Intel Unveils Panther Lake Architecture: First AI PC Platform Built on 18A

[Fred Chen]

View attachment 3734
someone made a edited version it's pretty nice yield

on a side not why semi companies don't label their graphs like do they have personal vendetta against graph axis?

Going by 0.2 at Q1'25, I would have added 0.02 to the estimates afterwards:
Q2'25: 0.17
Q3'25: 0.15
Q4'25-Q1'26: 0.13

But 0.15 still gets ~87% for 100 mm2 die

 

[KevinK]

* The small Xe chiplet is Intel Foundry’s learning path on GPU IP manufacturing?

Betting most heavily on this one.

 

[hist78]

Well, maybe 18A yield doesn't need press interpretation, Intel presented D0 trend chart here:

View attachment 3733

Intel's 18A production starts before TSMC’s competing N2 tech — here's how the two process nodes compare

Production starts at Fab 52.

www.tomshardware.com

In Q3 2024, D0 was said to be 0.4: https://www.tomshardware.com/tech-i...8a-is-healthy-potential-clients-are-lining-up

It’s interesting that the chart title indicates 18A production has begun at two U.S. fabs.

Is demand truly strong enough for Intel to start production at two fabs? Is there any particular reason behind this, positive or negative?

 

[siliconbruh999]

Is there such strong demand that Intel has begun 18A production at two US fabs?

Currently it is only being produced in Fab52 and yes they are planning to shift more production inHouse and Intel is a very big wafer consumer in Itself and 18A is supposed to support like 2-3 generation of products maybe more

 

[hist78]

Currently it is only being produced in Fab52 and yes they are planning to shift more production inHouse and Intel is a very big wafer consumer in Itself and 18A is supposed to support like 2-3 generation of products maybe more

But the title of the chart says Two Fabs started 18A production.

 

[osnium]

But the title of the chart says Two Fabs started 18A production.

Does the Oregon fab count as one of the two fabs?

 

[hist78]

Does the Oregon fab count as one of the two fabs?

I was thinking that too. But I thought Intel’s Oregon fabs were primarily for R&D. If Intel’s Oregon R&D fab needs to be used as a production site, that could indicate several possibilities:

1. 1. Demand is extremely strong, forcing Intel to shift the Oregon fab into a production role.
2. 
   
3. 2. Intel lacks sufficient EUV machines and/or other tools at the Arizona Fab 52, meaning Fab 52 alone doesn’t have enough capacity to meet demand, whether large or small.
4. 
   
5. 3. Intel has adopted a new capital allocation strategy and is trying to maximize the return on investment from its Oregon R&D fab.
6. 
   
7. 4. The yield of Intel 18A is not yet adequate, forcing Intel to use the R&D fab to make up for the shortfall.

 

[meo9725]

Meh...

 

[meo9725]

I was thinking that too. But I thought Intel’s Oregon fabs were primarily for R&D. If Intel’s Oregon R&D fab needs to be used as a production site, that could indicate several possibilities:

1. 1. Demand is extremely strong, forcing Intel to shift the Oregon fab into a production role.
2. 
   
3. 2. Intel lacks sufficient EUV machines and/or other tools at the Arizona Fab 52, meaning Fab 52 alone doesn’t have enough capacity to meet demand, whether large or small.
4. 
   
5. 3. Intel has adopted a new capital allocation strategy and is trying to maximize the return on investment from its Oregon R&D fab.
6. 
   
7. 4. The yield of Intel 18A is not yet adequate, forcing Intel to use the R&D fab to make up for the shortfall.

Most likely, a combination of 2 and 4.

 

[Artificer60]

Intel has always started production in Oregon early in a ramp. Most of the dice for the initial launch come out of Oregon. That isn't unique to 18A, it has always been that way (at least since D1B was opened in 1996). Oregon is typically running 3 technologies. The current node until at least the first factory is fully ramped, the next gen process that is getting all the "next great thing" press (in this case 14A), and the node beyond that which is starting to move out of pathfinding and into development.

 

[hist78]

Intel has always started production in Oregon early in a ramp. Most of the dice for the initial launch come out of Oregon. That isn't unique to 18A, it has always been that way (at least since D1B was opened in 1996). Oregon is typically running 3 technologies. The current node until at least the first factory is fully ramped, the next gen process that is getting all the "next great thing" press (in this case 14A), and the node beyond that which is starting to move out of pathfinding and into development.

If this has been a long-standing Intel practice, mixing R&D and production in the same environment, does it still make sense in today’s situation?

Competition is fierce, and every year Intel’s competitors are rolling out new products faster and more frequently. Can Intel’s R&D fabs truly concentrate on R&D under these circumstances? Unless Intel still believes its competitors aren’t fast enough to "catch up"?

Or perhaps some of Intel’s Oregon fabs do not participate in R&D and instead handle small quantity production runs. Could this be one of the reasons pushing Intel’s costs higher than other companies?

 

[Xebec]

* The small Xe chiplet is Intel Foundry’s learning path on GPU IP manufacturing?

Such an interesting thing to bring only that small GPU chiplet in house…

FWIW, The 4 Xe core verison of the iGPU is going to be the higher volume product by far.

Intel's showing 3 main configurations:

8 CPU + 4 GPU cores
16 CPU + 4 GPU
16 CPU + 12 GPU

The 12 GPU is definitely the more premium / limited quantity version (relatively speaking) where the 8+4 and 16+4 are going to be in huge volumes for corporate laptops, in addition to retail consumers.

 

[tomatoma]

If this has been a long-standing Intel practice, mixing R&D and production in the same environment, does it still make sense in today’s situation?

Competition is fierce, and every year Intel’s competitors are rolling out new products faster and more frequently. Can Intel’s R&D fabs truly concentrate on R&D under these circumstances? Unless Intel still believes its competitors aren’t fast enough to "catch up"?

Or perhaps some of Intel’s Oregon fabs do not participate in R&D and instead handle small quantity production runs. Could this be one of the reasons pushing Intel’s costs higher than other companies?

Occasionally, a research facility in Oregon will mass-produce the product for a short period of time before a mass-production factory is established.
I'm sure Meteor Lake was like that too.

 

[tomatoma]

I'm sure fab42 had enough space to install EUV.

 

[tomatoma]

So the press was told the yields and saw the yield at the show so now they know there are not yield issues?

just let me know when I have a choice of 5 different Notebooks with Panther lake that I can buy right away. I assume this will be March 2026. but let me know when it happens.

Please ask about the product cost at this months earnings report: "Will Panther lake have higher or lower cost than Arrow lake and Raptor Lake in Q1 2026? Will the product operating margins on Panther Lake (18A) be above or below zero in Q1 2026. They have the numbers.....

No, it won't reach us consumers until next year, but the PTL will definitely reach the OEMs within the year, so there's no doubt that the 18A will start production in the second half of 2025.

 

[tomatoma]

I don't recall many Intel products shipping for the first time in Q4. Mostly, Intel ships a new model in the spring. They do play games with the year of first shipment, backdating it to the prior year, probably as an accounting trick. I know some people think of that as how Intel claimed process leadership without delivering product, vaporware strategy, but I think it's accountants who dictate clearing old inventory in Q4.

Foundries don't have inventory, they don't own the wafers, so they claim an earlier date to realize revenue.

Rather than waiting until March, it may be possible to get it as early as January next year.

 

[eding42]

Betting most heavily on this one.

This can't be right though, Intel's been manufacturing its iGPUs on internal nodes since Sandy Bridge.

Most recently, Raptor Lake had some decently large 96 EU designs on Intel 7. Ofc the density kinda sucked but that's more of an Intel 7 not being optimized for that kind of usage.

Really excited for the rumored Xe3P GPU tile on 18a in Nova Lake.

 

[Artificer60]

If this has been a long-standing Intel practice, mixing R&D and production in the same environment, does it still make sense in today’s situation?

Competition is fierce, and every year Intel’s competitors are rolling out new products faster and more frequently. Can Intel’s R&D fabs truly concentrate on R&D under these circumstances? Unless Intel still believes its competitors aren’t fast enough to "catch up"?

Or perhaps some of Intel’s Oregon fabs do not participate in R&D and instead handle small quantity production runs. Could this be one of the reasons pushing Intel’s costs higher than other companies?

In order to truly develop a viable process you need to ramp it to some level in order to find the most obvious flaws. That is what people are referring to when they talk about "yield learning". You want your R&D team involved in that process early on, so I believe that doing this makes sense. It also makes economic sense, because you have invested in installing a process line to produce the wafers during development anyway. Knowing that it will take several months to get the next fab up and running on the new process why wouldn't you use that capacity to start building inventory for your initial launch while you wait for the HVM factory to come on line.

While Oregon has multiple fabs, they are in essence all really one big fab. While kind of hard to see exactly the red ovals in this picture show links connecting the clean rooms.

Just my opinion, but I feel like Intel's costs are higher because historically they didn't care. The goal was technology leadership and if that cost more, so be it. We live in a different world now and if Intel is going to be successful, they need to learn some hard lessons, including cost consciousness and developing robust processes with larger process windows.

 

[Michael Domnitei]

This is going to be interesting:

Intel's 18A production starts before TSMC’s competing N2 tech — here's how the two process nodes compare

Production starts at Fab 52.

www.tomshardware.com

 

[benb]

Or perhaps some of Intel’s Oregon fabs do not participate in R&D and instead handle small quantity production runs. Could this be one of the reasons pushing Intel’s costs higher than other companies?

Intel was able to do 5 nodes in 4 years because the lead fab has the mission to lead, seed and the seeds transfer to another fab. It used to be every other year a new node, before all the difficulty starting in the 2010 until present. The Ronler Acres Moore Park complex is a grand scale. There is probably a cheaper way to do R&D but that is the Intel way. It worked better when fab lite wasn't a viable option as it is now.

On the topic of taking fab lite off the table for Intel, so they focus on the old tick tock, now that Intel's capital structure is different, with Trump and SoftBank and Nvidia, perhaps they can sack the board members who still want to see fab lite happen? Make a statement that fab lite is dead?