Intel (Intel), which has been coveting TSMC's wafer foundry business, has finally resorted to the ultimate killer. A few days ago, there were public reports that Intel will open the X86 core license to compete head-on with TSMC to attract customers to use its wafer foundry business. As we all know, the X86 architecture is a notoriously closed ecosystem, and there is no third one that can be involved in this ecosystem except VIA and AMD.
Now, although VIA has not completely faded out of the industry vision, it is also an indisputable fact that the sense of industry presence is rounded to about equal to zero. Therefore, in the X86 architecture ecosystem, only Intel and AMD are left with the right to speak, of which the former is the big master.
Today's big boss actually takes the initiative to open up the license, mainly because the commercial value ceiling of the X86 ecological architecture has been pressed on the top of Intel.
At the recent San Francisco Investor Conference, Intel said that in 2023 and 2024, the company's operating income will grow between 5% and 9%. The single-digit growth doesn't look pretty, but it has exceeded analyst expectations, who forecast Intel to grow 1 percent in 2022, 3 percent in 2023, and 8 percent in 2024.
This result is really not enough to see. Due to the impact of the epidemic, the demand for semiconductors has increased significantly, the global semiconductor market has increased by 26% in 2021, Intel's old enemy AMD revenue has increased by 68% in the same period, the growth rate of Nvidia has also reached 61%, and Intel has dropped by 4%, not only outperforming two old rivals, but also the industry.
Equivalent to a class student exam, AMD, Nvidia tested close to full score, the former bully Intel even did not get the pass, its embarrassment can be imagined.
Reflecting on the specific market share, Intel is also slowly being cannibalized by AMD. According to data released by market research institute Mercury Research, amD's market share reached 25.6% in the global X86 CPU market, growing for 11 consecutive quarters, a new high in 15 years, while Intel was declining. And Intel's most lucrative and largest share of the server market has not been able to stop AMD's offensive pace, and its share has dropped from 92.9% in the fourth quarter of 2020 to 89.3% in the fourth quarter of 2021.
The real question on Intel's desk is how to find performance growth points and pull up the declining performance curve and stock price.
Continue to fight a tug-of-war with AMD in the original field, the cake is so big, obviously not the optimal solution for Intel, the optimal solution is to expand the incremental market and grab other people's cake. TSMC's wafer foundry business has been remembered by Intel.
However, to grab the rice bowl of TSMC, the outside world is not optimistic. On the one hand, Intel's manufacturing process has a gap compared with TSMC, which is enough to dissuade TSMC customers from switching to Intel; on the other hand, Intel has invested heavily in transformation, including investing as much as $100 billion to build the world's largest chip manufacturing base in Ohio, and spending $6 billion to acquire the Israeli chip company Tower Semiconductor, a large amount of money, but the rate of return is still unknown. Some analysts have calculated for Intel that its wafer foundry business will not make money until 2025. As a result, Intel's stock price fell 7% on the day after it announced its foundry business (2021).
Hard to grab TSMC rice bowl is not OK, Intel is now throwing out a smart solution: aren't you hungry for the monopoly position of my X86 architecture ecology? Then I will learn ARM, open licensing, on the condition that customers who want to license the service come to me to place a single wafer manufacturing. This trick is simply that you ask me to manufacture wafers, and I give you open CPU authorization.
It seems that the rice bowl of TSMC without chip design and ecology has begun to shake, but a careful analysis will find that Intel's ultimate killing move is not shocking enough.
TSMC's manufacturing process is currently state-of-the-art, and even though 5nm is not a true transistor gate width for marketing purposes, it does lead Intel in process. For Apple, Qualcomm, Nvidia and other large manufacturers, whether the manufacturing process is advanced is the priority, and the X86 open license is not attractive at all. As for AMD, it will not be stupid enough to go to the old opponent to place an order.
In this way, Intel can dig mainly TSMC's low-end customers, which are not sensitive to cutting-edge processes. However, under the publicly available Intel Licensed X86 program, it is limited to soft and hard IP. In other words, Intel does not open the X86 architecture instruction set, even if the licensed soft and hard cores, customers do not have the right to customize, but can be combined with ARM, RISC-V and other different cores, and then manufactured by Intel.
This is like the popcorn market, customers can buy two kinds of popcorn from Intel, one is unpackaged (soft core) and the other is packaged (hard core), but no matter which one, customers cannot make popcorn independently (design their own CPU core), nor can they change the recipe to make different flavors of popcorn (modify the CPU core to meet individual needs).
Therefore, Intel's open license X86 is different from ARM's open license (ARM can modify or customize the kernel), for TSMC, it is just a case to sell the CPU. Intel wants to sell CPU pull wafer foundry customers by changing shells, and it is obviously difficult to attract TSMC's customers who are not sensitive to cutting-edge processes.
After Intel launched the wafer foundry business, there was news that Qualcomm would become the first customer, but Qualcomm's new CEO implicitly denied this news in the earnings conference call, to the effect that there was no specific product plan for production at Intel. Amazon has become a customer of Intel, but it is only looking for Intel packaging chips, not manufacturing.
In fact, Intel is not the first time to get involved in the wafer foundry business. Since 2010, Intel has begun to get involved in wafer foundry, when it was still significantly ahead of TSMC in the process process, but intel voluntarily gave up in 2018. Now, Intel has restarted the wafer foundry business again, but the initiative of competition is in the hands of TSMC, and it is not easy for it to win this game.
——END——