Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/what-is-really-going-on-with-intel%E2%80%99s-18a-process.20944/page-6
        )

    [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] => 2021770
            [XFI] => 1050270
        )

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

What is really going on with Intel’s 18a process?

Intel 3 launched on time it is the only node that is not delayed along with Sierra forest today they are launching Granite Rappids on Intel 3 like their schedule

Companies want to see 18A which will decide fate of 14A
Delivering on technology is not the end goal. The end goal is being financially successful. Lots of previous, now no longer in existence companies have had technology lead but been inefficient and not financially viable.

The Stock crashed due to financial issues primarily caused by excessive spending and hiring from 2021-2023 and from a focus on foundry manufacturing which is Intel's least successful business (People got to see that in spring).

Lets see how Manufacturing P&L does and when 18A really ramps (20% of business) and then we will know the answer.

Has Intel given an update on what is going on with the Fab 52/62 situation?
 
Last edited:
Delivering on technology is not the end goal. The end goal is being financially successful. Lots of previous, now no longer in existence companies have had technology lead but been inefficient and no financially viable.

The Stock crashed due to financial issues primarily caused by excessive spending and hiring from 2021-2023 and from a focus on foundry manufacturing which is Intel's least successful business (People got to see that in spring).

Lets see how Manufacturing P&L does and when 18A really ramps (20% of business) and then we will know the answer.

Has Intel given an update on what is going on with the Fab 52/62 situation?


I started thinking about the tragic story of:

AT&T → Lucent - Bell Laboratories Innovations → Alcatel-Lucent → Nokia.

As of this moment, 11:38 EDT, 9/24/2024, Nokia’s stock price is $4.36 per share.
 
Delivering on technology is not the end goal. The end goal is being financially successful. Lots of previous, now no longer in existence companies have had technology lead but been inefficient and no financially viable.
This is why Gelsinger's approach gives me pause. He sees delivering on technology as the only means to be financially successful, and like you said, it's through Intel's least successful (and lowest margin?) business.

Why not set up foundry services on legacy nodes so you're not having to learn a new business and deliver on tech simultaneously (i.e., what they were looking to get with Tower)? Why not pitch design services only (without manufacturing bundled in) to build customer relationships and ecosystem IP? These are half-steps, sure, but they're steps to building a customer-focused operating model and still leave room for cross-sell/upsell.

Gelsinger is a CTO CEO, which is fantastic for Intel, but if the only tool you have is a hammer...
 
This is why Gelsinger's approach gives me pause. He sees delivering on technology as the only means to be financially successful, and like you said, it's through Intel's least successful (and lowest margin?) business.

Why not set up foundry services on legacy nodes so you're not having to learn a new business and deliver on tech simultaneously (i.e., what they were looking to get with Tower)? Why not pitch design services only (without manufacturing bundled in) to build customer relationships and ecosystem IP? These are half-steps, sure, but they're steps to building a customer-focused operating model and still leave room for cross-sell/upsell.

Gelsinger is a CTO CEO, which is fantastic for Intel, but if the only tool you have is a hammer...
Somewhere in all this, we need to remember that Intel is partly trapped in its heritage of being an outright tech leader and dominating markets. I cannot imagine that they really have the belief and desire to set up a relatively low margin trailing edge foundry business in which they are a #2 player at the very best and then only after years of struggle. It might make some sense (if they weren't Intel), but if their hearts aren't in it I'm dubious.

Also, the Intel legacy nodes - until the past few years - are on internal tools which are hard for customers to use. I think they're partnering to get more usable legacy nodes.
 
This is why Gelsinger's approach gives me pause. He sees delivering on technology as the only means to be financially successful, and like you said, it's through Intel's least successful (and lowest margin?) business.

Why not set up foundry services on legacy nodes so you're not having to learn a new business and deliver on tech simultaneously (i.e., what they were looking to get with Tower)? Why not pitch design services only (without manufacturing bundled in) to build customer relationships and ecosystem IP? These are half-steps, sure, but they're steps to building a customer-focused operating model and still leave room for cross-sell/upsell.

Gelsinger is a CTO CEO, which is fantastic for Intel, but if the only tool you have is a hammer...
100% agree. but thats just me

But Pat is trying to fight a different battle than I would fight .... "Make Intel THE dominant company again". His cheerleading is great. getting the numbers to line up and seeing what the spreadsheets say is concerning. Its a very different world than 2000
 
This is why Gelsinger's approach gives me pause. He sees delivering on technology as the only means to be financially successful, and like you said, it's through Intel's least successful (and lowest margin?) business.

Why not set up foundry services on legacy nodes so you're not having to learn a new business and deliver on tech simultaneously (i.e., what they were looking to get with Tower)? Why not pitch design services only (without manufacturing bundled in) to build customer relationships and ecosystem IP? These are half-steps, sure, but they're steps to building a customer-focused operating model and still leave room for cross-sell/upsell.

Gelsinger is a CTO CEO, which is fantastic for Intel, but if the only tool you have is a hammer...
My knowledge is quite limited, correct me if I am wrong. The reasons Intel's fab seems to have low margin (for now):

1. Intel uses similar TSMC process's pricing for its own Fab's pricing. For example, Intel used TSMC 7nm's current pricing for IFS Intel 7, and so on so forth. This puts IFS in a very bad look, as its Intel 7 was not a great node in the first place (cost is high), and then the pricing is dirty cheap, because TSMC has fully depreciated 7nm and can afford to offer it with a price.

2. There is human labor cost difference b/w Taiwan and U.S., but I think this is over-exaggerated. Taiwan is not exactly a poor place, and AZ does not pay silicon valley style salary. What is more, once a process is set up and running, human labor cost is less than 40% of Fab cost.

3. Intel is rushing to build shells, purchase equipment, but not generating much value with these build-outs yet, further depressing its margin.

4. In an imaginary world, if IFS did not exist, TSMC would have charged quite a bit more than current pricing, because they would claim that Intel is a very large customer, the supply is tight, blah blah.

As IFS transits to Intel 3 and Intel 18A, the above issues 1 and 3 will get resolved by their own. What is more, TSMC margin is a bit over 50%, I bet IFS would be content with 25%-30%.
 
+1 to what siliconbruh999 said.

A Clearwater Forest chip consists of 14 small chiplets, so it should yield quite well (probably even better than PTL) on the wafer level. However, the challenge is to do 3D chip stacking of so many chiplets.

A few weeks ago, IFS head said PTL yielded well, but did not use the same language for CWF. My speculation is that their 3d stacking technology was the bottleneck.
Recent news, this is supposed to be a Clearwater Forrest chip. Looks like they cut down the chiplet count.

 
My knowledge is quite limited, correct me if I am wrong. The reasons Intel's fab seems to have low margin (for now):

1. Intel uses similar TSMC process's pricing for its own Fab's pricing. For example, Intel used TSMC 7nm's current pricing for IFS Intel 7, and so on so forth. This puts IFS in a very bad look, as its Intel 7 was not a great node in the first place (cost is high), and then the pricing is dirty cheap, because TSMC has fully depreciated 7nm and can afford to offer it with a price.

2. There is human labor cost difference b/w Taiwan and U.S., but I think this is over-exaggerated. Taiwan is not exactly a poor place, and AZ does not pay silicon valley style salary. What is more, once a process is set up and running, human labor cost is less than 40% of Fab cost.

3. Intel is rushing to build shells, purchase equipment, but not generating much value with these build-outs yet, further depressing its margin.

4. In an imaginary world, if IFS did not exist, TSMC would have charged quite a bit more than current pricing, because they would claim that Intel is a very large customer, the supply is tight, blah blah.

As IFS transits to Intel 3 and Intel 18A, the above issues 1 and 3 will get resolved by their own. What is more, TSMC margin is a bit over 50%, I bet IFS would be content with 25%-30%.
For #1: Agree, Intel's manufacturing was never about efficiency though. Would be surprising if they're able to achieve both leading edge tech and operational efficiency. If Intel focused on legacy nodes for foundry, then depreciation is no longer an issue.

For #3: CapEx shouldn't impact margins though (although equipment depreciation will). Margins are down because revenue isn't offsetting operating costs.

For #4: I think this depends on how Intel approached it. If Intel signals early with a large enough order, TSMC could build sufficient capacity but Intel could also negotiate by volume and by threatening to go with Samsung. Does Intel take the Apple approach and pay a premium for first access, or the Nvidia approach and hold out for the best pricing?
 
My knowledge is quite limited, correct me if I am wrong. The reasons Intel's fab seems to have low margin (for now):

1. Intel uses similar TSMC process's pricing for its own Fab's pricing. For example, Intel used TSMC 7nm's current pricing for IFS Intel 7, and so on so forth. This puts IFS in a very bad look, as its Intel 7 was not a great node in the first place (cost is high), and then the pricing is dirty cheap, because TSMC has fully depreciated 7nm and can afford to offer it with a price.

2. There is human labor cost difference b/w Taiwan and U.S., but I think this is over-exaggerated. Taiwan is not exactly a poor place, and AZ does not pay silicon valley style salary. What is more, once a process is set up and running, human labor cost is less than 40% of Fab cost.

3. Intel is rushing to build shells, purchase equipment, but not generating much value with these build-outs yet, further depressing its margin.

4. In an imaginary world, if IFS did not exist, TSMC would have charged quite a bit more than current pricing, because they would claim that Intel is a very large customer, the supply is tight, blah blah.

As IFS transits to Intel 3 and Intel 18A, the above issues 1 and 3 will get resolved by their own. What is more, TSMC margin is a bit over 50%, I bet IFS would be content with 25%-30%.

Can you please explain a little bit about why #1 and #3 will get resolved by their own? Thanks.
 
For #1: Agree, Intel's manufacturing was never about efficiency though. Would be surprising if they're able to achieve both leading edge tech and operational efficiency. If Intel focused on legacy nodes for foundry, then depreciation is no longer an issue.
There are enough legacy node capacity in the world, perhaps a little bit too much already, and China is joining the competition there. IFS should not, and does not plan to (according to its roadmap) focus on that segment.

On the other hand, there are only two (or three) leading edge foundries (TSMC leading, IFS currently in distant second, and Samsung seemingly not able to ramp up its 3nm), the pricing power is strong, and IFS does not need to have TSMC's operational efficiency to enjoy a healthy margin there.

For #3: CapEx shouldn't impact margins though (although equipment depreciation will). Margins are down because revenue isn't offsetting operating costs.
Let's suppose Intel spend $3B to build shells in one year, but does not generate revenue from that spending. The shell will start to depreciate, with $0 revenue to offset, dragging down margin.

For #4: I think this depends on how Intel approached it. If Intel signals early with a large enough order, TSMC could build sufficient capacity but Intel could also negotiate by volume and by threatening to go with Samsung. Does Intel take the Apple approach and pay a premium for first access, or the Nvidia approach and hold out for the best pricing?
A monopoly is never good. Not good for Intel, not good for other foundry customers. I would argue that it is not good for TSMC, either, because a healthy amount of competitive pressure strengthens TSMC.

Can you please explain a little bit about why #1 and #3 will get resolved by their own? Thanks.
For issue 1, IFS has a bunch of under utilized (utilization is somewhere around 60%?) legacy node capacity (say, Intel 7 and even older nodes). Even when a portion of that capacity are utilized, the pricing is quite low, because TSMC sets the bar. Low utilization + subpar pricing kills margin.

Now, in a few years, IFS aims to have mostly Intel 3 and Intel 18A and 14A, the pricing will be 3x (if I remember correctly) of Intel 7, the utilization will be higher, because Intel products don't need to go to TSMC as much as today, and hopefully, the leading edge nodes could attract some outside customers, too.

For issue 3, please see my reply above. When the build out pace becomes normalized, the margin pressure will lessen.
 
Last edited:
Back
Top