Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/intel-at-deutsche-banks-2025-technology-conference.23497/
        )

    [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
)

Intel at Deutsche Bank's 2025 Technology Conference

hist78

hist78

Well-known member
  • Summary:

  • ~ Government Grants: Intel received $5.7 billion from the U.S. government, under the CHIPS Act, now being converted into a stake in the company. The $5.7B grant was received the night before the conference, bolstering Intel’s balance sheet and reducing the need for external capital in the near term. The U.S. government’s equity stake is seen as a positive endorsement that could help attract foundry customers, though execution—quality, capacity, and customer satisfaction—will ultimately determine success.

  • ~ SoftBank Investment: SoftBank has also injected $2 billion into Intel.

  • ~ Debt Reduction Plans: Intel aims to reduce debt by $3.8 billion this year, with additional deleveraging planned into the next year.

  • ~ CapEx and Margins: Maintaining CapEx at around $18 billion annually, with gross margins near 40%, and a goal to improve.

  • ~ Foundry Business: Intel is pushing its foundry segment, targeting volume production of its 18A process by 2028–2029, though securing external customer commitments remains a challenge.

  • ~ Product Roadmap: Upcoming launches include Panther Lake, Nova Lake, Lunar Lake, Diamond Rapids, and Coral Rapids. Adjustments are underway in the data center segment to address multi-threading gaps.

Audio Replay:
 
hist78 said:
Here is the Google NotebookLM AI generated audio overview:

Audio overview: https://notebooklm.google.com/noteb...tifactId=c510e4e2-1ae8-4880-8caa-8ec8babcf069
Click to expand...
Neat tool. Some insights about Foundry that seemed interesting to me:
The upcoming 0.5 PDK (Process Design Kit) for 14A will be a key milestone for customers to evaluate the technology.
Click to expand...
Intel is targeting high-performance compute (HPC) customers due to the advantages of its backside power solution, but also sees opportunities in the handset/mobile segment.
Click to expand...
CEO Lipbu will only declare a customer win for 14A when there's a "real commitment and signature". Intel expects next year (2026) to be a key period for evaluating customer wins, with potential wins also in 2027. The industry generally anticipates volume ramp for leading-edge customers at the 14A node in the late 2020s (2028-2030).
Click to expand...

Technology Progress (18A and 14A):
Click to expand...
18A has shown significant improvement, now being in a "good really good place on the performance" with "steady incremental improvement on yields," despite earlier elongation in yield stabilization
Click to expand...
14A was designed "from the ground up" as a foundry node, resulting in a significantly different maturity of PDKs and greater rigor in its ecosystem compared to 18A at a similar stage. It is "already out of the gate looking better".
◦ Learnings from 18A are being applied to 14A to enhance defect density and performance.
◦ Both 18A and 14A are noted for their "very good cost structures" relative to older nodes and better pricing.
Click to expand...
 
And on Product:
Challenges: There's more work needed to achieve strong performance across the entire data center spectrum and meet customer expectations. CEO Lipbu has "torn apart that strategy" and identified "pockets" or "gaps" in the roadmap, particularly around multi-threading, which are being addressed.
◦ Roadmap Adjustment: Adjustments are being made to deliver products that customers want and need, but this will be a multi-year process. Incremental improvements are expected over the next couple of years.
◦ Upcoming Products:
▪ Diamond Rapids is expected to be better than Granite Rapids but "doesn't get us quite there" in closing the performance gap, performing better in some cases but not others.
▪ Coral Rapids (which Lipbu reportedly named in some forum) is seen as the "real opportunity" to take a significant step forward in performance.
• Product Architecture and Foundry Use:
◦ Tile Architectures: Intel's architecture is increasingly moving towards tile structures, which provides flexibility for products to "pick and choose" where their silicon is produced. This means that while Intel Foundry will likely provide the "majority of our wafers" for products, Intel will "continue to be a significant purchaser of external wafers" for solutions that require them to achieve the best combination of price and performance.
◦ Foundry Support: Generally, Intel's data center products are done internally to help support the Intel Foundry business.
◦ Panther Lake: This product will utilize the 18A process from Intel Foundry.
• AI and GPU Mentions:
◦ The source mentions that SoftBank, a recent investor in Intel, is "very much invested in AI," and Intel is looking for opportunities to "intersect with that" by providing solutions useful for them or in concert with them.
◦ In the data center, Intel performs well in "the head node which is a CPU that runs alongside the GPU". This indicates Intel's CPU products are designed to complement GPUs, which are critical for AI workloads.
◦ Beyond these points, the source does not detail specific AI-focused products or discrete GPU product updates from Intel.Intel's Chief Financial Officer, Dave Zinsner, provided updates on various Intel product segments, though specific details regarding AI and GPUs were more general, particularly concerning the data center.
Click to expand...
So, Diamond Rapids still not catching up (to AMD), and we have to wait for Coral Rapids to tackle the performance deficit? Ouch.

Also, literally no update on GPU or any * Shores product… woof.
 
"Intel is pushing its foundry segment, targeting volume production of its 18A process by 2028–2029"

I thought 18A was targeting volume production in 2025? Or are they saying that they first expect to fab for external customers starting in 2028?
 
Brady said:
And on Product:

So, Diamond Rapids still not catching up (to AMD), and we have to wait for Coral Rapids to tackle the performance deficit? Ouch.

Also, literally no update on GPU or any * Shores product… woof.
Click to expand...
I think Intel's product roadmap will probably be released soon.
Well, the policy may already be decided within the company lol
 
siliconbruh999 said:
this is 14A iirc
Click to expand...
There are no significant 18A customers. So the plan is for 2nd wave customers. So 2028 Is the plan for external customers .....

Which leads us to my age old question: How long is it from the time you commit to a foundry until you tape out production stepping? If someone chooses Intel process in 2026, when would they be able to tape out? How long is it to port an existing design from TSMC to Intel?
 
Brady said:
I think Gemini hallucinated, but also is 2027-2028 for full 18A volume actually that far off? I thought when Gelsinger presented that stacked chart showing node wafer share over time, 18A taking over 7 and 3 was very long off and 2027 doesn’t sound wrong for when 18A actually becomes majority of leading edge Intel Foundry output?
Click to expand...
He stated that 18A will be a modest ramp in 2026. Intel 7 is currently the volume leader for both PC and DC processors.

That stacked chart (which I updated by fab) is gone. The spending was too much. They will publish the new product roadmap in October I am guessing.
 
MKWVentures said:
Which leads us to my age old question: How long is it from the time you commit to a foundry until you tape out production stepping? If someone chooses Intel process in 2026, when would they be able to tape out? How long is it to port an existing design from TSMC to Intel?
Click to expand...
TSMC to Intel Port will take time don't know exact time but if someone Tapes out in 2026 You can add 14-18 Months from Tape out date to availability in market.
MKWVentures said:
He stated that 18A will be a modest ramp in 2026. Intel 7 is currently the volume leader for both PC and DC processors.
Click to expand...
we can take Intel 7 as a reference it took like 3-4 years for 10nm to ramp up and surpass 14nm if we count from Ice Lake launch date we can apply this to 18A and it would roughly in good volume by 2028-29.
 
siliconbruh999 said:
TSMC to Intel Port will take time don't know exact time but if someone Tapes out in 2026 You can add 14-18 Months from Tape out date to availability in market.
Click to expand...
Agree on that. How long does it take to port over. I have heard 2+ years due to optimization but I know people have moved designs and taped out a year later. sometimes less.

Can Intel pull the trigger today and move a N3 product to Intel 3 and tapeout in 1 year? I get conflicting answers

siliconbruh999 said:
we can take Intel 7 as a reference it took like 3-4 years for 10nm to ramp up and surpass 14nm if we count from Ice Lake launch date we can apply this to 18A and it would roughly in good volume by 2028-29.
Click to expand...
Agreed
 
MKWVentures said:
Agree on that. How long does it take to port over. I have heard 2+ years due to optimization but I know people have moved designs and taped out a year later. sometimes less.
Click to expand...
For Intel or other companies ? cause Intel can do it in a lot less time vs other companies.
MKWVentures said:
Can Intel pull the trigger today and move a N3 product to Intel 3 and tapeout in 1 year? I get conflicting answers
Click to expand...
From N3 to I3 is a large difference in Area the Power and Performance can be similarish or slightly behind N3E but Area difference is large so the die would grow quite a bit and can't give a time on something i don't have concrete info on.

We have Xe3 IP(Their GPU IP) on both Intel 3 and N3E in Panther Lake that are being developed at the same time
 
Last edited:
hist78 said:
Summary:
  • ~ Government Grants: Intel received $5.7 billion from the U.S. government, under the CHIPS Act, now being converted into a stake in the company. The $5.7B grant was received the night before the conference, bolstering Intel’s balance sheet and reducing the need for external capital in the near term. The U.S. government’s equity stake is seen as a positive endorsement that could help attract foundry customers, though execution—quality, capacity, and customer satisfaction—will ultimately determine success.

  • ~ SoftBank Investment: SoftBank has also injected $2 billion into Intel.
  • ~ Debt Reduction Plans: Intel aims to reduce debt by $3.8 billion this year, with additional deleveraging planned into the next year.
  • ~ CapEx and Margins: Maintaining CapEx at around $18 billion annually, with gross margins near 40%, and a goal to improve.
  • ~ Foundry Business: Intel is pushing its foundry segment, targeting volume production of its 18A process by 2028–2029, though securing external customer commitments remains a challenge.
  • ~ Product Roadmap: Upcoming launches include Panther Lake, Nova Lake, Lunar Lake, Diamond Rapids, and Coral Rapids. Adjustments are underway in the data center segment to address multi-threading gaps.
Audio Replay:
Click to expand...

Interesting to hear but nothing new really. I’m wondering if 14A will use High NA EUV? It seems like a waste of CAPEX if 14A only uses it on a limited basis.
 
siliconbruh999 said:
For Intel or other companies ? cause Intel can do it in a lot less time vs other companies.

From N3 to I3 is a large difference in Area the Power and Performance can be similarish or slightly behind N3E but Area difference is large so the die would grow quite a bit and can't give a time on something i don't have concrete info on.

We have Xe3 IP(Their GPU IP) on both Intel 3 and N3E in Panther Lake that are being developed at the same time
Click to expand...
Good input. So Xe3 is being developed in parallel, so you could put it into N3 or I3 with no delay (Sizing is different I know). Correct?
 
hist78 said:
My guess is no, because Intel’s scale and financial strength won’t allow it to take on High-NA EUV alone, even if the technology is almost ready for deployment.
Click to expand...
However, in reality, Intel is hi-na But in reality, Intel bought Hi-Na EUV...
There is no option to not use it
 
You must log in or register to reply here.
Back
Top