Thread regarding Intel Corp. layoffs

@f1 Reuters keeps spewing inaccurate info on Intel, which often sounds like it was written by a hedge fund that barely understands the business.

Intel failed its own law at 10nm for several years. Why do you think it can be successful after that with even more complexities?

@fa The stock will respond to earnings growth, which can be helped by lowering cost but ultimately is about revenue growth and Intel may be able to take some share back from AMD here and there but ARM is taking from the TAM.

The wildcard is external customer volume and Intel could fill the newer fabs with just the volumes from (a few) TSMC customers wanting a second source.

Selling the older fabs helps the balance sheet so lowers the cost of debt and reduces the risk of carrying so much debt. The company used debt and buybacks to mask the worsening financials and any gain from that is long gone. Thanks, Bob Swan.

Others have said it. Intel needs to get about 50% smaller in order to grow again. The current effort is just the start of that process, because HC being over 100k was an attempt to grow again by getting bigger. Thanks Pat Gelsinger.

@fn thanks! I prefer that.

@fa LBT is making the company be sustainably profitable.

To be viable in the longer term, the company has to restart the capital cycle, which was broken with 10nm and the subsequent loss of the WSPW needed to get enough cash flow to pay for the next node.

Once AMD and ARM developers like Apple were able to produce on a newer node than Intel, the market dominance was broken (and can't be regained with x86). Whatever advantage Intel had with the highly customized design rules and nonstandard fab tools and recipes became a disadvantage.

This was 10 years ago, and everything since then has been marketing BS by the company. X86 will be around for some years but is virtually guaranteed to lose market share in CPU and no one will use that architecture for GPU because it is incapable of outperforming the simpler ARM architecture. ALL existing AI efforts are based on an x86 core, so are incapable of beating ARM.

With the x86 dominance a thing of the past, bringing in external customers is required to generate enough wafer volumes to pay for the next node. Either that or shed IFS and be fabless, but if still refusing to move on from x86, then that is a guarantee for long term decline.

This is why the Board is still trying to keep IFS, but it has to get through this transition to the industry standard PDK needed to get meaningful external customers, and that is 18AP at least if not 14A. This was always the situation regardless of how it was spun.

I apologize if you were instead looking for a mindless, troll answer.

@f3 you really think Intel will turn itself around. What will be the metrics for success?

You think our stocks will go back to the 50's in a few years?

Then 12A or 10B or 8C ….. you get the idea. This is plain and simple disgusting! A great towering American company that soon will be out of Nasdaq 100 and eventually S&P 500.

@f2 Nonsense of course.

Yes 18A was for Intel. So if Intel can't produce 18A for themselves, how the he'll are we going to make anything for external customers. We are a laughing stock.

Does anyone really think this layoff will magical inspire our eng or our process to produce 18A or 14A.

Lipbu will need to slowly hire outside talent and slowly let our failed talents go. Fill the top of the bucket with fresh clean water while poking a small hole at the bottom to let the cr-p out.

Note that the article/rumor came from Reuters, which appears to have some writers with a stock manipulation agenda, putting out a regular stream of unfounded rumors.

Not to say it isn't true and 18A was mostly for internal use and as a proof point to external customers that Intel was in fact transitioning to industry standard design rules.

This transition is hard stuff, undoing a codependency which previously existed between ProdCo and IFS. Prodco has since moved on, learning to work with standard PDK at TSMC, and now the production side is making that shift.

As the other poster mentioned, this involves both recipes and tool mods, so is kind of a big deal and by all appearances they are working through the changes reasonably well.

@a6 From anther site: 'Xe3 iGPU tile on some PTL SKU is on Intel 3'

18A has no external customers that can offset the development costs.

That’s a fact.

One can debate the reasons, but this is a fact one can easily look up in the many 10Q financial reports.

Or you can listen to the CFO

"We get test chips, and then some customers fall out of the test chips... So committed volume is not significant right now, for sure,"

Intel employees live in a reality distortion field where everything that Intel does is awesome.

Another inconvenient fact is that Pat sold 18a as the be all end all for IFS. Look, he even said. IN PUBLIC, that he bet the company on 18a and that has clearly failed.

The goalposts have moved and Intel has ZERO credibility with 14a.

@bc Consider that most of the posters on this site are either young fab technicians idling in the cafe and not really knowledgeable about the company, or outright trolls who may not know what Intel even does.

That said, the comments about the transition to industry standard PDK are correct. You can check out other, more reputable sites to research how important the issue is.

Standard PDK means it is based on the design rules that not only TSMC uses but Samsung and all the others. This is not a triviality.

I think LBT is in all manner of ways, rapidly transitioning the company so that it even has an option to sell off IFS and/or ProdCo. IFS is worthless to any other foundry until it is at least on 18A-P, because only then will it have tools and recipes configured to build an industry standard set of design rules. 18A is still wildly customized for CPU, in ways that by now are obviously pointless (as proven by TSMC & AMD).

18A-P will be the first time Intel has ever met industry standard design rules and that is no small thing. It may take until 14A for the transformation to be complete, and by then the external customers wanting a second source will come around. Right now they are just checking it all out.

Reading the comments here from Intel employees, it is clear that Intel is clueless about fondey business. Intel using non-TSMC KPD is not a show stopper, as long as the process performs reasonably well.

The simple fact is all foundry customers desperately want a second source to TSMC. So, if Intel 18A performs within reasons, customers will want to source therefrom, even if it costs 10s of millions to develop products on a second source fab. One negative example is Samsung foundry. Its process has terrible current performance, which is an absolutely no for mobile products. So Samsung can't even give the process away.

Intel have no customers really points to Intel 18A being having serious performance deficiencies, not just yield issue. Unless the yield is so bad that Intel can't afford to subsidise the customers.

@a4 Intel doesn't have to be ahead of TSMC, it merely needs to stay close.

18A is close enough but OP is right, it is build on non-standard design rules so not very useful for external customers.

There have been plenty of past examples at Intel and other semi fabs where processes exist basically for learnings and to test new technologies. 18A is definitely that, and appears ready for internal use as well, so not useless.

There have been plenty of test chips run for external customers, and that should lead to the start of customer volumes with 14A.

There is likely excess EUV capacity at the moment but that will be absorbed pretty quickly, as 14a development ramps.

Can't be overstated how important it is for Intel to demonstrate manufacturing which is based on a PDK which supports industry standard design rules. That makes the process useful to external customers and also makes IFS able to be spun off or sold outright.

By all appearances, everything LBT is doing is for that purpose.

@ak moore’s law isn’t even based on facts. It was a company vision.

Today going smaller is defying the laws of physics. I once interviewed to a different BU and got laughed at when I complained about decimal points in charts that were basically talking about Angstroms. The measure between atomic bonds. As a chemist it made zero sense why we should have that much control.

I told you earlier. This is a scam. Next year they will cancel 14A and jump to 10A and so on since Moore is dead.

It is sad but if we continue to waste time on 18A and then jump on 14A, our competition will be that much more ahead of us. If jump on 14A now, the gap may be less and faster for us to catch up.

OK who the heIl am I kidding. Intel will not catch up, 18A has delays and the people that might catch it up are probably being laid off next week.

If we cancel 18a, it’s over. The industry will not wait for or commit to more Intel pipe dreams and endless blatant lies.

@aa The site is infested with trolls, but that is a choice.

Lots of young fab technicians idling in the cafe, desperately trying to be funny and troll and failing most of the time.

Soon many of them will have even more free time, and rightfully so.

Wow, an actual informative thread with real employee insights instead of trolls? What site am I on?

@a7 I think you've hit the key point.

If LBT wants to sell IFS, he needs it to be on at least 18A-P if not 14A.

Preferably the board will also sell off all the older fabs, to make IFS more sellable.

The foundry that buys the older fabs would run the intel volume, then transition those older nodes to be able to manufacture to industry standard PDKs, a major retooling.

Keep in mind that many of the tools are modified by Intel just because of its use of non-standard design rules. Everything is (I hesitate to use this word) 'optimized' to Intel design rules.

In reality, every fab not producing 14A is really the culmination of 50 years of pointless, needless complexity built up by Product + TD. Whole careers have been built on being non-standard.

That stopped making sense about 15 years ago, when TSMC showed that it could outperform Intel.

Intel is embarrassing. Now I know why Lip-Bu wanted the Intel ticker removed from circuit.

@a4 18a was never expected to get Intel ahead of TSMC. It was about trying to catch up.

The fact that it was not built to a standard PDK shows the rushed nature of how it was designed.

So it was more about putting something out there that any external customer would even consider using.

No external customer is going to waste time with a PDK which is not based on industry standards, which is why even 7nm was not considered for external use.

I predict that Intel will find good internal use for 18a just as it is now heavily using 7nm.

This is more about the company facing the reality of 18A for external customer use. A number of companies have looked at it and I'm sure the feedback was that they found it to still be too customized for Intel.

By the time 18A-P is out there, it really is just a bridge to 14A, which probably explains why it may also be kept for internal use. By the time the company is ramping 14a, there should be a lot of EUV learnings as well as learnings which will come from the transition to manufacturing to industry design standards.

Manufacturing to industry standards may prove to be the toughest issue. I can imagine all manner of yield & performance issues associated with the shift away from the old design rules.

Once Intel has factories that manufacture to industry standard design rules, that may also make IFS more suitable for acquisition by another foundry.

Until 14a, IFS is useless to other foundries.

Isn't Xe3p on 18a?

18a-p is not competitive with TSMC.
14a may have some advantages so focus on that. The 18a cycle is done and it’s too late.

Now, you have to believe 14a will be competitive with TSMC A16/A14 in 2027. Good luck with that.

everyday only negative news. when will intel see light? lol

@a1 One of many falsehoods. Yield is on track and the cycle time is actually better than earlier processes at this stage.

The issue is that it was not designed for a standard PDK, and this is not in dispute.

18A-P was designed for a standard PDK, and if they decide not to market it then it may show that external customers want more proof that Intel can manufacture to a standard PDK (because it has never done that).

I agree with OP that what appears to be going on is yet another example of LBT trying to rip out the disfunction and delusion which has infected upper management.

That's why a line of EVPs have been leaving the company, and they did not quit.

18a cant even yield…