Intel CEO Pat Gelsinger doubles down on manufacturing, opens up foundry, shifts on strategy

Bob O'Donnell

Posts: 81   +1
Staff member
In context: The semiconductor industry has been getting a lot of attention at the macro and micro levels. From a big picture perspective, the shortages facing automakers and other industries have elevated chip-related stories to being a regular part of the evening news. On top of that, there’s been a great deal of focus on the geopolitical issues related to the supply chains, not only of the chip business, but of all the businesses that have become so dependent on a reliable semiconductor supply.

At a more detailed intra-industry level, there’s also been a great deal of scrutiny applied to the competitive landscape. In particular, there’s been a lot of focus recently on the ongoing challenges that long-time industry leader Intel has been facing in manufacturing and other areas.

Toss in the recent announcement of former Intel leader Pat Gelsinger coming back to the chip giant as CEO, and it seems the stage was nearly perfectly set for some blockbuster news on how the Gelsinger-led Intel could reinvigorate itself. Based on today’s announcements, the semiconductor stalwart did not disappoint.

In a significant strategic shift, Intel announced a series of major manufacturing related news that not only clarified any uncertainties about whether or not the company intended to keep building its own chips, but doubled down on its desire to increase its manufacturing capacity with a $20 billion investment to build two new state-of-the-art fabs focused on EUV-based (extreme ultraviolet) process technologies at 7nm and below in Arizona.

At the same time, Intel displayed a new willingness to work with other chip foundries on some of the company’s own chip designs. Finally, completing the trifecta, the company also unveiled plans to open up both its current and planned manufacturing capacity to other chipmakers through the launch of Intel Foundry Services.

Intel’s newest factory, Fab 42, became fully operational in 2020 on the company’s Ocotillo campus in Chandler, Arizona. Fab 42 produces microprocessors using the company’s 10nm manufacturing processes.

The strategy represents a bold new vision and surprisingly aggressive approach that is intended not only to help the company catch up to other major chip suppliers from a process perspective (e.g., reaching 7nm, 5nm and below), but also to work with them as needed for some of its own products until it reaches those goals, while starting to position itself as a competitive manufacturing partner to a much wider range of semiconductor companies.

"Intel designed and built its own chips—period. In fact, the company was the poster child example of an integrated device manufacturer, or IDM. With IDM 2.0, Intel is planning a significantly more open and comprehensive approach to chip manufacturing..."

For long-time Intel followers, this is an enormous change on many different levels. While the company has recently started working with several other chip foundry partners and has talked about building chips for a few outside companies, for decades it was essentially a closed loop.

Intel designed and built its own chips—period. In fact, the company was the poster child example of an integrated device manufacturer, or IDM. With IDM 2.0, Intel is planning a significantly more open and comprehensive approach to chip manufacturing that it hopes will once again return it to its leadership position in process technology and make it a significant worldwide competitor against TSMC, Samsung, GlobalFoundries, UMC and others.

Not only is the new Gelsinger-led Intel strategy tearing down the walls that kept Intel from working more aggressively with others, this strategy is opening up possibilities that were previously unheard of for the company, such as potentially licensing its x86 CPU cores to companies that use its manufacturing facilities.

Gelsinger discussed the possibility of doing things like using some of the company’s older, existing 22nm process based fabs to build chips for the automotive industry—something he said the company hadn’t really pursued until now

Based on conversations I participated in with Gelsinger to preview the news, it seems clear he’s open to almost any type of manufacturing and IP partnership possible, including things like using Intel’s chip packaging technologies to put together elements Gelsinger calls “tiles” (what others in the industry refer to as “chiplets”) built for other vendors at other chip foundries into SoCs (system on a chip) within Intel facilities.

In addition, Gelsinger discussed the possibility of doing things like using some of the company’s older, existing 22nm process based fabs to build chips for the automotive industry—something he said the company hadn’t really pursued until now. Along the way, he made the very interesting comment that Intel continues to use 86% of all the factories or fabs it ever built, though obviously some have been repurposed from their original intentions. For those who question whether semiconductor manufacturing, which progresses at such a blistering pace, can continue to be an effective strategy even for a company that has been doing it for decades, this should be a reassuring thought.

The world clearly needs more chip manufacturing capacity now and into future—something that the pandemic accelerated faster than anyone could have predicted.

Not only does the boldness and breadth of the strategy appear to be a much-needed shot in the arm for Intel, but the timing couldn’t be better.

The world clearly needs more chip manufacturing capacity now and into future—something that the pandemic accelerated faster than anyone could have predicted. In addition, the geopolitical issues driving renewed interest in US-based semiconductor manufacturing have hit a fever pitch—and with good reason.

As has been widely reported, the US share of global chipmaking has dropped from 37% in 1990 to just 12% today. Plus, over 60% of the world’s supply now comes from two major Taiwan-based foundries—TSMC and UMC—and that country’s growing tensions with China are raising red flags all over the world. In fact, not only did the US government recently pass legislation that earmarked several billion dollars in subsidies for US-based manufacturing, the EU has also recently voiced desires to dramatically increase the amount of chips built in Europe.

The timing of Intel’s announcements provides the company with an opportunity to take advantage of those potential subsidies, as well as competitive tax breaks and other incentives that other states in the US and countries in EU will likely provide as Intel continues with plans for additional manufacturing sites in the future.

As bold as the new Intel strategy may sound, it’s still critical that the company executes on that vision and regains some of the prestige and trust that it’s lost over the last few years. Catching up on process technology, in particular, is an extremely challenging task. Even though Intel also announced the tape-out of its first 7nm parts that are coming soon, Intel is going to need to provide a lot more definitive proof points that it can regain its long-held lead in manufacturing.

Gelsinger seems very confident that the company can do so, and successfully make the transition to simultaneously running a world-class independent foundry business (yet another big bet). If Intel can succeed on both fronts, it will undoubtedly be a very impressive story, but it’s a multi-year journey to get there with what will likely be a lot of unforeseen obstacles on the way. Without a doubt, it will be an interesting trip to watch.

Bob O’Donnell is the founder and chief analyst of TECHnalysis Research, LLC a technology consulting firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on Twitter .

Permalink to story.

 
It seems like intel is in deep denial after Apple broke up with them. If only they could ramp up their fabs quick enough to get ahead of the current capacity issues from TSMC but alas, This might at best give em a leg up on AMD. The looming threat at this point imo, it's still Apple.
 
This isn't the first time Intel has tried to open up their foundries to other chip designers (the previous attempt in 2015-2019 led to very little adoption) and the strategy will only work if 1) they can deliver a credible, competitive and reliable 7nm and beyond processes, and 2) they can let go of 'not invented here' syndrome and embrace industry standard tools and software so that designers can easily port across TSMC designs over to Intel.

Both elements stalled their previous efforts to become a credible foundry partner for external designers, and it will be interesting to see how Gelsinger (who is a dyed in the wool disciple of the 'Intel way' of approaching the industry, aka the opposite way to what they need to take) manages this transition, and brings the entire company culture along that ride.

The other thing will be how happy the likes of TSMC are to work with a competitor in the foundry business, assuming they pull that transition off. Modern designs are a very close collaboration between the designer (like AMD) and manufacturer (TSMC). It isn't as simple as sending a schematic to TSMC and saying 'make this', TSMC tooling and design libraries are integrated throughout the design process. I am sceptical that TSMC will be willing to extend this approach to Intel given the competitive nature of the two foundry businesses.
 
Last edited:
I applaud Intel for committing to invest in US fab capacity. In my view, offshoring our engineering and high-tech manufacturing is a massive national security risk. Having a viable domestic supply of semiconductors and electronics along with the domestic talent to run these operations is crucial.

This article also correctly points out the volatile relationship between Taiwan and China. Considering that China has long considered Taiwan to be a "renegade province", one has to wonder if an emboldened China would pursue economic or military actions against that state in the future... especially if the US sends ambiguous signals as to what consequences would occur. Having supply chain redundancy, if not supply chain independence, immensely strengthens our position and is the first (and possibly best) deterrent.
 
This isn't the first time Intel has tried to open up their foundries to other chip designers (the previous attempt in 2015-2019 led to very little adoption) and the strategy will only work if 1) they can deliver a credible, competitive and reliable 7nm and beyond processes, and 2) they can let go of 'not invented here' syndrome and embrace industry standard tools and software so that designers can easily port across TSMC designs over to Intel.

Both elements stalled their previous efforts to become a credible foundry partner for external designers, and it will be interesting to see how Gelsinger (who is a dyed in the wool disciple of the 'Intel way' of approaching the industry, aka the opposite way to what they need to take) manages this transition, and brings the entire company culture along that ride.

The other thing will be how happy the likes of TSMC are to work with a competitor in the foundry business, assuming they pull that transition off. Modern designs are a very close collaboration between the designer (like AMD) and manufacturer (TSMC). It isn't as simple as sending a schematic to TSMC and saying 'make this', TSMC tooling and design libraries are integrated throughout the design process. I am sceptical that TSMC will be willing to extend this approach to Intel given the competitive nature of the two foundry businesses.
If you haven’t already read it, Ben Thompson at Stratechary has had some interesting takes on Intel recently. Some might be subscriber only though. Some similar points to yours.
 
Licensing x86...

Interesting, but unless something drastic is done, I think that ARM/RISC will continue to be the future, if going by what apple pulled with the M1 and others are doing on the servers side.

At this point, ARM can only go up in power consumption and performance and intel failed in beating ARM on their own turf.
 
It seems like intel is in deep denial after Apple broke up with them. If only they could ramp up their fabs quick enough to get ahead of the current capacity issues from TSMC but alas, This might at best give em a leg up on AMD. The looming threat at this point imo, it's still Apple.

What?
 
I applaud Intel for committing to invest in US fab capacity. In my view, offshoring our engineering and high-tech manufacturing is a massive national security risk. Having a viable domestic supply of semiconductors and electronics along with the domestic talent to run these operations is crucial.

This article also correctly points out the volatile relationship between Taiwan and China. Considering that China has long considered Taiwan to be a "renegade province", one has to wonder if an emboldened China would pursue economic or military actions against that state in the future... especially if the US sends ambiguous signals as to what consequences would occur. Having supply chain redundancy, if not supply chain independence, immensely strengthens our position and is the first (and possibly best) deterrent.

It's worth noting that foundry jobs didn't so much move overseas as US based manufactures simply couldn't compete in regards to the price to build/maintain them versus the sales they could get out of them. That's what billions in government subsidies does.

While I agree that foundries being overseas is a MASSIVE security risk going forward, there really isn't much the Federal government can do about it. About the only thing they can do is directly subsidies foundry operation, but then of course "Socialism" gets cried out.
 
This is one massive change between both CEOs. One was accountant, the other actually knows how to and understand that chip maker has to make chips and not produce stream of embarrassing Power Point slides only (like AMD did for a decade).

Why was Bob Swan appointed in first place I have no idea. For good or bad at least that something. Commitment that Intel will not just roll over and let others do stuff. Intel soundly deserves all the thrashings. Ten years ago they could push mankind computing capabilities to new highs and tech level, but of course they choose they usual way - insatiable thirst for greed. It serves them well that AMD ate (for now) 30% of server segment and basically whole consumer, it serves them well that Apple dumped them for more promising platform.

BTW anyone remembers what Intel HEDT is? Yeah I though so. Since debacle X299 was and joke W-3175X 1kW CPU AMD released 3 platforms (1st, 2nd TR + TR Pro) with 4th on he horizon in form of 5000 TR. Of course I'm not happy that AMD cornered the market of TRPro only for Lenovo for handful of countries. I had to get TRX40 for that very reason, while I could get WRX80. Now I don't care, waiting for 5000 TR, just seeing IPC increases in 3rd gen EPYC. Glorious. (I'm sounding like Wendell ;) )
 
This is great news. More fabs, more competition, cheaper products.
Also moving production away from a country that commits genocide on racial minorities, send in tanks on its student population and annexes other regions completely ignoring agreed handovers (Hong Kong) gets my vote. China should be an economic pariah and not considered for any manufacturing or trade.
 
Licensing x86...

Interesting, but unless something drastic is done, I think that ARM/RISC will continue to be the future, if going by what apple pulled with the M1 and others are doing on the servers side.

At this point, ARM can only go up in power consumption and performance and intel failed in beating ARM on their own turf.
2013 called and wants its predictions back. Maybe by the time ARM is a legitimate server competitor to intel we will finally have fusion reactors.

The real threat to intel is not LOLARM, it's AMD's EPYC CPUs, which dominate intel in most server tasks. Unless AMD just decides to sit back and do nothing, intel will only continue to fall behind in performance and hand AMD a very lucrative pie to sink their teeth into.
 
This isn't the first time Intel has tried to open up their foundries to other chip designers (the previous attempt in 2015-2019 led to very little adoption) and the strategy will only work if 1) they can deliver a credible, competitive and reliable 7nm and beyond processes, and 2) they can let go of 'not invented here' syndrome and embrace industry standard tools and software so that designers can easily port across TSMC designs over to Intel.
From the standpoint of having worked for such a company, that was impossible for the company I worked for to do. At least Intel has a leg up on them, as it sounds like this CEO is not averse to reversing previous decisions. Once the company I worked for made a decision, they had what seemed like a corporate policy to NOT reverse them under any circumstances.

So, did Intel get rid of the business, but technically clueless, wonk they had in there for a year or two? Or is this guy, even though he is returning as CEO, as equally clueless, technically?

Opening up their foundry seems like a move of desperation - as if acting as foundry for other manufacturers will make up for a lack of CPUs capable of competing with AMD in the Intel product pipeline.

Oh well. That's what happens to a company that decides their shareholders are more important than giving their loyal product users meaningful new generations of products. As I see it, Intel backed itself into a corner, laughed at AMD when AMD was not competitive with them, then Intel's hubris caught up with them. It is a corner of Intel's own making.
 
Eeerrrmm, what? Sometimes, I do wonder how people get to such conclusions...
Fair enough this one is probably not going to pay off in the short term but long term I think that the PC market will likely shift away from personal computing as we know it with Apple eventually gaining a lot more ground and market share for consumers and even some professional use.

This change isn't surprising and I'd argue it has been on-going for some time now: the majority of the world is interconnected through mobile devices operating on Android or iOS as it is but as time goes on I think the closed model of Apple will find a lot more success with this regular consumers now that they're more serious about seemingly integrating the mobile experience with the comfort of the PC and Laptop form factor.
 
Unless AMD just decides to sit back and do nothing, intel will only continue to fall behind in performance and hand AMD a very lucrative pie to sink their teeth into.
I am sure Lisa Su does not mind the income EPYC is generating for AMD, however, I doubt Lisa Su will have anything to do with such a strategy - especially if she understands that what happened to Intel could happen to AMD if it is allowed by her and/or corporate policy. I am sure she is aware that following a course of hubris will do nothing for AMD in the long run - at least, I hope she is.
 
I am sure Lisa Su does not mind the income EPYC is generating for AMD
Although one can't know for certain what revenue EPYC has generated for AMD, the sector it comes under (Enterprise, Embedded, and Semi-Custom) saw a year-on-year growth of 176% in 2020. AMD attributes this to EPYC and semi-custom (I.e. console APU) sales.

In terms of Q4 revenue, EESC accounted for 39%, whereas Computing and Graphics was 60%. For Intel, in their comparable sectors, DCG (Data Center Group) accounted for 31% and their CCG (Client Computing Group) was 55%.
So roughly similar, but EPYC and the latest consoles have clearly been good AMD. That said, Intel's Q4 revenue was $20b to AMD's $3.24b (it was $5b for Nvidia), so there's still some way to go yet.
 
This shift on strategy is an important milestone for Intel.The sad thing, is that all of the announcements are reactions to: 1) slow monolithic corporate culture 2) lost product attractiveness.

That Intel lost the mobile train, is no news. What is much more worrisome, is how it managed to loose leadership in it's main product, all while watching AMD restructure years ago. Did they really think that the giant corporation will continue to rule in monolithic form like decades past? All the shaken up industries didn't tell anything to Intel? Where did all the R&D money go to? Certainly it didn't reach it's main products.

Now, it's doing the only moves they can. And throwing the towel. New plants in american soil are a great move, (more in you count the govmt. helping hand) but, hey, the capacity will be superior to what they can sell, so we will manufacture others products. They have to concentrate first on creating products that the market demands, with superior characteristics, Anything less, and Intel in condemned to be relegated to follower and manufacture for others. If that's the plan, other will fill the innovation void.

If Intel disapeared tomorrow, would we loose the best cpu, mobile or desktop? That what they have to do again, they have resources, but do they continue blinded toward the iceberg? As I see it, all the announcements go in the direction on adapting to a new reality, an Intel Corp standing squeezed by competitors and not much room to maneuver.
 
Fair enough this one is probably not going to pay off in the short term but long term I think that the PC market will likely shift away from personal computing as we know it with Apple eventually gaining a lot more ground and market share for consumers and even some professional use.

This change isn't surprising and I'd argue it has been on-going for some time now: the majority of the world is interconnected through mobile devices operating on Android or iOS as it is but as time goes on I think the closed model of Apple will find a lot more success with this regular consumers now that they're more serious about seemingly integrating the mobile experience with the comfort of the PC and Laptop form factor.
Wait a minute, you think, long term, everyone is just going to give up their PC's and move to Mac's?

Apple's closed eco-system is exactly why they won't break into the vast majority of businesses, Most businesses won't replace their entire infrustructure and re-code everything just to go with Apple.

Apple do not sell CPU's, Intel have absolutely nothing to fear, did they lose a big customer? Yes? I mean, Apple don't sell a huge amount of Mac's so Big but not a very high quantity customer. Are Apple a threat to AMD, Intel, Qualcomm, MediaTek etc... No.
 
Wait a minute, you think, long term, everyone is just going to give up their PC's and move to Mac's?

As I mentioned on my post, this has been an on-going process already as statistically speaking, most people primarily use a phone or a tablet and a not-insignificant amount of people only have these devices and have no personal computer outside of work and such.

Apple's closed eco-system is exactly why they won't break into the vast majority of businesses, Most businesses won't replace their entire infrustructure and re-code everything just to go with Apple.

I can concede that my post was mostly referring to the consumer market not the enterprise one because there IS a lot of truth to what you are saying here: there's just so much resistance to move away from "It works so good enough" mentality that keeps most enterprise stuff on Windows based that it might take a lot more time.

However...There are some solutions and we have kind of accelerated into those now with the pandemic. Remote work it's not just using VPNs there's also a good amount of Windows-only work that can be done simply by having a Windows VM to connect to. I have this exact set up for example since well I have to be remote and on a VPN anyway so they just cloned a VM that has all of the software environment I need done. Company did provided a windows rig to connect but it would work just as well on anything that accepts a remote desktop client.

What I am saying is that if for whatever reason the company decided "Hey we're giving you an M1 air as your work laptop" tomorrow basically nothing would change for me: the M1 works ok for basic office stuff like email and connecting to Teams and Zoom and running a remote desktop client to connect to my VM also works fine: my windows stuff is cheaper on a VM on a server.

Apple do not sell CPU's, Intel have absolutely nothing to fear, did they lose a big customer? Yes? I mean, Apple don't sell a huge amount of Mac's so Big but not a very high quantity customer. Are Apple a threat to AMD, Intel, Qualcomm, MediaTek etc... No.

Not having an important name like Apple supporting x86_64 is important imo. It's not just that they lost a big customer (Big enough to hurt their bottom line imo) but that they did so because they think they can compete with x86.

And just yes I realize Apple has tried this exact thing before in the past with power pc. But now that they're many, many times bigger as a company plus many of the other circumstances I've been mentioning I think they have much better chances at succeeding.
 
Not having an important name like Apple supporting x86_64 is important imo. It's not just that they lost a big customer (Big enough to hurt their bottom line imo) but that they did so because they think they can compete with x86.
I am not a rabid cult member of the Apple church, but their chip is simply amazing and as a trend setter, this is opening the doors even more to ARM and RISC-V.

X86 design is simply too complex, hence needs more resources, like bigger die, more power, etc.

Hence why I said, X86 itself needs to evolve in a drastic way.

If not, ARM will eventually take over.
 
Intel lost its leadership, is pretty much f-ed; I sold all my intel stocks a few months ago. Its brand image is not associated with their own team members like AMD or NVIDIA;the only intel person I could think of was the USB inventor.
 
I am not a rabid cult member of the Apple church, but their chip is simply amazing and as a trend setter, this is opening the doors even more to ARM and RISC-V.

X86 design is simply too complex, hence needs more resources, like bigger die, more power, etc.

Hence why I said, X86 itself needs to evolve in a drastic way.

If not, ARM will eventually take over.

M1 is good for some workloads, partly because it has huge L1 level caches, L1 instruction cache is whopping 192 KB. But. Is bigger cache always better? With Zen2 AMD reduced L1 instruction cache from 64kB to 32kB because according to AMD, internal testing showed 32kB cache with double associativity was overall (but not always) faster choice. Also worth to note that current AMD and Intel x86 CPU's are NOT designed single thread but single core performance in mind. M1 is nothing sort of "amazing", it's just suited for different needs.

x86 design needs More resources for What? Now let's see. Current x86 CPU's are basically RISC CPU's internally. About only thing that makes them "too complex" or "needing more resources", are decoders that translate x86 instructions into micro-ops and other way around. That takes very little amount of resources. And that's basically everything that makes x86 "too complex" or "resource hog" vs ARM CPU's.

Have been hearing this "x86 is dead" mantra at least 30 years already.
 
M1 is good for some workloads, partly because it has huge L1 level caches, L1 instruction cache is whopping 192 KB. But. Is bigger cache always better? With Zen2 AMD reduced L1 instruction cache from 64kB to 32kB because according to AMD, internal testing showed 32kB cache with double associativity was overall (but not always) faster choice. Also worth to note that current AMD and Intel x86 CPU's are NOT designed single thread but single core performance in mind. M1 is nothing sort of "amazing", it's just suited for different needs.

x86 design needs More resources for What? Now let's see. Current x86 CPU's are basically RISC CPU's internally. About only thing that makes them "too complex" or "needing more resources", are decoders that translate x86 instructions into micro-ops and other way around. That takes very little amount of resources. And that's basically everything that makes x86 "too complex" or "resource hog" vs ARM CPU's.

Have been hearing this "x86 is dead" mantra at least 30 years already.
Many good points.

I'm just looking at the current performance in single core combined with power consumption, which right now, its very good on M1 compared to AMD and Intel.

And what I meant with resources, you cannot deny that as of right now, a M1 is as fast as an equivalent X86 meanwhile it consumes a lot lets power.

I'm not claiming X86 is dead, but we again, cannot ignore the potential that M1/ARM are currently showing.
 
Back