Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/has-intel-stumbled-again.15965/page-2
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Has Intel Stumbled Again?

Wow:

Gelsinger took over as CEO of Intel in Feb. 2021, and received a compensation package worth $178.59 million later that year, according to a financial filing. The compensation included over $1 million in salary, a $1.75 million bonus, over $140 million in stock awards, and nearly $30 million in option awards.

Q1 2022 vs. Q1 2021 Revenue
Intel: -7% (-1% non GAAP, -7% GAAP)
Total value aside - if 3% of his compensation is salary, and 97% is stock based - that's not necessarily a terrible motivator for the CEO to do something for the shareholders..

Turnarounds also take years - so it's going to be interesting to see if the shareholders (who somewhat rewarded the previous CEO's behavior because of buybacks and dividends) continue to side more with short term profits over the long term viability of the historical company model which Gelsinger is trying to restore. Strategically, Intel having it's own fabs can give them advantages others can never have, but the risk profile for this advantage only increases with time.

(If you want real "CEO Wows" - look at former Lockheed CEO Kubasik, who was fired for improperly promoting subordinates he was sleeping with, and paid many $10s of millions to leave.. and then hired by another defense contractor as their CEO - L3 - that has it's own wiki page internally making excuses for his behavior previously.. lol!)
 
Total value aside - if 3% of his compensation is salary, and 97% is stock based - that's not necessarily a terrible motivator for the CEO to do something for the shareholders..

Turnarounds also take years - so it's going to be interesting to see if the shareholders (who somewhat rewarded the previous CEO's behavior because of buybacks and dividends) continue to side more with short term profits over the long term viability of the historical company model which Gelsinger is trying to restore. Strategically, Intel having it's own fabs can give them advantages others can never have, but the risk profile for this advantage only increases with time.

(If you want real "CEO Wows" - look at former Lockheed CEO Kubasik, who was fired for improperly promoting subordinates he was sleeping with, and paid many $10s of millions to leave.. and then hired by another defense contractor as their CEO - L3 - that has it's own wiki page internally making excuses for his behavior previously.. lol!)
"Strategically, Intel having it's own fabs can give them advantages others can never have, but the risk profile for this advantage only increases with time."

From time to time I'm thinking what exactly the advantages Intel has by owning fabs? Especially in the fast changing market, does that really matter?
 
I missed seeing this earlier...but apparently some Intel employees are working long hours and are rather unhappy. They want to unionize!
 
"Strategically, Intel having it's own fabs can give them advantages others can never have, but the risk profile for this advantage only increases with time."

From time to time I'm thinking what exactly the advantages Intel has by owning fabs? Especially in the fast changing market, does that really matter?
Fair - my thinking is :

Potential for higher margins than fabless. Risk : utilization of fabs

Potentially less barriers to raising capacity for future products. (don't need to compete against other fabless). Risk : capital, tech level.

Pivoting faster in some ways (you control the fabs) but slower in others (your own commitments limit you to what you can do quickly). The mitigation here is to both outsource and insource to try to get both benefits.
 
Fair - my thinking is :

Potential for higher margins than fabless. Risk : utilization of fabs

Potentially less barriers to raising capacity for future products. (don't need to compete against other fabless). Risk : capital, tech level.

Pivoting faster in some ways (you control the fabs) but slower in others (your own commitments limit you to what you can do quickly). The mitigation here is to both outsource and insource to try to get both benefits.
"Potentially less barriers to raising capacity for future products. (don't need to compete against other fabless). Risk : capital, tech level."

As an IDM, Intel will have to ramp up the production when time is good and ramp down the production when time is bad. When they ramp down the output, all those financial modeling and assumptions will be thrown out of the window. It's not an easy task.
 
"Potentially less barriers to raising capacity for future products. (don't need to compete against other fabless). Risk : capital, tech level."

As an IDM, Intel will have to ramp up the production when time is good and ramp down the production when time is bad. When they ramp down the output, all those financial modeling and assumptions will be thrown out of the window. It's not an easy task.
Agreed - definitely not easy. During 'bad' times they could still at least (if financially justified) keep production flat to try to take share but I definitely get your point.

I think Intel gives up a key "opportunity" if they go fabless, and there's still a chance their fab tech could pull ahead again in which case it's pretty justified.. We've been shrinking since the 1960s, so it's a long haul..
 
Why has this thread dried up? This seems like a big deal.

SemiWiki is the quintessential assembly of all that is semiconductor. If it’s not here, it’s suspect.

I keep hearing regurgitated rumblings from hack sources about Intel not being ready to supply the Meteor Lake CPU, using their Intel 4 node, and since TSMC is supplying most other chiplets, they’re an obvious choice.

So, will Intel 4 be ready or not? Will Intel delay Meteor Lake until Intel 4 is ready? How late could Intel 4 be?
 
Why has this thread dried up? This seems like a big deal.

SemiWiki is the quintessential assembly of all that is semiconductor. If it’s not here, it’s suspect.

I keep hearing regurgitated rumblings from hack sources about Intel not being ready to supply the Meteor Lake CPU, using their Intel 4 node, and since TSMC is supplying most other chiplets, they’re an obvious choice.

So, will Intel 4 be ready or not? Will Intel delay Meteor Lake until Intel 4 is ready? How late could Intel 4 be?
Intel has been developing 4, 3, 20A, and 18A in parallel. It is likely hoping to push the most advanced process it can complete as soon as possible. So even 18A is no longer gated by High-NA availability. So they are even going back to multipatterning. That said, if increasing EUV usage is what they had hoped for, then they will get stuck at 4 presumably.
 
Intel has been developing 4, 3, 20A, and 18A in parallel. It is likely hoping to push the most advanced process it can complete as soon as possible. So even 18A is no longer gated by High-NA availability. So they are even going back to multipatterning. That said, if increasing EUV usage is what they had hoped for, then they will get stuck at 4 presumably.

Thanks for those data, Fred.

Let me rephrase the questions; Given Intel is not in risk production of Meteor Lake for many reasons, the primary reason being the lack of 13 EUV tools at Fab 42 in Chandler Arizona (where Intel 4 will start?).

1.) When will Fab 42 have sufficient EUV tools, and other, to start Risk Production of Intel 4?

2.) How long will Intel 4 be in risk production, given EUV inexperience and other, before Volume Production starts?
 
Fab 34 getting the EUV tools from Oregon but not even online until 2023.

Thanks Fred, I understood that Intel 4 was to start at Fab 42, I guess those were just prototypes (test shuttles).

Since Intel 4 will be produced at Fab 34 in Ireland and given Fab 34 received their first UEV tool from Oregon in April ’22, then keeping in mind that Intel is only receiving 1 EUV tool per month, Fab 34 will not have received all 13 EUV tools required for Meteor Lake until May ’23.

Once the last tool is installed, an additional 3-4 months for calibration is required, which pushes the beginning of Risk Production to 4Q23 ???

This schedule seems to indicate that Meteor Lake will not start Volume Production in ’23.
 
To answer hist78 question, what is the advantage of Intel owning fabs?
- Cost and quality: You can potentially have a lower cost (take out the middle man) and better quality (higher yields)
- Improved supply: You can have supply when others have no supply. Timeliness.
- Leverage to inflation: Inflation drives up costs while your fab investment, made previously, remains the same, so you make more money

The last one is clearly an incentive to build like mad when inflation is high as it is today.

Downsides of owning fabs:
- Capital intensity increases risk of insolvency
- Business cycles; booms and busts
- Ongoing research and development risk; failure to develop a functioning next node
 
Thanks Fred, I understood that Intel 4 was to start at Fab 42, I guess those were just prototypes (test shuttles).

Since Intel 4 will be produced at Fab 34 in Ireland and given Fab 34 received their first UEV tool from Oregon in April ’22, then keeping in mind that Intel is only receiving 1 EUV tool per month, Fab 34 will not have received all 13 EUV tools required for Meteor Lake until May ’23.

Once the last tool is installed, an additional 3-4 months for calibration is required, which pushes the beginning of Risk Production to 4Q23 ???

This schedule seems to indicate that Meteor Lake will not start Volume Production in ’23.
It seems hard to swallow, doesn’t it, this “Intel 4 in 2022” claim. Yet the signs are there that It is happening.

There are starting to be Intel Meteor Lake teardown-style analyses discussing the many tiles (and as many process nodes as tiles, seemingly) as a done deal. So it seems to be a finished product.

Intel 4 may not be that dense in SRAM. Intel seems to be adjusting process technology content dynamically to hit dates.

Tiles appear to be a form of fab lite. So as Intel shifts the majority of tiles to TSMC, they will be ramping down production, to some degree. So the expectation of HVM1 in 2022 may not apply anymore. Maybe Intel can essentially supply all their Intel 4 tiles from Mods 1, 2 and 3 of D1X.
 
Teardown of Meteor Lake by SemiAnalysis and Locuza: https://semianalysis.substack.com/p/meteor-lake-die-shot-and-architecture?s=r

Less than 25% of Meteor Lake is Intel 4. It may still be mostly Intel older nodes (SoC tile).
Really interesting, and I wonder: will the ability to use lagging nodes on the tiles that don't need it (SoC) allow that lower cost to offset the new packaging complexity/cost? And not to mention the cost of the TSMC N3-based GPU tile. Similar to how AMD has been able to use GF for their SoC / IO chiplet allows them to min/max the cost/availability of those expensive leading edge TSMC chiplets (and will continue to use N6 for GPU / IO chiplet and N5 for compute).

And going back to the discussions on "will Intel have the EUV fab capacity to ramp Meteor lake": this relatively small amount of Intel 4 usage seems to be a good sign for launch, in that Intel really does not need nearly as much Intel 4 fab capacity to launch Meteor Lake (disaggregated, Intel 4, N3, Intel 16) as they needed Intel 7 fab capacity to launch Alder Lake (monolithic). But related, will the packaging fabs be ready for the Meteor Lake ramp as well? They may need a small amount of EUV / Intel 4, but they certainly need *a lot* of EMIB / Foveros packaging...
 
It seems hard to swallow, doesn’t it, this “Intel 4 in 2022” claim.

Do you have any reputable reference to an “Intel 4 in 2022” claim?

Pat has only promised Intel 4 in 1QA23, that I know of.

If Pat releases Intel 4 before 2Q23, they will likely be from low yield (<20%) Risk Production, similar to Samsung 4nm (<20% yield).
 
Do you have any reputable reference to an “Intel 4 in 2022” claim?

Pat has only promised Intel 4 in 1QA23, that I know of.

If Pat releases Intel 4 before 2Q23, they will likely be from low yield (<20%) Risk Production, similar to Samsung 4nm (<20% yield).

Unfortunately Intel's culture has changed. Now the bragging rights are often more important than the real technical achievement.

I have posted a thread about an Intel Vietnam manufacturing change and consequently it increased Intel revenue by $2 billion. Intel PR department uses this as an example to tell people the greatness of Intel IDM business model.

What Intel didn't say is that their 1% 2021 revenue growth is a very rare and miserable performance in the whole semiconductor industry.

Thread 'Intel believes IDM model is better' https://semiwiki.com/forum/index.php?threads/intel-believes-idm-model-is-better.16119/
 
Intel did not stumble "again", it's only Pat Liesinger made you all think he is magician and this time it will be better.
 
Intel performance was based on their process lead. I only use three things on my labtop zoom or skype, word processsing, and the browser. The ryzen 7 is better for what I need. Intel goes it alone while amd uses the ecosystem. I'm not paying $1000 for a laptop with a celeron processor, 10th or 11th generation processor and another $60 for microsoft. I5 and i7 is good but that's as much as a macbook. If you're paying for intel's peak performance it's a short time frame until something better comes out.. Intel is living off their legacy and customers that don't know better.
 
Intel performance was based on their process lead. I only use three things on my labtop zoom or skype, word processsing, and the browser. The ryzen 7 is better for what I need. Intel goes it alone while amd uses the ecosystem. I'm not paying $1000 for a laptop with a celeron processor, 10th or 11th generation processor and another $60 for microsoft. I5 and i7 is good but that's as much as a macbook. If you're paying for intel's peak performance it's a short time frame until something better comes out.. Intel is living off their legacy and customers that don't know better.

But the reality now is a few generations old Celeron at $1000+ as a mainstream sale you see in a mall.

All big Taiwanese ODMs keep closing their plants in China as Taiwanese labour cost now has an advantage vs. South China
 
Back
Top