There was a quarrel! AMD executives sarcastically said that Intel's foundry business was "doomed to failure"?

"Do you think Intel's IFS (foundry business) strategy will be a success?"

"Of course not. Divesting manufacturing and R&D operations is more conducive to growth. ”

In the face of questions from the media, Darren Grasby, executive vice president of AMD and president of EMEA, gave a very clear answer. Darren Grasby's remarks to some extent represent the negative attitude of the outside world towards Intel's model of manufacturing, design, and delivery, as well as the ill-fated foundry business.

Objectively speaking, Intel IFS does face quite a few challenges. For example, the overall shrinkage of the wafer foundry market, TSMC and Samsung, which have always had a solid dominance, and high factory construction costs. Even Intel CEO Pat Gelsinger has been humble lately, arguing that IFS's results over the past two and a half years can only "give a passing grade."

However, modesty does not mean that Intel will allow its old opponents to provoke him, nor does it mean that Intel will easily give up this new business with heavy investment. On the contrary, Intel IFS has been constantly moving during this period, fully arming itself from the dimensions of technology, production capacity, and commercialization.

The golden age of PCs is long gone, and Intel, as the CPU hegemon, is clear that there is no turning back. GPU, wafer foundry, any growth point is impossible to give up. The negative comments of AMD executives will only prompt Intel to speed up again.

(Image via UNsplash)

Behind the public ridicule, Intel's foundry business once again provoked AMD's sensitive nerves

As we all know, AMD and Intel have been fighting for half a century, whether it is customer competition in the mall, you chase me on the technical side, or verbal battles in private are common.

Going further, AMD founder Jerry Sanders and several co-founders of Intel Robort Noyce and Gordon Moore all worked at Fairchild Semiconductor before starting their business, and they were exposed to similar technologies and overlapping network resources. This also makes the outside world keen to compare Intel and AMD, two companies that are less than a year apart in their founding, and the former seems to have shown superior strength and popularity from the beginning.

After the establishment of the company, the frontal conflicts between the two sides were endless. In the 80s, in order to compete for IBM's exclusive cooperation, Intel terminated AMD's 80386 series chip technology license in advance, and the latter took Intel to court in a lawsuit. The legal dispute between the two sides, as well as Intel's technology blockade, to some extent delayed AMD's research and development process, and indirectly caused the latter to miss the golden age of the PC.

Closer to the past, after entering the 21st century, AMD finally sounded the clarion call for counteroffensive and constantly narrowed the gap between it and Intel. For example, in 2004, the market share of AMD desktop processors exceeded Intel for the first time, and in 2009, Intel shifted its focus to the GPU business in an early step, and then launched the Ryzen series with performance and cost performance far beyond similar products of Intel, all laying the foundation for overtaking in the future.

Jerry Sanders once ridiculed himself that Intel can raise $5 million in 5 minutes, while AMD can raise $50,000 in 5 million minutes, which also lays the groundwork for the differential treatment of the two in the capital market. So much so that when AMD's market value exceeded Intel for the first time in the future, various media came down to create momentum, bluntly saying that this was "the end of an era."

Given the decades-long feud between the two sides, it's no surprise that executives at Darren Grasby's level are sarcastic about old wrongdoers. What the outside world should really pay attention to is the faint anxiety of AMD behind this remark: finally complete the corner overtaking, should Intel really rely on wafer foundry to make a comeback?

As we all know, Intel launched the first generation of 14nm process processors in 2014, more than one position ahead of TSMC and Samsung. If it weren't for Intel's unwillingness to give up the illusion of CPU at that time, it would not have let TSMC and Samsung come later.

Of course, the technical advantages of TSMC and Samsung are based on crazy money. Between 2013 and 2017, when the two invested the most crazy, Samsung spent an average of $25.5 billion in annual capital expenditure, Intel less than 10 billion, and most of the budget was allocated to the design department.

It's not too late to wake up. According to official forecasts, Intel's capital expenditure as a percentage of revenue after increasing the foundry business may be close to 35% at most, compared with about 20% in the past 10 years. Whether it is recruiting a research and development team or building a factory, Intel has smashed its blood.

AMD's fear of Intel's foundry business is based on two considerations: one is the boost of revenue and profit, that is, the huge market size of the wafer foundry industry; Second, there is still no room for technological upside in the ceiling, and doing a good job in the foundry business is also beneficial to its GPU business - and the GPU is the key to AMD surpassing Intel.

The former point is easy to understand, and the latter worry is mainly based on TSMC and Samsung's slightly tight production capacity, and the importance of foundry technology to GPUs. AMD's GPU can rise, it is from the beginning of holding hands with TSMC, Intel also announced in 2021 that the two series of independent graphics of Ruixuan and Ponte Vecchio will be handed over to TSMC. Once the foundry technology catches up with or even surpasses TSMC and Samsung, Intel can in turn jam AMD's neck.

Of course, Darren Grasby's deep thoughts are one thing, and it's still too early for Intel IFS to talk about surpassing TSMC. Intel also does not need to pay too much attention to the former's words, and it is most important to do its job well.

Increase investment, attract customers, crazy efficiency, Intel will not give up IFS

As mentioned at the beginning of the article, Intel IFS has suffered a lot of troubles recently.

On the one hand, the acquisition of Israeli chip manufacturer Tower Semiconductor failed, and Intel regretted missing out on a large number of technology patents, outstanding talents and customers, and paid a large amount of breakup fees. However, the two sides still joined hands to reach a number of cooperation after the failed acquisition, such as Tower's investment of $300 million in Intel's Mexico State plant to improve equipment and acquire some fixed assets, which also offset some losses.

On the other hand, the talent shortage in the foundry business is still very prominent, which has also slowed down Intel's factory construction plan in Germany. In the past two years, Intel has poached many executives from TSMC and Samsung, such as Hong Hao, former senior vice president of Samsung, and Michael Chang, deputy general manager of the original design infrastructure management department who worked for TSMC for more than 30 years, but unfortunately still cannot meet the talent needs under large-scale expansion.

However, it is certain that no matter how many obstacles lie ahead, Intel has never intended to reduce its investment in IFS, nor will it waver in its original course. What Intel IFS needs to do is always inseparable from the three keywords of technology, capacity, and orders.

The first is, of course, to strive to make a breakthrough in technology, and put aside the gap and cooperate with direct competitors such as TSMC.

Pat Gelsinger has said that starting from March 2022, Intel aims to break through 5 technology nodes within 4 years and advance the foundry process from Intel 7 to Intel 18A. Accelerating penetration from simple foundry to packaging, testing and other links is also one of its goals. For example, in September this year, Intel announced cooperation with TSMC to jointly develop multi-chip packaging chips.

Secondly, Intel opened a crazy factory construction mode in the first half of this year, with a high proportion of foundries, in order to reserve production capacity as much as possible.

Among them, the most concerned are the two chip manufacturing plants in Magdeburg, Germany, with a total investment of about 33 billion US dollars, which is also the largest foreign investment in German history. In addition, Intel will also invest $25 billion in Israel to build a high-specification foundry, which costs more than two factories in Germany.

Despite the failure of the acquisition of Tower Semiconductor, Intel's love for Israel's semiconductor hot land has not disappeared, and there is even a trend of making it an overseas base of IFS. What Intel is looking at is the rich talent resources of more than 200 local chip companies.

Finally, Intel is also fully competing for customers, frequently releasing goodwill to former wrongdoers such as NVIDIA and Qualcomm.

As of the end of the last fiscal year, Intel's foundry business accumulated orders of more than $4 billion, and TSMC and Samsung are certainly not in the same order of magnitude, and there is a lot of progress. Among them, Amazon's AWS and Qualcomm are Intel's two most important external customers now, and Intel's internal orders are also very critical to IFS.

Speaking of which, we have to mention the point put forward by Darren Grasby: Is the asset-heavy operation model that integrates chip design and foundry, as well as the internal digestion of foundry orders, really conducive to the development of enterprises?

As Darren Grasby points out, Intel's IFS route is indeed debatable. One of the biggest problems is the heavy cost burden, and Intel also acknowledges that the focus on the foundry business will lead to a spike in capital expenditure.

Striking a balance between cost and profit is a constant challenge. The size of the foundry market is vastly different from when AMD abandoned the foundry business, but the competition has also become more intense. When Intel's huge investment in the early stage will enter the harvest period, no one can give a definite answer.

On the spin-off topic of the foundry business, as well as the independent development model of the foundry and chip design business, AMD's previous experience is indeed worth Intel's reference.

PSG

, IFS have become independent, can the spin-off method help Intel return to the top?

Turning the clock back to 2008, the uproar of AMD's spin-off plan landed, and AMD's chip production division accepted a total of $8.4 billion investment from the UAE ATIC and MUBADALA funds, officially moving towards independence, and the new company was named Doundry Co. - that is, everyone knows Grofoundries. AMD continued to reduce its holdings in the following years, and finally completely abandoned its stake in GLOBALFOUNDRIES, and the two turned into a pure partnership.

AMD's experience in the past ten years shows that this spin-off is indeed a good move: after throwing off the OEM baggage and traveling lightly, AMD quickly reversed the unfavorable situation of consecutive losses of more than $3 billion in 2007-2008 and achieved profitability; Investing the saved money in GPU R&D projects is also the right decision.

The independent development of IFS has long been put on the agenda by Intel. In June this year, Intel announced a restructuring in which IFS will operate independently, settle separately, and continue to work with other Intel business units as a third-party foundry.

However, compared with the spin-off of AMD and Globalfoundries, the independence of IFS is not complete enough.

According to the official release of information, the spin-off is mainly to "improve efficiency": including improving the self-decision-making efficiency of the IFS management team, and the contact and communication efficiency of third-party OEM customers in order to improve productivity. However, due to its self-hematopoietic ability, IFS may not be able to become independent in the short term at the capital expenditure level, although Intel expects to save about $3 billion in costs after the spin-off is completed in 2023.

Of course, there is no single standard answer to whether to spin off or not, the specific spin-off method and the development path after the spin-off. In the end, AMD and Intel now want to find a balance in cost, profit, market share, or explore the best compromise. Looking to the future, with the spin-off of IFS and the implementation of the IDM 2.0 model, the improvement of the operating efficiency of Intel's foundry business is still very promising to offset the increase in costs.

Under Pat Gelsinger's vision, IFS will play a similar pivotal role in the IDM 2.0 model, working closely with other business units in the group to provide foundry, tape-out, testing and other services, and settle fees at market prices. In addition to improving communication efficiency and controlling costs, the collision of various departments on the technical side will not be interrupted, and the communication between IFS and external customers will become more, which is more conducive to understanding market prices and technological changes.

(Image from Intel)

Ideally, in Pat Gelsinger's words, IDM 2.0 could completely change Intel's operating model. External customers can let their guard down and fully trust IFS's confidentiality of core technologies under the information isolation system.

It is worth mentioning that after IFS, Intel's Programmable Solutions Group (PSG) was also exposed in early October to operate independently, and is expected to IPO independently within 2-3 years, and Intel does not mind selling part of its equity. It can be seen that Intel also realizes that its business line is getting bigger and bigger, and it is time to break the situation of eating big pot rice.

It remains to be seen whether the spin-off of Dafa and IDM 2.0 will be a great success in the future. But at least, the first step has now been bold.