Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/how-intel-plans-to-rival-tsmc-and-samsung-as-a-chip-supplier.17048/
        )

    [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] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

How Intel plans to rival TSMC and Samsung as a chip supplier

Daniel Nenni

Admin
Staff member
Geopolitics could give U.S. tech giant's contract chipmaking strategy a boost

PALO ALTO, U.S./TAIPEI -- Since returning to Intel as CEO in early 2021, Pat Gelsinger has been on a mission: transform America's biggest semiconductor company into a major contract chipmaker.

Pulling off such a transformation would reshape the global chipmaking landscape at a time of unprecedented geopolitical tensions over control of this crucial technology. The U.S. is pouring billions into its domestic chip sector and clamping down on China's own ambitions, while Beijing is desperately trying to break its reliance on foreign technology.

But manufacturing chips for outside customers -- also known as the foundry business -- is new territory for Intel, which has built its reputation designing and making its own cutting-edge semiconductors for PCs and servers. Asian rivals, moreover, have dominated the global foundry market for decades.

Foraying into contract chipmaking also puts Intel in direct competition with two of its own suppliers, Taiwan Semiconductor Manufacturing Co. and Samsung Electronics.

And it is an expensive strategy. Since Gelsinger, who previously worked at Intel from 1979 to 2009, announced the pivot in March 2021, the company has planned spending of over $70 billion for building and expanding its chip fabrication facilities, or fabs.

https%253A%252F%252Fs3-ap-northeast-1.amazonaws.com%252Fpsh-ex-ftnikkei-3937bb4%252Fimages%252F_aliases%252Farticleimage%252F4%252F8%252F4%252F6%252F42896484-1-eng-GB%252FIntel-D1X-Oregon-2re.jpg
Intel is attempting a pivot into the foundry business, or making chips for outside customers, an area long dominated by Asian rivals. © Intel

None of that has stopped Intel from setting its sights high.

"Our ambition is to be the No. 2 foundry in the world by the end of the decade, and [we] expect to generate leading foundry margins," Randhir Thakur, the president of Intel Foundry Services, told Nikkei Asia. IFS was set up last year to turn Gelsinger's vision into a reality.

For Intel, such a move is not only a new potential revenue source, but also a way to regain a technological edge in chip manufacturing lost to Asia over the past decades.

Investors, however, do not seem entirely convinced: Intel's share price has more than halved since it embarked on its foundry transformation.

The company's spending in this area includes $20 billion for a chip facility in Ohio and 17 billion euros ($16.8 billion) to build a plant in Germany, as well as $3.5 billion to expand its chip packaging facility in New Mexico, a $20 billion investment in Arizona fabs and a 17 billion euro expansion in Ireland. On top of that, Intel acquired Israeli foundry Tower Semiconductor for $5.4 billion in February.

Meanwhile, slowing global demand for chips has weighed on Intel's top line. The company reported a 20% year-over-year drop in third-quarter revenue last week, and lowered its 2022 full-year revenue outlook to between $63 billion and $64 billion, down as much as $4 billion from its previous guidance.

Coupled with the heavy spending on its foundry business, Intel is now expecting to end 2022 with a negative $2 billion to $4 billion free cash flow, compared to the negative $1 billion to $2 billion it projected earlier this year.

https%253A%252F%252Fs3-ap-northeast-1.amazonaws.com%252Fpsh-ex-ftnikkei-3937bb4%252Fimages%252F_aliases%252Farticleimage%252F9%252F7%252F7%252F2%252F42892779-1-eng-GB%252F20221104-Intel-revenue-Col.png


While investors may have doubts, however, analysts and industry insiders say Intel's foundry strategy is "sensible" and massive upfront investments are simply the price the company must pay to make it work.

"For an IDM [integrated device manufacturer] like Intel to really survive, there are only a couple of things that they can do. They can either grow and get scale, or they specialize," said Wayne Lam, senior director of research at CCS Insight, an industry analysis company.

David Crawford, head of the global technology and cloud services practice at consultancy Bain & Co., agreed.

"I think it's a very sensible strategy for Intel to scale into the fabrication business, yet it is not for the faint of heart," Crawford said. "You're either in or you're out. The worst strategy for someone like Intel is to be middling."

But to scale up, Intel must win over customers, something it has so far struggled to do.

Intel has previously said that Qualcomm, Amazon's AWS and MediaTek have signed up to use its manufacturing services. It did not announce any new foundry customers for the July-September quarter.

During the same period, IFS revenue dropped 2% on the year to $171 million, following a 54% year-over-year drop in the previous quarter.

One reason Intel has struggled to win over customers is the formidable competition from the market leaders.

TSMC -- the world's largest contract chipmaker -- controlled over 53% of the global foundry market in the first half of this year, research company Trendforce said. Samsung was in second place, with a 16.5% share. TSMC only makes chips for others and counted most of the world's top chip developers -- from Apple, Qualcomm and Nvidia to Broadcom -- as key customers

https%253A%252F%252Fs3-ap-northeast-1.amazonaws.com%252Fpsh-ex-ftnikkei-3937bb4%252Fimages%252F_aliases%252Farticleimage%252F8%252F1%252F8%252F9%252F42859818-3-eng-GB%252F20221104-Biz-Spot-Foundry-market-share-Bar-stack.png


Samsung mainly builds chips for its in-house use but has become more serious about its foundry business over the years. In 2019 it announced it would spend 133 trillion won ($115 billion) by 2030 to expand its chip design unit and foundry business.

Intel lags behind both of these companies in semiconductor manufacturing technology as gauged by how many transistors they can squeeze onto a single chip. This is measured in nanometers -- a smaller number means less space between transistors and, generally speaking, indicates a more advanced chip.

TSMC and Samsung both began production of industry-leading 3-nanometer chips this year and aim to put 2-nm chips into production by 2025. Intel has still not been able to mass-produce 5-nm chips, which are widely used in advanced electronics like smartphones.

For its foundry business to succeed, Intel will have to do more than catch up on the tech front, according to Charles Shi, a semiconductor analyst with Needham & Co. After decades of making chips primarily for its own use, the company must learn how to serve a wide range of external customers with varying needs, he said. Intel also needs to build up a third-party intellectual property portfolio, design services, and a chip packaging and testing ecosystem with partners to make it easier for customers to use Intel's manufacturing process.

"Intel has taken initiatives to address these changes, but so far, there is no evidence to show any of these hurdles have been overcome," Shi said.

https%253A%252F%252Fs3-ap-northeast-1.amazonaws.com%252Fpsh-ex-ftnikkei-3937bb4%252Fimages%252F_aliases%252Farticleimage%252F2%252F8%252F3%252F6%252F42896382-1-eng-GB%252FSamsung_foundrys_first_3nm_chip_production_3re.jpg


Samsung of South Korea has begun mass-producing chips using the industry-leading 3-nanometer production technology, something only Taiwanese rival TSMC has also achieved so far. © Samsung

Intel is still working on getting its 5-nm mass production going, and says it will begin manufacturing Intel 3 -- its answer to TSMC's 3-nm tech -- in the second half of 2023. Intel 18A production, intended to compete with TSMC's 2-nm chips, is slated to start in the second half of the following year.

Any major setback to those plans would mean more pressure on the company's cash flow and from investors.

"If they run into another incident where they fall back two generations [in chip production technology], they are done for," said Lam at CCS Insight.

Amid all the challenges, one external factor could give Intel's strategy a boost: geopolitics.

"If you asked me five years ago whether Intel could get some business in the foundry space, there is no doubt I would have said no," a veteran Taiwanese chip executive told Nikkei Asia, declining to be named because his company has business with both TSMC and Intel. "But if you asked me now, I think Intel does have some opportunity because of Washington's rhetoric of diversification amid mounting geopolitical pressure in Asia, especially between Taiwan and China."

The U.S. government has been ramping up efforts to bring chip production onshore amid tensions with China and a global chip shortage that emerged amid the COVID-19 pandemic.

The question now, the Taiwanese executive said, is how well Intel can monetize those concerns and grow its foundry business.

https%253A%252F%252Fs3-ap-northeast-1.amazonaws.com%252Fpsh-ex-ftnikkei-3937bb4%252Fimages%252F_aliases%252Farticleimage%252F3%252F3%252F4%252F6%252F42896433-1-eng-GB%252FAP22253052693866re.jpg
Intel CEO Pat Gelsinger, left, and U.S. President Joe Biden attend the groundbreaking ceremony of the chipmaker's manufacturing facility in New Albany, Ohio, on Sept. 9. © AP

Intel itself is optimistic it can seize this opportunity.

"As we have engaged with foundry customers since launching IFS, it has become abundantly clear that many of these companies see the need for a more resilient and geographically balanced semiconductor supply chain," IFS President Thakur told Nikkei Asia.

Intel was a key advocate for the passage of the CHIPS Act, which will funnel more than $50 billion in federal money into the semiconductor industry. State money has also supported Intel's foundry venture. The company has received grants and tax breaks from Ohio, New Mexico and Arizona.

But TSMC and Samsung are also expanding their production footprints in the U.S., and they have long relationships and proven track records with their clients.

Still, shifting political tides could give Intel the chance it has been waiting for, said Needham & Co.'s Shi.

"The opportunity is always there for Intel, as the geopolitical risk is rising not only for TSMC but also for Samsung," Shi said. "We may be heading toward some fundamental changes to the geopolitical landscape in Asia over the next decade."

 
From my naive investor's view, "Why bother?" It's hard to (impossible for me) to gauge all the risks and rewards of all of Intel's businesses, whether foundry, data center, PC, or automative. It's tough to see five years, without lots of guesses.

Best quote from the article:

"I think it's a very sensible strategy for Intel to scale into the fabrication business, yet it is not for the faint of heart," Crawford said. "You're either in or you're out. The worst strategy for someone like Intel is to be middling."

Just looking at foundry, Crawford's "middling' is the most likely. Key challenges: 1) Hiring and retaining the best talent at all levels, especially now that TSMC is doubling down in Arizona. Without the best talent, yield drops, taking down profitability and ROI. 2) Customers will move slowly before they commit large volume deals. TSMC's and Samsung's high volume customers will ramp quickly to new processes. That will force Intel, due to internal and US government political pressure, to focus on niche deals. When Intel turns on a $10B fab, they must have some customers. Of course Intel foundry will sell to Intel, but that's not the goal. 3) Intel missed mobile, except for automotive. (I have no idea how that market will shake out, but units shipped will remain a fraction of mobile.)
 
From my naive investor's view, "Why bother?" It's hard to (impossible for me) to gauge all the risks and rewards of all of Intel's businesses, whether foundry, data center, PC, or automative. It's tough to see five years, without lots of guesses.

Best quote from the article:

"I think it's a very sensible strategy for Intel to scale into the fabrication business, yet it is not for the faint of heart," Crawford said. "You're either in or you're out. The worst strategy for someone like Intel is to be middling."

Just looking at foundry, Crawford's "middling' is the most likely. Key challenges: 1) Hiring and retaining the best talent at all levels, especially now that TSMC is doubling down in Arizona. Without the best talent, yield drops, taking down profitability and ROI. 2) Customers will move slowly before they commit large volume deals. TSMC's and Samsung's high volume customers will ramp quickly to new processes. That will force Intel, due to internal and US government political pressure, to focus on niche deals. When Intel turns on a $10B fab, they must have some customers. Of course Intel foundry will sell to Intel, but that's not the goal. 3) Intel missed mobile, except for automotive. (I have no idea how that market will shake out, but units shipped will remain a fraction of mobile.)
Assuming Intel can hit it's targets with 20/18A, I would not be shocked if many of Samsung's customers jump ship. It is a not TSMC fab, has 100% of it's capacity in the west, and a PPA advantage over TSMC. If intel undershoots, there is a very good chance they at the very least have a lead over Samsung.

Intel needs IFS if they want to succeed. Without external customers they will not have the scale to continue scaling transistor sizes. IFS charging Intel will also drive the design side to not have a dozen stepings for every product, and reduce issues from constantly varying wafer throughput and fab priority. I also doubt that layoffs are going to impact IFS or it's roadmaps, given the criticality of it to the company's survival and how much non essential fat there is to trim.

From my naive investor's view, "Why bother?" It's hard to (impossible for me) to gauge all the risks and rewards of all of Intel's businesses, whether foundry, data center, PC, or automative. It's tough to see five years, without lots of guesses.
What is wrong with the risk reward of DCAI, AXG, and CCG? Intel agrees with you on automotive and sold of some of it's shares to get all of it's money back (and a little bit extra) as well as formalizing Mobileye's De facto autonomy.
 
Intel is not the same as TSMC or Samsung, because they compete with their own customers for foundry services.
They will never give external customers access to any new leading edge node, because they require 100% of that node for their own products. Foundry is only 1% of Intel's business, they get most of their profits from CPU products and those products will always come first.
 
Intel is not the same as TSMC or Samsung, because they compete with their own customers for foundry services.
They will never give external customers access to any new leading edge node, because they require 100% of that node for their own products. Foundry is only 1% of Intel's business, they get most of their profits from CPU products and those products will always come first.
This is not right. Intel’s first two foundry nodes — Intel 3 and Intel 18A — will be their absolute bleeding edge technology for the time.

And as for them using all their leading edge foundry supply for their own internal designs, that too cannot be right else IFS would not be a business…
 
Not really. IFS is business, but just for older nodes. For leading edge maybe they give a very small capacity to external customer, but it won't be until after the initial ramp up
 
I don’t think QCOM would be investigating leading edge products on 18A if IFS told them “you can get manufacturing supply but only after we have shipped all of our client/server/hpc products on this node”.

What you describe is a terrible way to throw away billions of dollars bootstrapping a new business and then never even remotely catch up to TSMC / Samsung Foundry.
 
I don’t think QCOM would be investigating leading edge products on 18A if IFS told them “you can get manufacturing supply but only after we have shipped all of our client/server/hpc products on this node”. What you describe is a terrible way to throw away billions of dollars bootstrapping a new business and then never even remotely catch up to TSMC / Samsung Foundry.

True, that's not really how it works. Intel will use 20A for internal CPU chiplets and 18A for IFS customers and internal use. So Intel will have a good year of 20A before 18A comes out. Also Intel is now using the chiplet (tiles) methodology, like AMD, so they won't have the big monolithic chips to jam up their fabs. Chiplets are very disruptive and will change the way we do business, absolutely.
 
when a new node ramps up the capacity is always constrained, so Intel don't even have enough capacity for their own products, let alone external customers. You can't just say "chiplets" and wave your hands. Only servers are currently using chiplets but they still are big enough (400mm2)
Plus look at the bottom line. Intel make $18 billion quarterly revenue from mostly CPU business and only $170 from foundry.
 
when a new node ramps up the capacity is always constrained, so Intel don't even have enough capacity for their own products, let alone external customers. You can't just say "chiplets" and wave your hands. Only servers are currently using chiplets but they still are big enough (400mm2)
Plus look at the bottom line. Intel make $18 billion quarterly revenue from mostly CPU business and only $170 from foundry.

Say chiplets three times and click your heels... :LOL: But anyway you are wrong starting at 18A. This is the new Intel not the old Intel. The new Intel will have too much capacity unless they cut CAPEX in half moving forward, which could certainly happen.
 
when a new node ramps up the capacity is always constrained, so Intel don't even have enough capacity for their own products, let alone external customers. You can't just say "chiplets" and wave your hands. Only servers are currently using chiplets but they still are big enough (400mm2)
Plus look at the bottom line. Intel make $18 billion quarterly revenue from mostly CPU business and only $170 from foundry.
I'm with Lefty on this one. I too think that Intel won't have the capacity for themselves, let alone external clients. I just don't see it.
You two have got to be kidding. Do you really think Intel would be negotiating with Apple, Qualcomm, and Amazon to fab chips in the 18A process if they thought there was little chance they would succeed? Intel may fail to get high volumes and high yields from the new processes, but it is irrational to think that the plan going in is to prioritize 18A volumes for themselves and allocate leftovers for the three previously mentioned companies, or perhaps just Intel 7 capacity.
 
I'm with Lefty on this one. I too think that Intel won't have the capacity for themselves, let alone external clients. I just don't see it.
Due to bad process yield or limited manufacturing capacity? Everything I have heard in the PDK ecosystem is good for 18A thus far. Fingers crossed. I hope there is a three way race for GAA foundry business.
 
Intel is not the same as TSMC or Samsung, because they compete with their own customers for foundry services.
Wrong. Samsung competes with most of their customers, and those customers get their leading edge nodes as soon as Samsung debgugs them. Also you forget that most of Samsung’s capacity is also for itself.

I too think that Intel won't have the capacity for themselves, let alone external clients. I just don't see it.
Also wrong. Right now intel is struggling from overproduction and likely their fabs are being underutilized. When products like meteorlake come out that will further reduce demand for Intel silicon. With this information (and the continued loss of market share to AMD Apple and maybe Qualcomm) I can't imagine Intel would need two expansions at AZ, one at Israel, one at Ireland, and two new fabs at Ohio to satisfy its internal needs. Pat has literally also said that he wants customers on the leading edge, and that if a customer really wants to they can even use yield learning nodes (something TSMC exclusively reserves for Apple). However you are probably right on the count that in the early days of i4 and 20A free IFS capacity will likely be low to nill. i3 might also be tight depending on how many EUV tools they can get. Because most of Intel's fabs are at least already EUV compatible Intel does NOT need to build new fabs to support nodes that have EUV. In short if Intel had no plans to sell chips to foundry customers all of these new fabs they are building would be a waste, given their current capacity is more than intel needs right now. The only other possible explanation is that Intel expects to get ~100% PC/server market share back, make a big splash in the handheld business and push AMD and Nvidia around in the graphics space. Which of these explanations is less outlandish to you?

Plus look at the bottom line. Intel make $18 billion quarterly revenue from mostly CPU business and only $170 from foundry.
That is literally the whole point, the business is brand new. Also the purpose of foundry is to provide extra scale and make intel’s wafer costs lower (not for intel to stop making products). I will also note that once upon a time Apple's handheld devices were MUCH smaller than their Mac business, while today the Mac business is one of Apple’s smallest. Not that I think IFS will eclipse CCG or DCAI, but if 10-20 years from now you said "External IFS contracts would be one of the major money makers for Intel" I would have no trouble believing that.
 
Wrong. Samsung competes with most of their customers, and those customers get their leading edge nodes as soon as Samsung debgugs them. Also you forget that most of Samsung’s capacity is also for itself.
I do not really see this. Samsung apart from RAM, NAND and displays along and with the relevant chips, such as controllers or image processors, is not that big in the mobile sector (and it even buys tons of chips from competitors to use in its products, possible as a way to secure more deals) and mostly absent from a lot of other sectors. Intel on the other hand competes heavily with all possible companies that are in dire need of the most advanced nodes. High performance CPUs, data center accelerators, GPUs, high speed networking, FPGAs, etc etc. Who knows though? It is a paranoid world the one we are living in... Maybe Uncle Sam will pay companies to manufacture their chips in Intel fabs, but if I were AMD three to four years in advance of my new data center processor that my company may live or die with its success or failure, I would feel very uncomfortable taking it to my direct competitor (and practically my only competitor!). And even if it is true that there will be silos between Intel Design and Intel Foundry, if my product is a success would I feel comfortable contributing to the financial success of my competitor and giving it money to keep on competing with me? It sounds a little bit strange, considering that AMD also has a track record of suing Intel for bad business practices. QCOM on the other hand is famous on being aggressive with any opportunity that it can get and is already setup to work with multiple foundries. Still there are many issues that I can think of, but I digress. I see them as practically one of the very few to be able to get to IFS (and also because of their multi-foundry setup structure, I can see that among the big players, they take the minimum risk).
In the end, I think that Intel needs to execute flawlessly and above that it needs a unicorn. It needs a (completely) external company of considerable size and product volume to come and deliver a decisively competing chip manufactured by IFS. This will make or break the story.
By the way, Intel preparing to use external foundries more heavily in the future - despite making such huge investments and marketing on IFS - would also make me feel uncomfortable if I were to sign a multi-million or even billion check to them or even worse if I had to bet the future of my company to them. So I would prefer not to be the frontman in the line. At least till we get that flawless execution and unicorn product.
 
I do not really see this. Samsung apart from RAM, NAND and displays along and with the relevant chips, such as controllers or image processors, is not that big in the mobile sector (and it even buys tons of chips from competitors to use in its products, possible as a way to secure more deals) and mostly absent from a lot of other sectors. Intel on the other hand competes heavily with all possible companies that are in dire need of the most advanced nodes. High performance CPUs, data center accelerators, GPUs, high speed networking, FPGAs, etc etc. Who knows though? It is a paranoid world the one we are living in... Maybe Uncle Sam will pay companies to manufacture their chips in Intel fabs, but if I were AMD three to four years in advance of my new data center processor that my company may live or die with its success or failure, I would feel very uncomfortable taking it to my direct competitor (and practically my only competitor!). And even if it is true that there will be silos between Intel Design and Intel Foundry, if my product is a success would I feel comfortable contributing to the financial success of my competitor and giving it money to keep on competing with me? It sounds a little bit strange, considering that AMD also has a track record of suing Intel for bad business practices. QCOM on the other hand is famous on being aggressive with any opportunity that it can get and is already setup to work with multiple foundries. Still there are many issues that I can think of, but I digress. I see them as practically one of the very few to be able to get to IFS (and also because of their multi-foundry setup structure, I can see that among the big players, they take the minimum risk).
In the end, I think that Intel needs to execute flawlessly and above that it needs a unicorn. It needs a (completely) external company of considerable size and product volume to come and deliver a decisively competing chip manufactured by IFS. This will make or break the story.
By the way, Intel preparing to use external foundries more heavily in the future - despite making such huge investments and marketing on IFS - would also make me feel uncomfortable if I were to sign a multi-million or even billion check to them or even worse if I had to bet the future of my company to them. So I would prefer not to be the frontman in the line. At least till we get that flawless execution and unicorn product.
Samsung is the largest mobile phone maker in the world. It is also one of, if not the largest, makers of consumer electronics and appliances in the world. Most of the low end SOCs in appliances, TVs, and cheaper smartphones are Samsung’s own chips. Even for their flagship devices a significant portion use Samsung SOCs.

Samsung memory/display also gives Samsung devices an advantage over other OEMs who are less vertically integrated. This hurts the customers of Qualcomm and Mediatek. This also hurts Google/Foxcon. The equivalent for Intel would be if Intel started making and selling laptops versus their OEM partners. Samsung was also literally accused of IP theft by apple at 28nm or something like that. One could even argue that Samsung causes more harm to their customers than Intel does to AMD (not that I would personally agree with such an argument).

Going back out to IFS, I don’t think an AMD or Apple would use IFS for most or any of their products. Although this is mostly because they are in the TSMC inner circle and enjoy the awesome benefits that this entails. People outside the circle like the cloud folks, Qualcomm, Mediatek, analog companies, broadcomm, cisco, Nvidia, ect don’t have extensive overlap with Intel or don’t have IP that Intel would want to steal. Furthermore with the move to heterogeneous architectures it will be easy for folks who are still paranoid to only fab part of a design at intel and use their packaging to stitch it all together.

I don’t know if UCIE is changing this but I thought I remember seeing that TSMC requires all logic die they package to be from TSMC. If this is the case I can see this as a big selling point/money maker for IFS. We will also have to see what happens with intel licensing x86 ip out for IFS. If they open up recent architectures or the instruction set I can see alot designers having some interest in that (even in-spite of the long prophesied “immanent death of x86”).
 
This conversation reminds me of the time I was having lunch with a bunch of fellow neanderthals in National Semiconductor's cafeteria when Brian Halla bought Cyrex. "What are we doing getting involved in the digital stuff? We are an analog company!" Brian took the blame when we failed, but we failed because we were a bunch of losers. He was right in buying that company. You play to win. Why wouldn't Intel play to win? They have a nice product and they have a good fab. They probably brought in the Israelis to replace the losers/wokesters (I agree TJ). Mr Gunslinger is a leader. The nay sayers are namby pambies. Make adjustments, and play to win.
 
Last edited:
This conversation reminds me of the time I was having lunch with a bunch of fellow neanderthals in National Semiconductor's cafeteria when Brian Halla bought Cyrex. "What are we doing getting involved in the digital stuff? We are an analog company!" Brian took the blame when we failed, but we failed because we were a bunch of losers. He was right in buying that company. You play to win. Why wouldn't Intel play to win? They have a nice product and they have a good fab. They probably brought in the Israelis to replace the losers/wokesters (I agree TJ). Mr Gunslinger is a leader. The nay sayers are namby pambies. Make adjustments, and play to win.
I totally agree Pat has the right strategy as it’s better to swing for the fences and at least have a chance rather then prioritizing boosting quarterly EPS as per the Brian and Bob era at Intel. My concern is just that Samsung and TSMC have vastly greater resources to call on. For instance Intel just had to go 50/50 on a fab expansion project with Brookfield partners. I get the strategy and it’s better then the alternative but my personal opinion and I am very well likely wrong is that Intel just doesn’t have the financial firepower to continue down this road and especially if there are to be any issues with IFS. Again, it’s the best available strategy but given Intels dismal financial performance this year and upcoming guidance, on top of the fact TSMC is actually accelerating its financial strength and margin growth in a down year for the semi cycle, I personally find the odds unfavourable. As for Samsung I just know they have an ungodly amount of resources to call on as well. As always I welcome criticism of my musings.

 
I totally agree Pat has the right strategy as it’s better to swing for the fences and at least have a chance rather then prioritizing boosting quarterly EPS as per the Brian and Bob era at Intel. My concern is just that Samsung and TSMC have vastly greater resources to call on. For instance Intel just had to go 50/50 on a fab expansion project with Brookfield partners. I get the strategy and it’s better then the alternative but my personal opinion and I am very well likely wrong is that Intel just doesn’t have the financial firepower to continue down this road and especially if there are to be any issues with IFS. Again, it’s the best available strategy but given Intels dismal financial performance this year and upcoming guidance, on top of the fact TSMC is actually accelerating its financial strength and margin growth in a down year for the semi cycle, I personally find the odds unfavourable. As for Samsung I just know they have an ungodly amount of resources to call on as well. As always I welcome criticism of my musings.

A few things that I think should help. Intel 4 and 3 will probably help with Intel’s economics. 10nm has excellent yield and P&P these days, however the process complexity inherent to the process lowers wafer output and speed for a given amount of tools. The sooner intel can dumpster 14/10nm for i16 and i4/3/20A/18A the better intel’s financials will look. Outsourcing part of their SOCs to TSMC should also help with cost in the short term. Fabs and design teams operating as distinct business units should also force both sides to be more efficient and not make choices that screw the other over. Hopefully intel’s the new methodologies and increased pre silicon verification allows the designers to not keep dropping the ball. After all products like Sapphire rapids (and even alchemist products) would be generally competitive if it came out when it was supposed to (as opposed to being situationally useful today).
 
Back
Top