!!!

The failure of Pat was his audacity to believe he could reel in other whales and build Ohio and fantasy of Germany. The BoD failed the smell test there and why he was likely one reason his retirement.

Seems very reasonable strategy to get IFS and LTD to deliver I18A and I14A and I10A with fab build out to support the the full stack of the product team. That is still a 40-60B revenue depending on how much MS they can claw back from AMD. As to AI that ship of direct compete with Nvidia has sailed like x86 two decades ago for the other client and HPC dreamers. Continue to build out the foundry capabilities and become an alternative to not T on secure soil, there should be a few billion there and grow from there. Don’t pretend they are a IDM in transition to open Foundry with I14A the real starting line and spend and staff like that.

Hunker down and focus on what you do best and change the culture to be customer oriented and embrace risk, flexibility and openness to everything!! They still have a lot of dinosaurs and useless roadblockers there.

 

[Jure]

Why everyone says AI ship has sailed? I think there's still a lot of opporunities for Intel in AI space. In future there will likely be a shift towards chain of thought reasoning, which means shift towards inference. This will be an opporunity for AMD, Intel and others to offer cost competitive solutions. Granted, margins will not be Nvidia-like. Intel is able to offer good software support for their products. If Falcon Shores is a decent performing chip, then I see Intel's market share in AI inference growing, especially because OneAPI may offers good programability and c++ / python interface (a must for developers). Same goes for edge AI. As far as I know, Intel's gpus are mostly well supported by oneapi and thus very accessible.

From what I know, OneAPI has no general support for Gaudi and imo, this is one of the biggest issues. With Falcon Shores Intel should solve this. We'll see, but I think general programability and ease of use is one of the biggest advantages of cuda.

 

[OneEng]

I normally agree with Dylan but I think the question is whether you think Intel has a better competitive position in Products or Foundry. I think Intel is better competitively in product markets as the leader in PC and DC CPUs. Intel needs to solve a lot of cost issues in manufacturing even thought they have tremendous skill and knowledge. Financing that will not be easy and it is hard to see how this ends well. I don't think the Intel Product guys think Intel Fabs give them an advantage.... I think the Intel Product guys think TSMC gives Apple, Nvidia, and AMD an advantage. Just a thought....

I tend to agree with this; however, I also agree with @siliconbrush999 that Intel has enjoyed a big advantage having vertical integration with their design and fab:

I think you don't realise how much their own fabs give them the advantage even with inferior products

I believe that the transition to a services company for Intel for their fabs is a WAY more daunting task (and financially difficult as well) than learning to produce their designs on TSMC nodes.

Just like Apple with its closed environment and tight OS to application integration, there certainly has to be similar advantages to Intel's vertical integration approach. The failure of the model, I believe, lies in the inability of Intel to spread the cost of new fab equipment across nearly as many chips and customers as TSMC can. It's kind of a perfect storm for Intel in that the exponentially increasing cost of new nodes now coincides with lower market share/total number of chips being produced.

AMD appears to be straddling the issue of increasing wafer costs by limiting their main stream (high volume) designs to an older node while only producing lower volume, higher margin parts (DC) on the leading edge node.

I don't think that PC consumers are going to suddenly decide they are willing to pay much more for a desktop or laptop (at least not the lions share of them) for incrementally smaller performance increases, so that kinda limits the strategy moves for the chip industry as a whole IMO.

Having said all this, I don't believe it is practical for Intel to continue to lose money at the rate they currently are for the length of time that they now need to peruse a strategy of selling fabrication of others chips for profit (I am guessing another 2 to 3 years before it becomes sustainable if all technical hurdles are overcome).

As a result, it seems to me that the only viable path is to split the company up. The design side can likely compete profitably, but the fab will require several years to become profitable IMO.

 

[count]

It's being hopeful that Intel can still keep what has defined them for decades (i.e., being an IDM and using that scale to set industry standards). AMD/GF never had that.

The problem is that what has worked for decades will not work in the future.

The way I see this is through the framework of Clay Christensens disruption theory. This is a classic case. One of the key takeaways if you read the book is that what makes a technology disruptive is not the technology itself, but the changes in business model enabled by new technology. Companies holding onto the old business model fail.

This is Intel's problem, they have an IDM business model in a fabless world. Ironically it was Daniel Nenni's book, Fabless, that opened my eyes to this, but Daniel himself doesn't quite see it this way. But if you read the "Innovators Dilemma" and "Fabless" back to back, you will see it.