Intel Layoff, Data Center Group, 2026-July

[Daniel Nenni]

Intel is gearing up for another round of layoffs targeting its Data Center Group as the chip giant keeps tightening the screws on costs and chasing efficiency. It is the latest move in a multi‑year wave of cuts that has already thinned the company’s global headcount, and it is a development Portland‑area tech workers have learned not to shrug off.

As reported by the Portland Business Journal, the new reductions will focus on DCG operations as Intel searches for additional savings. The outlet notes the company cut roughly 40,000 jobs globally in 2024 and 2025. Intel told the paper the moves are part of a “broader strategy to become a more focused and efficient company.”

Portland’s exposure​

Intel remains one of the Portland metro’s biggest private employers, and The Oregonian has reported the company employs roughly 20,000 people across its Washington County campuses. Earlier layoff rounds in 2024–25 triggered WARN notices and sent shock waves through local contractors and service firms, a reminder that any fresh cuts would spill well beyond Intel’s own badge‑wearing workforce. Local workforce and economic‑development officials are likely to keep a close eye on new filings and any outreach from the company in the days ahead.

Why the cuts are happening​

Company materials and investor presentations this year have hammered on tighter spending and simpler organizational charts as Intel reorients around higher‑margin products and AI‑related chips. In its recent investor communications, including Intel’s Q4 2025 earnings materials, the chipmaker has signaled a narrower set of priorities and ongoing workforce reductions. That backdrop makes it clear the new DCG‑focused cuts fit into a broader corporate reset rather than a one‑off belt‑tightening move.

What workers can expect​

In past rounds, Intel has issued individual notices to affected employees and provided severance packages. The Oregonian reported last year that laid‑off workers received at least 13 weeks of pay, plus additional weeks based on years of service. State WARN filings and local job centers typically list resources for displaced workers, and city and county officials often coordinate with major employers to organize rapid re‑employment support.

For employees who may be directly in the crosshairs of the new cuts, the immediate to‑do list is straightforward if not exactly welcome: watch for formal notices from Intel, review any severance details closely, and reach out to HR or local workforce agencies to understand what help is available.

https://hoodline.com/2026/07/intel-swings-layoff-ax-at-portland-data-center-powerhouse/

 

[NY_Sam2]

 

[Daniel Nenni]

 

[hist78]

View attachment 4909

Intel is gearing up for another round of layoffs targeting its Data Center Group as the chip giant keeps tightening the screws on costs and chasing efficiency. It is the latest move in a multi‑year wave of cuts that has already thinned the company’s global headcount, and it is a development Portland‑area tech workers have learned not to shrug off.

As reported by the Portland Business Journal, the new reductions will focus on DCG operations as Intel searches for additional savings. The outlet notes the company cut roughly 40,000 jobs globally in 2024 and 2025. Intel told the paper the moves are part of a “broader strategy to become a more focused and efficient company.”

Portland’s exposure​

Intel remains one of the Portland metro’s biggest private employers, and The Oregonian has reported the company employs roughly 20,000 people across its Washington County campuses. Earlier layoff rounds in 2024–25 triggered WARN notices and sent shock waves through local contractors and service firms, a reminder that any fresh cuts would spill well beyond Intel’s own badge‑wearing workforce. Local workforce and economic‑development officials are likely to keep a close eye on new filings and any outreach from the company in the days ahead.

Why the cuts are happening​

Company materials and investor presentations this year have hammered on tighter spending and simpler organizational charts as Intel reorients around higher‑margin products and AI‑related chips. In its recent investor communications, including Intel’s Q4 2025 earnings materials, the chipmaker has signaled a narrower set of priorities and ongoing workforce reductions. That backdrop makes it clear the new DCG‑focused cuts fit into a broader corporate reset rather than a one‑off belt‑tightening move.

What workers can expect​

In past rounds, Intel has issued individual notices to affected employees and provided severance packages. The Oregonian reported last year that laid‑off workers received at least 13 weeks of pay, plus additional weeks based on years of service. State WARN filings and local job centers typically list resources for displaced workers, and city and county officials often coordinate with major employers to organize rapid re‑employment support.

For employees who may be directly in the crosshairs of the new cuts, the immediate to‑do list is straightforward if not exactly welcome: watch for formal notices from Intel, review any severance details closely, and reach out to HR or local workforce agencies to understand what help is available.

https://hoodline.com/2026/07/intel-swings-layoff-ax-at-portland-data-center-powerhouse/

Do we know the headcount of Intel’s Data Center Group?

 

[tim_b]

View attachment 4908

Do we know the headcount of Intel’s Data Center Group?

Since Pat took office, DCG's revenue has declined significantly, with the business steadily ceding ground to AMD and NVIDIA across key markets.

 

[siliconbruh999]

Since Pat took office, DCG's revenue has declined significantly, with the business steadily ceding ground to AMD and NVIDIA across key markets.

Not entirely his fault that there was already broken stuff GNR is the first Good Xeon product in a while that got fixed and DMR as much LBT is saying it's bad and in terms of performance improvement it's very decent and is sellable.

On Layoffs i hope they use their brain and don't ayoff from the teams working on Core Product and don't do a Massive one cause delay is inevitable than.

 

[Xebec]

Since Pat took office, DCG's revenue has declined significantly, with the business steadily ceding ground to AMD and NVIDIA across key markets.

True, but "the ceding" started earlier -- 2017 or 2018. By 2020, Intel was more than one full generation behind in servers vs AMD.

(AMD Rome and Milan were 7nm chiplet based server architectures - launched in 2019 and 2021, while Intel was fielding 14nm server CPUs during this time).

 

[dkr1986]

Not entirely his fault that there was already broken stuff GNR is the first Good Xeon product in a while that got fixed and DMR as much LBT is saying it's bad and in terms of performance improvement it's very decent and is sellable.

On Layoffs i hope they use their brain and don't ayoff from the teams working on Core Product and don't do a Massive one cause delay is inevitable than.

Clearly, the delay in NVL, DMR partial cancellation/delays indicate they didn't use their brain on the past layoffs. Also, no Crescent Island yet or any other AI DC GPUs. Maybe it's their GPU guys who are getting the axe inside DCAI as Intel seemed to have partnered with Sambanova for their DC AI needs.

 

[swka]

Clearly, the delay in NVL, DMR partial cancellation/delays indicate they didn't use their brain on the past layoffs. Also, no Crescent Island yet or any other AI DC GPUs. Maybe it's their GPU guys who are getting the axe inside DCAI as Intel seemed to have partnered with Sambanova for their DC AI needs.

when any company gets to that scale, and some "company-wide" active is to be taken, no matter how reasonable CEO is, it will be done in a mechanical way. So yes, it will be done "without using their brain"

 

[dkr1986]

when any company gets to that scale, and some "company-wide" active is to be taken, no matter how reasonable CEO is, it will be done in a mechanical way. So yes, it will be done "without using their brain"

Which I have repeatedly said will happen when "targeted" layoffs were announced last year but some here were giddy at the thought of "right" sizing by the new CEO. They will be okay long term, but it's just another thing to deal with when they are already trailing on product competitiveness. Missing product launch window can be deadly in DCAI when the competitor wins the socket. These customers are inelastic and winning that socket in the next refresh cycle is difficult. This customer inelasticity is what kept Intel from dying in the enterprise segment. That is about to change in my opinion. Intel is just lucky the Agentic AI tailwind led DC CPU shortage came to the rescue. Based on all things I see, AMD Venice EPYC CPUs would absolutely dominate Granite Rapids and will launch at least 6-9 months ahead of DMR 16ch. But Intel can tread water comfortably due to CPU shortages even with 0.5-0.6x performance vs AMD for a while.

AMD has invested and built a go to market team for win Enterprise customers in the last year. With Intel cancelling DMR 8ch and EPYC Venice eating Granite Rapids Xeons for lunch, they will lose lot of sockets to AMD. Maybe the memory price inflation lessens the blow as enterprise looks to delay their server refreshes.

To make things worse, CEO himself declared their next gen Xeon - Diamond Rapids as not competitive product outright in ER calls non-stop in the name being "Honest" & "Transparent". There were lot of reports that this caused concerns with customers (Check ServeTheHome reporting on this). He could have said that DMR a very good product but still needs work to be a leadership product (you know like they talked about Granite Rapids- They did change the messaging a bit in last couple of ER calls). There is a difference between managing expectations and self-sabotaging.

Come to think of it, not even a single product has launched on time since 2025, and I blame the layoffs for the sake of "right sizing" the company to save ~$2B a year on Operating Expenses (Intel missed on their OpEx guidance and also missed their employee count target btw). Without the demand for the DC CPUs now due to AI, they would have lost billions for years because of this disruption.

 

[siliconbruh999]

Clearly, the delay in NVL, DMR partial cancellation/delays indicate they didn't use their brain on the past layoffs. Also, no Crescent Island yet or any other AI DC GPUs. Maybe it's their GPU guys who are getting the axe inside DCAI as Intel seemed to have partnered with Sambanova for their DC AI needs.

Crescent Island launch would likely be end of the year.

AMD has invested and built a go to market team for win Enterprise customers in the last year. With Intel cancelling DMR 8ch and EPYC Venice eating Granite Rapids Xeons for lunch, they will lose lot of sockets to AMD. Maybe the memory price inflation lessens the blow as enterprise looks to delay their server refreshes.

To make things worse, CEO himself declared their next gen Xeon - Diamond Rapids as not competitive product outright in ER calls non-stop in the name being "Honest" & "Transparent". There were lot of reports that this caused concerns with customers (Check ServeTheHome reporting on this). He could have said that DMR a very good product but still needs work to be a leadership product (you know like they talked about Granite Rapids- They did change the messaging a bit in last couple of ER calls). There is a difference between managing expectations and self-sabotaging.

Agreed here it could have been framed differently in Public and than the CEO goes and simply throws the product that he wants to sell under bus.

 

[Xebec]

To make things worse, CEO himself declared their next gen Xeon - Diamond Rapids as not competitive product outright in ER calls non-stop in the name being "Honest" & "Transparent". There were lot of reports that this caused concerns with customers (Check ServeTheHome reporting on this). He could have said that DMR a very good product but still needs work to be a leadership product (you know like they talked about Granite Rapids- They did change the messaging a bit in last couple of ER calls). There is a difference between managing expectations and self-sabotaging.

This seems to be a pattern with LBT. "We may stop development on 14A without customers" clearly signals that any consideration of using 14A is very high risk.

LBT has done a good job pumping the stock short term, and I think he's doing a better job than any Intel CEO before in terms of trying to win foundry customers, but it's not clear how well products are going to perform under his watch.

 

[dkr1986]

This seems to be a pattern with LBT. "We may stop development on 14A without customers" clearly signals that any consideration of using 14A is very high risk.

LBT has done a good job pumping the stock short term, and I think he's doing a better job than any Intel CEO before in terms of trying to win foundry customers, but it's not clear how well products are going to perform under his watch.

With the kind of connections LBT commands in the industry, I believe he is the perfect guy to lead Intel at this point of time to win external customers for Intel Foundry. Many companies who started as a startup is cozy with LBT because he was a VC and he does not have the baggage PG likely had while his time in Intel during the glory days.

But his 14A cancellation ruse (TBH I don't think it was ruse, I think he truly was prepared to cancel it) was a shitshow! who in their right mind would consider a foundry or process node that can be cancelled anytime if Intel cannot make economics work. The economics works only if a big whale (like Apple or Qualcomm or Nvidia or AMD) in addition to Intel Products decide to use that node in high volume (at least in the beginning). So, unless you hear news like that, low volume players developing hyperscalers in-house silicon (supposedly the best cohort to adopt Intel Foundry initially likes of Broadcom/ Marvell etc) will never touch that node seriously. What if down the line Intel comes and say, sorry economics don't work because Intel Products + your volume does not make sense. An entry foundry player needs to take a leap of faith before the customers takes a leap of faith (otherwise called Built it & they may come but if you don't build it, they will never come). Anyway, now LBT says stuff like Foundry customers don't sign up for one node, they sign up for a roadmap... I laughed so hard hearing him say that in the recent interview.

Couple of other things he walked back after announcing it as his pivotal strategy- 1) Selling non-core business, i.e. Selling Networking to not selling networking because you know, realizing networking is important for AI workloads, Nvidia is making shit ton of money selling networking products. 2) Prudent investments - sure cut the investment in capacity expansion, capex cut to dangerously low level, prepare to pay the fine in the billions for missing construction milestone to SCIP partner, miss the Server CPU demand for couple of quarters, sell the scraps in the inventory to makeup, realize the fines payable to SCIPS are almost as high as buying back the stake with a premium, now walk back and say we are investing more in capacity expansion.

It is clear that Intel BoD & LBT decided to reset market/ stockholder expectations by shitting on "Intel under previous management" as much as they can before unveiling "magical improvements" in business operations in 12 months (I met 5 yr goals set by BoD in 15 months = oh wow so humble (bragging)...He brags a lot all the while advocating humility, he says "I" a lot too). Anyone with an ounce of knowledge on working for or following Semiconductor Design & Manufacturing companies know these ships turn slow, products take years to influence etc. But it clearly helped with market sentiment. But hey! I am not complaining I made shit ton of money as an investor in Intel, but I am not blind to not see what is actually happening. What is surprising to me was the "Industry analysts" gobbling this shit up same as Wall St Analysts. I have actually grown wary of trusting "Industry Analysts" nowadays and put them in the same bucket as the "Wall St Sell Side Analysts".

Now that Nvidia is saying Single Core CPU performance is the most important thing for Agentic AI CPUs in Datacenter, I am waiting for LBT & Intel to repeat that line of talk about DMR lacking SMT focusing on Single threaded performance being ideal for AI workloads and walk back previous statements.

Having said all that, I am really glad it was LBT who succeeded PG. It could have been way worse like Hock Tan!

 

[blueone]

With the kind of connections LBT commands in the industry, I believe he is the perfect guy to lead Intel at this point of time to win external customers for Intel Foundry. Many companies who started as a startup is cozy with LBT because he was a VC and he does not have the baggage PG likely had while his time in Intel during the glory days.

LBT is still a VC. He is still a founding managing partner of Walden International, Celesta Capital, and Walden Catalyst Ventures.

But his 14A cancellation ruse (TBH I don't think it was ruse, I think he truly was prepared to cancel it) was a shitshow! who in their right mind would consider a foundry or process node that can be cancelled anytime if Intel cannot make economics work. The economics works only if a big whale (like Apple or Qualcomm or Nvidia or AMD) in addition to Intel Products decide to use that node in high volume (at least in the beginning). So, unless you hear news like that, low volume players developing hyperscalers in-house silicon (supposedly the best cohort to adopt Intel Foundry initially likes of Broadcom/ Marvell etc) will never touch that node seriously. What if down the line Intel comes and say, sorry economics don't work because Intel Products + your volume does not make sense. An entry foundry player needs to take a leap of faith before the customers takes a leap of faith (otherwise called Built it & they may come but if you don't build it, they will never come). Anyway, now LBT says stuff like Foundry customers don't sign up for one node, they sign up for a roadmap... I laughed so hard hearing him say that in the recent interview.

Agreed. LBT's statement about 14A's future was inexcusable for a senior executive of his experience and stature.

Couple of other things he walked back after announcing it as his pivotal strategy- 1) Selling non-core business, i.e. Selling Networking to not selling networking because you know, realizing networking is important for AI workloads, Nvidia is making shit ton of money selling networking products. 2) Prudent investments - sure cut the investment in capacity expansion, capex cut to dangerously low level, prepare to pay the fine in the billions for missing construction milestone to SCIP partner, miss the Server CPU demand for couple of quarters, sell the scraps in the inventory to makeup, realize the fines payable to SCIPS are almost as high as buying back the stake with a premium, now walk back and say we are investing more in capacity expansion.

Intel still has no credible networking strategy. Intel is still working with Google on the second generation of the IPU chip, but that's hardly a networking strategy.

It is clear that Intel BoD & LBT decided to reset market/ stockholder expectations by shitting on "Intel under previous management" as much as they can before unveiling "magical improvements" in business operations in 12 months (I met 5 yr goals set by BoD in 15 months = oh wow so humble (bragging)...He brags a lot all the while advocating humility, he says "I" a lot too). Anyone with an ounce of knowledge on working for or following Semiconductor Design & Manufacturing companies know these ships turn slow, products take years to influence etc. But it clearly helped with market sentiment. But hey! I am not complaining I made shit ton of money as an investor in Intel, but I am not blind to not see what is actually happening. What is surprising to me was the "Industry analysts" gobbling this shit up same as Wall St Analysts. I have actually grown wary of trusting "Industry Analysts" nowadays and put them in the same bucket as the "Wall St Sell Side Analysts".

I didn't invest in INTC. I've always been a buy and hold sort of guy, and Intel's financial fundamentals haven't supported anything like a $100/share price. Obviously, with hindsight, I wish I invested though. As for financial industry analysts regarding pretty much anything in the semi industry... don't get me started.

Now that Nvidia is saying Single Core CPU performance is the most important thing for Agentic AI CPUs in Datacenter, I am waiting for LBT & Intel to repeat that line of talk about DMR lacking SMT focusing on Single threaded performance being ideal for AI workloads and walk back previous statements.

Having said all that, I am really glad it was LBT who succeeded PG. It could have been way worse like Hock Tan!

Hock Tan has been a disaster for product customers (e.g. VMWare), but he's been great for shareholders. And the AVGO financial fundamentals support the stock's performance.

 

[MKWVentures]

Back to the layoff decisions:
1) LBT wants Intel to be smaller than 85K even when they return to success someday.
2) There is targetting of certain roles and certain areas. Other areas may increase. Laying people off and hiring new people at same time is likely as well
3) A number of people who "retired" or were "poached" .... were actually laid off (fired) and will be replaced.... LBT is REALLY remaking leadership.

So Intel headcount may go down overall and some of those people/roles will be replaced by new people in new roles.

just an opinion.

 

[MKWVentures]

Agreed. LBT's statement about 14A's future was inexcusable for a senior executive of his experience and stature.

I think the 14A comments were honest and lined up with what needed to be looked at. When he made the comment....
- Intel had no customers on 18A
- Intel had no customers on 14A
- Intel balance sheet was high risk
-IFS was losing 10B per year.

When GF stopped doing advanced nodes.... it was the correct decision and it hurt their future business. Both were true.

With the shortage today, it make IFS look very necessary. If the shortage didnt exist, or if it goes away, Intel would be looking at the "What to do with IFS?" again until it is a financial successful entity.

 

[blueone]

I think the 14A comments were honest and lined up with what needed to be looked at. When he made the comment....

If I was the CEO of a chip design company, being courted by Intel to bet one of my future product lines on 14A, which would probably cost me well into 9 digits ultimately, and LBT said what he said in that public statement, my next call would be to my legal and business people to sign the proposal from TSMC. I'd also create a rule in Outlook to send emails from anyone with an @intel.com address to my junk mail folder.

 

[MKWVentures]

If I was the CEO of a chip design company, being courted by Intel to bet one of my future product lines on 14A, which would probably cost me well into 9 digits ultimately, and LBT said what he said in that public statement, my next call would be to my legal and business people to sign the proposal from TSMC. I'd also create a rule in Outlook to send emails from anyone with an @intel.com address to my junk mail folder.

Agreed. Intel was looking at cancelling IFS completely at the time so that would have been fine.

This is the paradox with being a new foundry...... to build capacity, you need commits.... or you are not doing that node. to get commits, you need to promise capacity. Pat tried it one way (if you build it they will come) and it was a disaster (4 years no customers). LBT went the other way (commit and I will finiash the fab) and it seems like it might work.

the correct way is to have a foundry company and slowly add customers and capacity .... but intel doesnt have a foundry business yet.

And commits don't mean revenue or profits so we wont know the true outcome for 3+ years most likely. lets see what they say at earnings report

 

[dkr1986]

I think the 14A comments were honest and lined up with what needed to be looked at. When he made the comment....

No one questions that if Intel can't secure external customers to make Foundry at least break even, they have to discontinue the operation but announcing that in ER calls to satisfy investors was reckless imo. Adding that to the risk section in 10Q was basically Intel BoD & C-Suite covering their behinds from future Investor led lawsuits.

Now let's say 14A was discontinued, that would require them to keep operating the existing fabs (to keep supplying existing products) and outsource the new product wafers to TSMC at the bleeding edge to compete with AMD. We all know how this worked out in 2023 & 2024 for Intel. This scenario is little different - If 14A & future nodes are discontinued completely, that would cut lot of R&D spend, likely more layoffs in Foundry & low growth capex spend - still need to spend the maintenance capex to keep the light on the existing fabs - but what what they gain on opex savings will be offset by gross margin hit due to outsourcing & previous capex amortization schedule. Fully outsourcing the new products to TSMC while still running the older fabs (& newer 18A fabs) would mean Intel Products team's gross margin is impacted (no more low priced wafers from IDM, Intel needs to pay that juicy TSMC gross margin tax) due to outsourcing + remaining IFS operation perpetually posts negative gross margins as wafer volume shifts away to TSMC (it gets worse & worse when the depreciation schedule really starts kicking in for fabs & tools). Only way out of this is to spin off Intel Foundry and leave it to its own fate and then hope Intel products alone over time recovers Gross margin through competitive products (a big ? mark with AMD already gaining volume and cash balance to match Intel on nodes at TSMC - imo). They also need to stand in line to get wafer capacity allocation at TSMC.

Second, imo Intel Product engineers would struggle (at least initially) to deliver competitive products on TSMC nodes vs AMD (imo, they seem to have squandered the node density advantage of N3 in ARL (vs N4P for AMD) mainly because it is their first time using TSMC nodes for CPUs).

When GF stopped doing advanced nodes.... it was the correct decision and it hurt their future business. Both were true.

We all know Mubadala was there to bank roll Global Foundry when they got spun off from AMD. Intel Foundry would have needed someone like that to survive without the Capex $$ coming from Intel Products. CFO said Intel need about $9B Maintenance capex (I believe most of that is Foundry), who is going to shell out ~$9B every year to keep the lights on at Intel Foundry as a separate entity making negative gross margins with nodes that don't have industry standard PDKs to win other customers than Intel Products and need incremental capex to develop those PDK & IP tech to even sell those mature nodes to make money. Not to mention, your biggest customer is diversifying away from you to TSMC (Intel Products). That is a death spiral.

With the shortage today, it make IFS look very necessary. If the shortage didnt exist, or if it goes away, Intel would be looking at the "What to do with IFS?" again until it is a financial successful entity.

IFS itself was conceived on the tail of supply issues related to COVID. Also, main objective of IFS is to share the burden of developing advanced nodes under the IDM business model. Intel's IDM wafer volume is not enough anymore; they need to extend the life of these nodes to extract the initial investments. Breaking even on Operating income basis on 18A or 14A itself would have been successful outcome for Intel. Also, America First US government admin & real risk of China blockading Taiwan Strait are opportunities for Intel to pursue the Intel Foundry opportunity.

I didn't invest in INTC. I've always been a buy and hold sort of guy, and Intel's financial fundamentals haven't supported anything like a $100/share price. Obviously, with hindsight, I wish I invested though. As for financial industry analysts regarding pretty much anything in the semi industry... don't get me started.

I first bought my INTC position in Oct'22 after listening to Pat Gelsinger interview here, The next month ChatGPT hit the market and sucked up all the cloud capex $$ away from CPUs. Also followed is the Client PC downturn Armageddon in 2023. It has been a roller coaster waiting for 2026 for 18A to ramp, I got rewarded more than I thought with some stock pumping thanks to Trump but I have significantly reduced my position now. Still maintain a significant exposure but diversified quite a bit.

Pat Gelsinger came back to turn Intel around — here’s how it’s going

The Intel CEO is betting big on a new manufacturing site in Ohio.

www.theverge.com

You will hear the same Engineering culture change or Leadership restructuring corporate speak that we are hearing now. So this is not my first rodeo with Intel BoD & C-Suite "Culture Change". Only difference is LBT is lucky enough to be followed by someone who walked through fire to build up Intel's actual technology capability rather than sit on their own laurels & focus on returning shareholder value.

 

[dkr1986]

This is the paradox with being a new foundry...... to build capacity, you need commits.... or you are not doing that node. to get commits, you need to promise capacity. Pat tried it one way (if you build it they will come) and it was a disaster (4 years no customers). LBT went the other way (commit and I will finiash the fab) and it seems like it might work.

When you are sitting on fabs space built already & tools already ordered by someone, you can easily cut the future capex and say you saved the money and stopped the bleed in cash and say firmly "I don't believe in Build it, they will come". LBT can go the this other way because the other guy invested quite of bit of money on fab space (shell ahead), invested in R&D and get the tech ready. Imagine in 2021, LBT comes in as CEO of Intel announces IFS, goes to say Broadcom and say I will do the R&D, develop 18A, when you run test chips on 18A and commit, I will build Fab 52/62, buy the WFE tools and ramp the fabs. It will take 5+ years for that. I think Broadcom CEO will laugh at LBT and book some wafers at TSMC. All the while Intel Products team is sitting there waiting for the CEO to tell what node to build their next gen CPU on and what capacity to plan their road map of products.

Name one capital intensive business where a new entrant comes in and say, commit to buy my output and I will build my business/infrastructure and provide service to you. (while also road map of my other segment is also dependent on the same infrastructure).

the correct way is to have a foundry company and slowly add customers and capacity .... but intel doesnt have a foundry business yet.

Isn't this what they have been doing so far? Can you do this while your product team is also dependent on your roadmap on node process? You will have to match your biggest customers product release cadence, no? I will give you that they made some mistakes (required EDA & IP- CEO admitted as much) and things didn't go to exactly as they planned (under execution by their foundry team on test chips, revenue dried up on DCAI & CCG, stress on BS due to said reduction in revenue, CEO didn't manage expectations on Foundry business) . But it seems they need 2 fabs to support Intel Products on each node, each has to ramp one after the other and before that gen of products reach maturity, then need the next set of fabs to ramp the next gen of products. It went Intel 7 (Fab 28 & Fab 42) to Intel 4/3 (Fab 34 only because most of that gen got outsourced) to Intel 20A(not productionized)/18A (Fab 52 & 62) to 14A (Fab 27 in Ohio) (Since they lost so much market share to AMD, I think this aspect of their planning is also affected now) The one thing Intel can't do anymore like they used to do, under IDM 2.0 is rip out the tools of a fab and upgrade it to next gen node, rinse & repeat, because they need that older fab to extend the life of that node to generate incremental revenue from external customers to recoup their investment of developing the nodes.

If IFS started out as an independent new foundry with a rich cash balance, they can build only one fab, install 1 full line of tools required, ask customers to run test chips, ramp capacity only after customer commits & successfully demonstrating the capability and 1 fab worth of capacity. Even then they need to significantly invest before hand before customers commit. Once they prove themselves, then they can start building another fab and slowly ramp operations. This is not possible when Intel product team's road map of releasing a client & server CPU each year or two years, is tightly coupled with Intel Foundry developing and successively ramping your nodes one after the other. It is a very complex planning problem of being an IDM & a Foundry at the same time.