Intel Reportedly Places 2nm Orders For Nova Lake At TSMC; Foundry Division Likely To Be Left Out For Now

[fansink]

Intel has reportedly placed 2nm orders at TSMC for their Nova Lake CPUs, as Team Blue plans out a massive comeback in the desktop processor segment.

Intel Plans To Stick With Its "Dual-Sourcing" Approach With TSMC For Future Desktop CPUs, Signaling A Strategic Shift​

Well, it seems like Team Blue won't hold back when it comes to the process node it utilizes for future CPUs, as it looks like, under the new leadership, Intel has made it a priority to ensure that it delivers the best to its consumers. According to a report by Taiwan Economic Daily, it is revealed that Intel has outsourced Nova Lake's compute tile to TSMC, where it will leverage the foundry's 2nm process. Apparently, Intel plans to go all-out with the Taiwanese giant, which raises doubts about the fate of Intel's 18A process node, which is "advertised" to be better than TSMC N2.

When it comes to using TSMC for semiconductor needs, Intel's Products CEO Michelle Johnston Holthaus made it clear that the firm won't hesitate to opt out of Intel Foundry, stating that Team Blue needs to meet consumer expectations. So, the possibility of acquiring cutting-edge nodes from TSMC was always there, but the one thing that makes this development a bit confusing is the way Team Blue would use IFS for its product lineup, given that Intel would certainly not rely on it for its flagship offerings, so a dual-source approach will likely be adopted.


TSMC's 2nm process has received much attention from industry giants, given that Apple, AMD, and now Intel are in the race to acquire the node. AMD recently announced that they were the first customer of 2nm, integrating the semiconductor for its 6th-generation EPYC "Venice" processors. Similarly, Apple plans to adopt the process for its A20 chip intended for the iPhone 18 series, and now, Intel will also use it for Nova Lake, so it is safe to say that attention is there, but it is important to note that 2nm integration will turn out to be much more costlier compared to previous-gen processes, especially with the supply chain uncertainty.

As far as Intel Foundry is concerned, well, we are going to see 18A with Panther Lake SoCs and Clearwater Forest Xeon, which means that Intel doesn't plan on leaving its in-house processes alone. However, the prospect would likely be decided by how PTL-S and other products play out, but for now, it is safe to say that the company would stick with a dual-sourcing approach.

Intel Reportedly Places 2nm Orders For Nova Lake At TSMC; Foundry Division Likely To Be Left Out For Now

Intel has placed 2nm orders at TSMC for their Nova Lake CPUs, as Team Blue plans out a massive comeback in the desktop processor segment.

wccftech.com

 

[fansink]

I thought Lip-Bu Tan was going to bring transparency back (as though it was ever there), because being transparent would mean coming clean about the health of 18A.

With Nova Lake ~18 months away, if 18A is doing well NOW (as we're being led to believe), surely it will be ROCKING in 18 months. Sounds as transparent as a black hole.

 

[MKWVentures]

I thought Lip-Bu Tan was going to bring transparency back (as though it was ever there), because being transparent would mean coming clean about the health of 18A.

With Nova Lake ~18 months away, if 18A is doing well NOW (as we're being led to believe), surely it will be ROCKING in 18 months. Sounds as transparent as a black hole.

Lets see what they say. 18A could be completely healthy and high performance and TSMC could still be the best business decision for some products. A mix is a good idea until the financials make the long term decision obvious. Arrow lake was N3 then it was 20A then it was N3..... based on financials.

 

[fansink]

Lets see what they say. 18A could be completely healthy and high performance and TSMC could still be the best business decision for some products. A mix is a good idea until the financials make the long term decision obvious. Arrow lake was N3 then it was 20A then it was N3..... based on financials.

That's the crock of shite, "we'll use TSMC because it's the best business decision".

The "best business decision" is CODE for "because what we have internally is crap".

Why don't you give us a logical reason for, if 18A is doing well NOW (as we're being led to believe), surely it will be ROCKING in 18 months.

 

[siliconbruh999]

Well apparently from the leaks the 8+16(This is the least volume product but a flagship product) Die is N2 and 4+8+4/4+0 die is 18AP(This is most volume product and mainstream) for Nova Lake also there is unified HUB/SOC that is shared between desktop and Mobile this time.

 

[fansink]

Well apparently from the leaks the 8+16(This is the least volume product but a flagship product) Die is N2 and 4+8+4/4+0 die is 18AP(This is most volume product and mainstream) for Nova Lake also there is unified HUB/SOC that is shared between desktop and Mobile this time.

We all know that the ~leak is not from TSMC, ~it’s Intel’s form of Transparency.

It’s as Clear as Water, Panther Lake is for mobile platforms; it can use less than optimum 18A.

Clearwater Forest is server-focused for data centers; it cannot be on a less than optimum 18A.

Intel has <9 months to get 18A “running on all cylinders” for Clearwater Forest.

If intel had faith in 18A for Clearwater Forest (for servers), why would they hedge their bet by having Nova Lake (for desktops) produced by TSMC’s 2nm.

Because Intel has insufficient Faith in 18A, currently.

So instead of Transparency, we get ~leaks about Nova Lake being on 2nm, and very shortly, we will be getting Transparent ~leaks about Clearwater Forest being on 2nm as well.

 

[siliconbruh999]

We all know that the ~leak is not from TSMC, ~it’s Intel’s form of Transparency.

It’s as Clear as Water, Panther Lake is for mobile platforms; it can use less than optimum 18A.

Clearwater Forest is server-focused for data centers; it cannot be on a less than optimum 18A.

Intel has <9 months to get 18A “running on all cylinders” for Clearwater Forest.

First Clearwater Forest was delayed due to packing it uses Hybrid bonding it can very well be the issue.

If intel had faith in 18A for Clearwater Forest (for servers), why would they hedge their bet by having Nova Lake (for desktops) produced by TSMC’s 2nm.

Because Intel has insufficient Faith in 18A, currently.

So instead of Transparency, we get ~leaks about Nova Lake being on 2nm, and very shortly, we will be getting Transparent ~leaks about Clearwater Forest being on 2nm as well.

This is impossible otherwise Clearwater forest will be launching end of 2026 or start of 27 not Q1 26 as for Nova Lake on TSMC N2 this was hinted by pat Gelsinger in Q3 24 earnings call so not really a suprise.

 

[hist78]

We all know that the ~leak is not from TSMC, ~it’s Intel’s form of Transparency.

It’s as Clear as Water, Panther Lake is for mobile platforms; it can use less than optimum 18A.

Clearwater Forest is server-focused for data centers; it cannot be on a less than optimum 18A.

Intel has <9 months to get 18A “running on all cylinders” for Clearwater Forest.

If intel had faith in 18A for Clearwater Forest (for servers), why would they hedge their bet by having Nova Lake (for desktops) produced by TSMC’s 2nm.

Because Intel has insufficient Faith in 18A, currently.

So instead of Transparency, we get ~leaks about Nova Lake being on 2nm, and very shortly, we will be getting Transparent ~leaks about Clearwater Forest being on 2nm as well.

In early March, Intel clarified this multi foundries strategy at a Morgan Stanley conference:

"Q - Joe Moore (Semiconductor Industry Analyst at Morgan Stanley)

So, you went through a CEO transition, lots of transformational stories coming out every day. But it seems like from the way you're talking about on the earnings call and the meetings you've been doing, strategy is still basically the same. And so maybe give us some perspective on what's happening at Intel now from more of a strategic perspective? It seems like you're still aiming towards the same general goals that you had under, Pat?

John Pitzer (Corporate Vice President, Investor Relations at Intel Corporation)

Yes. I mean, it's a good question. I think the core of the strategy continues to be standing up a world-class fabless company and a world-class foundry. I think on the margin, as you think about Dave and Michelle in the interim CEO roles, they have been given full agency to go off and prosecute the strategy, I think on the margin. I think from an Intel products perspective, I think there's a strong understanding that without that being healthy first. It's really hard to have a healthy Intel foundry.

So, I do think Michelle has been given a bit more agency to go off and make decisions around what she thinks is best for the product road map and for longer-term market share there. And I think the biggest sort of manifestation of that today is she probably has more agency to use TSMC longer and for more than she might have had six or nine months ago. And to be clear, we have roughly 30% of our wafers outsourced today. That is probably a high watermark for us.

But to the extent that I think a year ago, we were talking about trying to get that to zero as quickly as possible. That's no longer the strategy. We think it's always good to have at least some of our wafers with TSMC. They're a great supplier. It creates a good competition between them and Intel Foundry. Not quite sure what the right sort of level set is. Is it 20? Is it 15? We're working through that. But we will use, I think, external foundry suppliers longer kind of under this new strategy."

Source:

It's official: Intel will continue outsourcing up to 30% of its wafer production to TSMC

"Q - Joe Moore (Semiconductor Industry Analyst at Morgan Stanley) So, you went through a CEO transition, lots of transformational stories coming out every day. But it seems like from the way you're talking about on the earnings call and the meetings you've been doing, strategy is still...

semiwiki.com

 

[dkr1986]

First Clearwater Forest was delayed due to packing it uses Hybrid bonding it can very well be the issue.

This is impossible otherwise Clearwater forest will be launching end of 2026 or start of 27 not Q1 26 as for Nova Lake on TSMC N2 this was hinted by pat Gelsinger in Q3 24 earnings call so not really a suprise.

View attachment 3061

I was about to post it, but you beat me to it. Michelle (CEO of Products) said they have the flexibility to choose foundry partner based on PPA, time to market window & cost targets, who knows what led them to choose N2 .

Could be just iGPU tiles for NVL-H laptops too (less likely). So far, people here have said 18A performance is somewhere between N3P and N2, with AMD choosing N2, may be Intel needed that extra juice to stay competitive.

EDIT - Just to clarify I think Intel is only using N2/N2P only for the top end NVL-SK SKU per the leaks and everything else is likely inhouse on 18A or 18A-P (Unless Intel repeats the ARL 20A to N3B switch at the last minute again, which I think is unlikely).

 

[hist78]

In early March, Intel clarified this multi foundries strategy at a Morgan Stanley conference:

"Q - Joe Moore (Semiconductor Industry Analyst at Morgan Stanley)

So, you went through a CEO transition, lots of transformational stories coming out every day. But it seems like from the way you're talking about on the earnings call and the meetings you've been doing, strategy is still basically the same. And so maybe give us some perspective on what's happening at Intel now from more of a strategic perspective? It seems like you're still aiming towards the same general goals that you had under, Pat?

John Pitzer (Corporate Vice President, Investor Relations at Intel Corporation)

Yes. I mean, it's a good question. I think the core of the strategy continues to be standing up a world-class fabless company and a world-class foundry. I think on the margin, as you think about Dave and Michelle in the interim CEO roles, they have been given full agency to go off and prosecute the strategy, I think on the margin. I think from an Intel products perspective, I think there's a strong understanding that without that being healthy first. It's really hard to have a healthy Intel foundry.

So, I do think Michelle has been given a bit more agency to go off and make decisions around what she thinks is best for the product road map and for longer-term market share there. And I think the biggest sort of manifestation of that today is she probably has more agency to use TSMC longer and for more than she might have had six or nine months ago. And to be clear, we have roughly 30% of our wafers outsourced today. That is probably a high watermark for us.

But to the extent that I think a year ago, we were talking about trying to get that to zero as quickly as possible. That's no longer the strategy. We think it's always good to have at least some of our wafers with TSMC. They're a great supplier. It creates a good competition between them and Intel Foundry. Not quite sure what the right sort of level set is. Is it 20? Is it 15? We're working through that. But we will use, I think, external foundry suppliers longer kind of under this new strategy."

Source:

It's official: Intel will continue outsourcing up to 30% of its wafer production to TSMC

"Q - Joe Moore (Semiconductor Industry Analyst at Morgan Stanley) So, you went through a CEO transition, lots of transformational stories coming out every day. But it seems like from the way you're talking about on the earnings call and the meetings you've been doing, strategy is still...

semiwiki.com

Typically, a corporate IR (Investor Relations) person - such as John Pitzer, Corporate Vice President of Investor Relations at Intel - knows how to speak carefully, as their words can have significant financial and legal implications for the company.

For Intel’s John Pitzer to publicly state that Intel needs TSMC (because TSMC is a great supplier) and that Intel Foundry needs healthy competition is a very unique situation. It suggests that Intel Foundry may be facing serious issues that the company must address with the help of TSMC.

 

[Daniel Nenni]

From what I can see there will be significantly more N2 tape outs than N3 at this point in time which is quite impressive considering N3 is a dominant node for TSMC. That may be hard to do without Intel using N2 and yes I agree there should be some transparency here. The TSMC Symposium is tomorrow, we should know more then.

 

[fansink]

First Clearwater Forest was delayed due to packing it uses Hybrid bonding it can very well be the issue.

This is impossible otherwise Clearwater forest will be launching end of 2026 or start of 27 not Q1 26 as for Nova Lake on TSMC N2 this was hinted by pat Gelsinger in Q3 24 earnings call so not really a suprise.

View attachment 3061

Are you intimating that 18A HVM will be before TSMC's 2nm HVM.

If so, you would be absolutely incorrect.

 

[fansink]

From what I can see there will be significantly more N2 tape outs than N3 at this point in time which is quite impressive considering N3 is a dominant node for TSMC. That may be hard to do without Intel using N2 and yes I agree there should be some transparency here. The TSMC Symposium is tomorrow, we should know more then.

Are you referring to a D0/Q point in time, or an N2 is ~6 months ahead of 18A point in time, or other?

 

[MKWVentures]

That's the crock of shite, "we'll use TSMC because it's the best business decision".

The "best business decision" is CODE for "because what we have internally is crap".

Why don't you give us a logical reason for, if 18A is doing well NOW (as we're being led to believe), surely it will be ROCKING in 18 months.

Its really simple [hypothetically]: 18A is a high performing process with BSPD. It is very expensive due to output per tool limitations that Intel is discovering. IFS foundry customers have not committed to it. As a result, the volume is low and if we keep it limited to one production fab, we can maybe break even by 2027 on 18A. If we need to ramp another fab, it will cost us billions in capex and the new fab will not be fully loaded. If the product gets delayed or doesn't ramp as expected, we lose billions. With TSMC we spend no capex, we can move wafer starts up and down we can adjust loadings on other products. Our volume is <10% of the N2 output.... lots of flexibility.

Pats whole goal was foundry to help keep costs down. But apparently Intel still does not have 2 or 3 key customers according to Tan. Given the design timelines. its getting too late.

I have said all along.... and Intel will tell you soon.... its not about the technology. Its about the finances .... You can easily do the spreadsheet to show this.

So how will the operating margins look for 18A process and products at Intel in 2026?

 

[Daniel Nenni]

First Clearwater Forest was delayed due to packing it uses Hybrid bonding it can very well be the issue.

This is impossible otherwise Clearwater forest will be launching end of 2026 or start of 27 not Q1 26 as for Nova Lake on TSMC N2 this was hinted by pat Gelsinger in Q3 24 earnings call so not really a suprise.

View attachment 3061

Intel can use TSMC N3 for the 18A supporting tiles, no problem. Last year TSMC mentioned that all but one N3 customer has signed up for N2. It was my understanding that Intel was that customer. This was last year at the first TSMC event so maybe things have changed. I do know that Intel signed up for a lot of N3 wafers in the first agreement, it was very optimistic from what I remember.

 

[Daniel Nenni]

Here is the source article:

TSMC 2nm Intel orders, legal person optimistic about second half performance​

Market sources say Intel has joined TSMC's first batch of 2nm customers, which will be used in next-generation PC processors. (Reuters)

2025/04/22 02:06:44
Economic Daily reporter Zhong Huiling/Taipei report

TSMC plans to mass produce 2nm process in the second half of the year. Recently, related customers have come to light. After AMD announced its order for TSMC's 2nm, the market has heard that Intel has also joined the ranks of TSMC's first batch of 2nm customers. It will be used to produce next-generation PC processors and is currently preparing for trial production at TSMC's Hsinchu plant to facilitate subsequent adjustments to the yield rate.

In response, TSMC said it would not comment on market rumors or the business of specific customers. Intel also declined to comment on the news. Analysts are optimistic that as 2nm customer orders continue to pour in, TSMC's performance will continue to improve in the second half of the year.

Intel is already one of TSMC's major customers for advanced processes. In February last year, Intel's then-CEO Henry Kissinger confirmed that the company had handed over the key compute tiles of two processors to TSMC for the first time. These two products were the later Intel Core Ultra 200V series laptop processors (code-named Lunar Lake) and the company's first AI PC desktop processor, the Intel Core Ultra 200S series (code-named Arrow Lake).

Among them, the computing chip blocks are produced using TSMC's N3B process, the GPU chip blocks are produced using TSMC's N5P process, and the SoC and I/O chip blocks are produced using TSMC's N6 process.

Intel's CEO is now led by Lip-Wu Chen, and the latest news is that the company will continue to use TSMC's 2nm production for its own products.

It is understood that the two parties are currently only cooperating on one product in the 2nm process. It is speculated that this may be the computing chip block of Intel's PC processor Nova Lake that will be launched next year.

Regarding the progress of 2nm, TSMC disclosed in its latest shareholder meeting report that the development of 2nm process technology is proceeding according to plan and has made good progress. It uses the first-generation nanochip transistor technology to provide performance and power consumption improvements for all process nodes.

TSMC also mentioned that major customers have completed 2nm silicon intellectual property (IP) design and started verification. TSMC is also developing low-resistance reset conductor layers and ultra-high-performance metal layer capacitors to continue improving the performance of its 2nm process technology. Its 2nm process technology leads the industry in meeting customers' insatiable demand for energy-efficient computing, and nearly every IC innovator is working with it.

台積電2奈米 英特爾下單 法人看好下半年業績熱轉 | 產業熱點 | 產業 | 經濟日報

台積電規劃下半年量產2奈米製程，近期相關客戶陸續浮上檯面，繼超微宣布下單台積電2奈米後，市場傳出，英特爾也加入台積電2奈...

money.udn.com

 

[MKWVentures]

I think there is a lot of parallel design efforts still going on. And Sometime Intel renames one lake to another lake for marketing purposes.
Remember when Arrow lake was 20A. Then it was TSMC N3 on some skus. Then it was N3 on all skus. There are 1 or 2 refreshes coming as well.

Don't get too married to what Intel says or even what they are sampling. The roadmap will change significantly by mid 2026

 

[fansink]

Intel's CEO is now led by Lip-Wu Chen, and the latest news is that the company will continue to use TSMC's 2nm production for its own products.

It is understood that the two parties are currently only cooperating on one product in the 2nm process. It is speculated that this may be the computing chip block of Intel's PC processor Nova Lake that will be launched next year.

Oh, now I get it.

Intel's Clearwater Forest, for Data Centers, will use 18A (critical for boosting their current less than stellar image); and then later

Intel's Nova Lake, for Desktops, will use N2 (where their image can take a back-seat due to financial constraints per @MKWVentures)

It's all clear now, thanks.

 

[kevin01]

Oh, now I get it.

Intel's Clearwater Forest, for Data Centers, will use 18A (critical for boosting their current less than stellar image); and then later

Intel's Nova Lake, for Desktops, will use N2 (where their image can take a back-seat due to financial constraints per @MKWVentures)

It's all clear now, thanks.

Many of these decisions were made a while ago.

And Clearwater Forest is not really a high-volume high-value product, BTW. According to Intel, these E-core Xeons have limited interest from the industry partners. We could corroborate that statement with Ampere Computing's poor performance.

 

[fansink]

Many of these decisions were made a while ago.

And Clearwater Forest is not really a high-volume high-value product, BTW. According to Intel, these E-core Xeons have limited interest from the industry partners. We could corroborate that statement with Ampere Computing's poor performance.

I can totally relate, it wouldn't be prudent to use 18A on a high-volume or high-value product.

Especially since N2's lifecycle is ~6 months ahead of 18A's lifecycle.

N2 is clearly the prudent move for Intel's high-volume or high-value products.