Intel internal foundry model and IDM 2.0

[hist78]

Source: https://www.intel.com/content/www/u...mbraces-internal-foundry-model.html#gs.fj6qgt


Intel CEO Pat Gelsinger sent the following statement announcing creation of an internal foundry model for external customers and Intel product lines and the creation of the IDM 2.0 Acceleration Office, under the leadership of Stuart Pann, to all company employees on Oct. 11:

When I returned to Intel in 2021, I introduced our IDM 2.0 strategy: a multiyear journey to regain unquestioned technology leadership, manufacturing scale and long-term growth. Today, we begin the next phase of our IDM 2.0 journey.

In the first phase of our transformation, we made significant progress on our process roadmaps and capacity. We remain on track to deliver five process nodes in four years, and we’ve invested in the capacity required to meet the industry’s demand for semiconductors, bringing much-needed balance to the global supply chain.

The next phase of our IDM 2.0 journey requires a fundamental shift in mindset. We must embrace an internal foundry model, not only for our external customer commitments but also for our Intel product lines. This is a significant evolution in how we think and operate as a company, but the systems and infrastructure that served us well in the IDM 1.0 world will not enable us to achieve the full potential of IDM 2.0.

Internal Foundry Model Explained

Implementing an internal foundry model means establishing consistent processes, systems and guardrails between our business unit, design and manufacturing teams. This will allow us to identify and address structural inefficiencies that exist in our current model by driving accountability and costs back to decision-makers in real time. It will also put Intel’s product groups on a similar footing as external Intel Foundry Services customers and vice versa.

For example, our business unit and design teams will be able to consider the potential impact on their margins if they want to run an additional product stepping, while the manufacturing team will be able to assess requests based on actual costs and impact on factory output. This will give teams the tools and transparency they need to find the most effective and cost-efficient solutions before implementation in silicon, ultimately helping us maximize factory output, reduce costs and shorten design cycles.

We will also create a foundry accounting model that encompasses manufacturing, technology development and Intel Foundry Services. This will give us more transparency into our financial execution and will allow us to fully benchmark and drive ourselves to best-in-class foundry performance.

IDM 2.0 Acceleration Office (IAO)

To facilitate this effort, I have established an IDM 2.0 Acceleration Office, which will be led by Stuart Pann, Senior Vice President of our Corporate Planning Group (CPG). In addition to his responsibility leading CPG, Stu will take on the role of Chief Business Transformation Officer, reporting to Chief Financial Officer Dave Zinsner and joining the executive leadership team. The IDM 2.0 Acceleration Office will work in close collaboration with all business units and functional teams to bring this new internal foundry model to life.

Execution Excellence

Alongside initiatives like the next evolution of our tick-tock model and our Gladius program for driving excellence in product development, embracing an internal foundry model will enable us to deliver the competitive cost structure and predictable cadence of leadership products essential to our success.

We will continue to provide regular updates on these initiatives and our progress in restoring our culture of execution excellence.

This is our time to take action as One Intel to unleash IDM 2.0’s full potential. Thank you in advance for your hard work and commitment to our strategy.

 

[hist78]

I think Intel is working on an option to split itself into two independent companies, a foundry business and a fabless product company.

This is an option Intel may not take eventually but they can't afford not to prepare for it.

 

[blueone]

When I read this announcement the first thought that came to my mind was that the CPU guys were being pushed to reduce their number of all-layer steppings to get a CPU to GA. Sapphire Rapids supposedly has done 5 all-layer steppings, and 12 total. They're apparently on E5...

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

overclock3d.net

(Since this information isn't sourced from Intel it is hearsay, but previous Intel CPU histories make the information seem quite credible.)

This is one reason I've always been a fan of tiled dies for big chips. With one huge die and massive complexity this is what happens. I wonder how Nvidia, the other high volume massive die company, handles this problem...

 

[hist78]

When I read this announcement the first thought that came to my mind was that the CPU guys were being pushed to reduce their number of all-layer steppings to get a CPU to GA. Sapphire Rapids supposedly has done 5 all-layer steppings, and 12 total. They're apparently on E5...

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

overclock3d.net

(Since this information isn't sourced from Intel it is hearsay, but previous Intel CPU histories make the information seem quite credible.)

This is one reason I've always been a fan of tiled dies for big chips. With one huge die and massive complexity this is what happens. I wonder how Nvidia, the other high volume massive die company, handles this problem...

On the other hand, Intel's product design division might be pointing the finger at Intel foundry for issues that caused the delay, bugs, and/or cost increase. With the current market condition and IDM 2.0 strategy, it's hard to keep these two sister divisions happy under the same Intel roof.

Also, I assume Pat Gelsinger has also given certain freedom to Intel Product Design division to use outside foundries.

 

[Daniel Nenni]

I think Intel is working on an option to split itself into two independent companies, a foundry business and a fabless product company.
This is an option Intel may not take eventually but they can't afford not to prepare for it.

IDM 2.0 is evolving. Samsung did the same thing on paper. They spun out the logic fabs and call them foundry even though quite a bit of that capacity is for Samsung internal chips. If Intel does that and counts the revenue of internal chips as foundry business, like Samsung does, Intel will be the #1 foundry.

 

[blueone]

IDM 2.0 is evolving. Samsung did the same thing on paper. They spun out the logic fabs and call them foundry even though quite a bit of that capacity is for Samsung internal chips. If Intel does that and counts the revenue of internal chips as foundry business, like Samsung does, Intel will be the #1 foundry.

Bigger than TSMC? Since TSMC's revenue will be comparable to Intel's total revenue, how can this be?

 

[hist78]

Bigger than TSMC? Since TSMC's revenue will be comparable to Intel's total revenue, how can this be?

In August Intel reduced its 2022 revenue forecast from $76 billion to between $65 and $68 billion. TSMC 2022 revenue is expected to reach $76 billion. Please see my post in anther thread.

TSMC Earnings 3Q22

Financial results are out... https://investor.tsmc.com/english/quarterly-results/2022/q3 ==== Hsinchu, Taiwan, R.O.C., October 13, 2022 -- TSMC today announced consolidated revenue of NT$613.14 billion, net income of NT$280.87 billion, and diluted earnings per share of NT$10.83 (US$1.79 per...

semiwiki.com

 

[Daniel Nenni]

Bigger than TSMC? Since TSMC's revenue will be comparable to Intel's total revenue, how can this be?

In August Intel reduced its 2022 revenue forecast from $76 billion to between $65 and $68 billion. TSMC 2022 revenue is expected to reach $76 billion. Please see my post in anther thread.

What an interesting turn of events. That will be a big celebration! That would also put TSMC ahead of Samsung? Will the Mobileye IPO reduce Intel revenues even more?

 

[benb]

Interesting news, could be significant, time will tell what it means.

It’s a reorg, but the scope to be defined by Stu Penn?

Also in the background is the reality that Intel cost structure is too high for a Foundry, and they need to cut a lot of fat somehow. More contractors, fewer direct hires?

 

[blueone]

Also in the background is the reality that Intel cost structure is too high for a Foundry, and they need to cut a lot of fat somehow. More contractors, fewer direct hires?

I agree that Intel's cost structure is too high. IMO, the problem can't be solved with contractors, it'll take reducing the fat in non-engineering and manufacturing roles, like sales, marketing, planning, HR, and finance. There are still too many managers with just a few direct reports. There's also been a lot of grade-level bloat, even in engineering. Intel is just plain fat.

 

[blueone]

Will the Mobileye IPO reduce Intel revenues even more?

Probably by about $2B the first year. For the past year their revenue was $1.4B, and growing at ~40% annually.

 

[kingmouf]

I believe that splitting Intel into two companies, foundry and all the rest, is inevitable. If Intel Design is given freedom to choose other foundries, this is a huge blow to Intel Foundry. Unless they become separate entities, this will not work. The same goes the other way around: a delay in important products, that can be considered halo products for a new node, could severely impact Intel Foundry - no matter if the node is ready and working. Additionally, Intel has a very broad portfolio of products that practically competes with almost all the companies that may want to use leading edge manufacturing. I am still skeptical how this is going to work unless competitors are somehow forced to use Intel Foundry (e.g. if they have taken money from a certain funding source, if they are restricted to manufacture in the US/Europe for specific products, etc). We need to wait and see, but if I had to make a bet, I would say that the clock is ticking for Intel to split.

 

[tooLongInEDA]

I think Intel is working on an option to split itself into two independent companies, a foundry business and a fabless product company.

This is an option Intel may not take eventually but they can't afford not to prepare for it.

I agree. It leaves both options open until the data is gathered to make a final decision (or keep running both models). But does not imply that Intel splits yet. It is notable that he's ordered Intel to be pin compatible with the industry and not expected the industry to change one iota to accomodate Intel. That might be a first. Also that he quite explicitly says this is about benchmarking Intel's capabilities against the industry.

It all makes sense to me.

But the scale and pace of the change here is very ambitious.

 

[nghanayem]

I believe that splitting Intel into two companies, foundry and all the rest, is inevitable. If Intel Design is given freedom to choose other foundries, this is a huge blow to Intel Foundry. Unless they become separate entities, this will not work. The same goes the other way around: a delay in important products, that can be considered halo products for a new node, could severely impact Intel Foundry - no matter if the node is ready and working. Additionally, Intel has a very broad portfolio of products that practically competes with almost all the companies that may want to use leading edge manufacturing. I am still skeptical how this is going to work unless competitors are somehow forced to use Intel Foundry (e.g. if they have taken money from a certain funding source, if they are restricted to manufacture in the US/Europe for specific products, etc). We need to wait and see, but if I had to make a bet, I would say that the clock is ticking for Intel to split.

I suppose this is the point of putting up walls within the company. That way intel needs to be reasonably competitive with IFS customers, and so those customers can feel at ease sharing data with IFS. As for the IDM foundry, I don't see it being a showstopper, after all many ARM SOC makers work with Samsung foundry without issue or conflicts of interest. To get there though intel must put up these walls to ensure that customers have that confidence (as well as having good products that are price competitive).

 

[hist78]

I suppose this is the point of putting up walls within the company. That way intel needs to be reasonably competitive with IFS customers, and so those customers can feel at ease sharing data with IFS. As for the IDM foundry, I don't see it being a showstopper, after all many ARM SOC makers work with Samsung foundry without issue or conflicts of interest. To get there though intel must put up these walls to ensure that customers have that confidence (as well as having good products that are price competitive).

I thought conflict of interest was one of the reasons Samsung Foundry got into troubles and lost Apple's business?

To build a meaningful firewall between Intel Foundry division and Intel product/design division is very hard.

They need to have:

1. Two independent accounting departments. Otherwise how do they negotiate price and contracts with each other when they know each other's cost structure down to the last penny?

2. They need to have two totally independent departments for IT, HR, Procurement, and Training. Otherwise confidential informant can go across from one innocent training session or a simple inter company technical Q&A posting.

3. Each of them need to have their own cafeterias, gyms, libraries, conference rooms, elevators, parking lots and visitors entrance. Otherwise the firewall between them is as thick as a piece of paper. There is no privacy or secrecy between them.

4. Inter divisions job transfer is highly discouraged or prohibited. Otherwise how do they assure their own customers, suppliers, and partners there is no conflict interest and confidential information are well kept?

5. They need to have strict rules and enforcement on participating other division's functions. Such as New Year party, retirement party, golf outings, tradeshow's receptions and vendors/partners conferences.

After going through all these troubles, it's probably much easier to be two independent companies to begin with.

 

[nghanayem]

I thought conflict of interest was one of the reasons Samsung Foundry got into troubles and lost Apple's business?

To build a meaningful firewall between Intel Foundry division and Intel product/design division is very hard.

They need to have:

1. Two independent accounting departments. Otherwise how do they negotiate price and contracts with each other when they know each other's cost structure down to the last penny?

2. They need to have two totally independent departments for IT, HR, Procurement, and Training. Otherwise confidential informant can go across from one innocent training session or a simple inter company technical Q&A posting.

3. Each of them need to have their own cafeterias, gyms, libraries, conference rooms, elevators, parking lots and visitors entrance. Otherwise the firewall between them is as thick as a piece of paper. There is no privacy or secrecy between them.

4. Inter divisions job transfer is highly discouraged or prohibited. Otherwise how do they assure their own customers, suppliers, and partners there is no conflict interest and confidential information are well kept?

5. They need to have strict rules and enforcement on participating other division's functions. Such as New Year party, retirement party, golf outings, tradeshow's receptions and vendors/partners conferences.

After going through all these troubles, it's probably much easier to be two independent companies to begin with.

Apple left for A7 yes. But they came back for dual sourcing with A9 (because Samsung's 14nm was faster to market than N16) and they then left again after there was FUDD around the battery life of the Samsung versions. Besides Apple all of the other ARM SOC folks use Samsung without fear.

As for the accounting issue that is exactly what Pat is talking about; having separate accounting for better accountability between the divisions. A similar deal was already done with AXG getting a share of SOC revenues based on their contribution to the iGPUs within said SOC.

As for the physical separation that seems way overkill. Intel's processes/pdks are developed/headquartered in Oregon (and they share mask tapeouts with their old/repurposed California fabs I believe); if memory serves the design division is primarily in Israel these days - with a smaller footprint in California and Texas. All IP is likely already compartmentalized on a need to know basis (given this is standard practice at basically all businesses).

As for inter company job transfers could you give an example of how this would be an IP issue? While an extreme example it's not like a process engineer will transfer to the CPU design division, and even if they did they would know almost nothing about the products they are fabing.

 

[Daniel Nenni]

Apple left for A7 yes. But they came back for dual sourcing with A9 (because Samsung's 14nm was faster than N16) and they then left again after there was FUDD around the battery life of the Samsung versions. Besides Apple all of the other ARM SOC folks use Samsung without fear.

As for the accounting issue that is exactly what Pat is talking about; having separate accounting for better accountability between the divisions. A similar deal was already done with AXG getting a share of SOC revenues based on their contribution to the iGPUs within said SOC.

As for the physical separation that seems way overkill. Intel's processes/pdks are developed/headquartered in Oregon (and they share mask tapeouts with their old/repurposed California fabs I believe); if memory serves the design division is primarily in Israel these days - with a smaller footprint in California and Texas. All IP is likely already compartmentalized on a need to know basis (given this is standard practice at basically all businesses).

As for inter company job transfers could you give an example of how this would be an IP issue? While an extreme example it's not like a process engineer will transfer to the CPU design division, and even if they did they would know almost nothing about the products they are fabing.

Apple sued Samsung for IP theft and patent infringement and won a large settlement. It was a big deal. Apple still does business with Samsung for other parts of the phone but only if they have no other choice. I see zero possibility of Apple using Samsung foundry again for multiple reasons but competing with customers is a real problem in the foundry business. No matter how you spin it IDM foundries and Pure-play Foundries are two different things.

 

[nghanayem]

Apple sued Samsung for IP theft and patent infringement and won a large settlement. It was a big deal. Apple still does business with Samsung for other parts of the phone but only if they have no other choice. I see zero possibility of Apple using Samsung foundry again for multiple reasons but competing with customers is a real problem in the foundry business. No matter how you spin it IDM foundries and Pure-play Foundries are two different things.

For sure, and I also doubt Apple will use anything besides TSMC (unless something very bad happens). However I can't ignore that evidently it isn't a complete showstopper to have an IDM foundry since Apple went back to Samsung (even if it was only as a second source rather than the primary) for one generation after the lawsuit, and that this hasn't stopped Qualcomm or MediaTek (companies that are more direct competitors with Samsung than Apple is) from using Samsung to supply part of their demand.

Regardless this is still a MAJOR part of TSMCs value proposition (one you can't really put a price to). But if Samsung can snag customers that they literally compete with for SOCs within Samsung's own phones and appliances, than I don't see why intel can't put up enough walls in their operation for folks like Qualcomm, MediaTek, Broadcom, CISCO, and NVIDIA to feel comfortable enough to consider IFS.

 

[hist78]

Apple left for A7 yes. But they came back for dual sourcing with A9 (because Samsung's 14nm was faster than N16) and they then left again after there was FUDD around the battery life of the Samsung versions. Besides Apple all of the other ARM SOC folks use Samsung without fear.

As for the accounting issue that is exactly what Pat is talking about; having separate accounting for better accountability between the divisions. A similar deal was already done with AXG getting a share of SOC revenues based on their contribution to the iGPUs within said SOC.

As for the physical separation that seems way overkill. Intel's processes/pdks are developed/headquartered in Oregon (and they share mask tapeouts with their old/repurposed California fabs I believe); if memory serves the design division is primarily in Israel these days - with a smaller footprint in California and Texas. All IP is likely already compartmentalized on a need to know basis (given this is standard practice at basically all businesses).

As for inter company job transfers could you give an example of how this would be an IP issue? While an extreme example it's not like a process engineer will transfer to the CPU design division, and even if they did they would know almost nothing about the products they are fabing.

I'm talking everything, not just IPs. For example, if there's a Intel Product Division financial analyst who was just transfered from IFS. What will he/she say during a market competition analysis for the Intel product division?

If he/she knew through the experience at IFS that IFS is working with Mediatek for a secret product that will eat into 30% of Intel desktop CPU market. And there's another secret project that IFS is working with Qualcomm for a new laptop processor that can potentially kill some new Intel product designs?

Will he/she keep quiet?

 

[nghanayem]

I'm talking everything, not just IPs. For example, if there's a Intel Product Division financial analyst who was just transfered from IFS. What will he/she say during a market competition analysis for the Intel product division?

If he/she knew through the experience at IFS that IFS is working with Mediatek for a secret product that will eat into 30% of Intel desktop CPU market. And there's another secret project that IFS is working with Qualcomm for a new laptop processor that can potentially kill some new Intel product designs?

Will he/she keep quiet?

This is a good example, but a few questions on the probability of such an occurrence. The capability of said MediaTek product is bound by the intel node and packaging techniques that are being used, so how secret could it really be? Also I wouldn't know but presumably wouldn't spies know about this product and their targets in a similar timeframe (if not maybe faster)?

But yeah this sort of thing is why it would be hard to imagine someone like AMD using intel (even ignoring how good their standing/treatment with TSMC already is) no matter how strong the walls separating divisions are. But it is much easier to see for people like Qualcomm, MediaTek, or NVIDIA - 2nd sourcing some of their GeForce products (in a similar fashion to their Samsung relationship) - because these companies have a very small area of overlap.