Updated to reflect the latest situation in the semiconductor war.
1. March 18. Putin unveils the Kinzal hypersonic cruise missile. That Kinzal shot this time towards Ukraine.
2. Kinzal, which means dagger, was launched from a MiG 31 and hit a target more than 2,000 km away at a speed of Mach 10.
3. It was not a ballistic missile that fell just because it went out of the atmosphere like an ICBM, but rather hit a target at hypersonic speed in the atmosphere.
4. Russia is viewed as the successor to the Soviet Union, which had been at war for 45 years as the two major powers in the past, and the US acknowledged the increase in military power.
5. China is different.
6. The fact that China has hypersonic missiles capable of hitting targets of more than 2,000 kilometers means that the naval hegemony of the United States through the carrier front will be directly challenged.
7. China also succeeded in launching a Mach 6 hypersonic missile and touched the nerves of the United States.
8. The US believes that the key to determining future military hegemony is aerospace based on hypersonic missiles and artificial satellites.
9. The core of aerospace is basic precision machinery technology and parts industry, 5G, core research capabilities and semiconductors.
10. Precision machinery can be managed because Japan and Germany are the US allies at the forefront.
11. 5G caught Huawei to make time, and it is catching up with Chinese by blocking Chinese short-term technology research institutes in the US and Japan.
12. Among the core sectors of aerospace, the remaining unmanaged are semiconductors.
13. Among semiconductors, design is managed by the US patents and US companies, but the foundry, which can be viewed as a manufacturing plant, is a problem.
14. Currently, TSMC is the top foundry, and Samsung Electronics is following.
15. As Intel fell, TSMC and Samsung Electronics became a two-way battle.
16. Intel makes very well up to 10nm after 14nm. He was called a craftsman of 14 nanometers and even called a craftsman of 10 nanometers.
17. In the same 10nm semiconductor performance, if Intel is 100, Intel is good at 10 nanometers to such an extent that Samchun is 52 and TSMC is 47.
18. However, Intel failed to enter 7nm.
19. Other latecomers, SMIC, are 12nm, UMC is 22nm, etc., which is more backward than Intel. 7nm, the current cutting-edge process, is a two-way system between Samsung Electronics and TSMC.
20. Samsung Electronics succeeded in entering 7nm, but there was a problem.
21. Samsung Electronics boasted that TSMC was eliminated by making 7nm as a more advanced EUV facility rather than a DUV.
22. But the results were disastrous.
23 . Apple left for TSMC, and even Mvidia, which had supplied new shipments three years ago, gave up the Samsung Electronics 7-nano and left.
24. Yield was not found.
25. I poured a lot of money to expand and open a restaurant, but a group of customers went to the store next door.
26. This is one of the main reasons why the number of samjeon, which was running for 100,000 electrons, fell to 60,000.
27. Currently, semiconductors are nano-competition.
28. 1 nano is 1/100,000th the thickness of a human hair.
29. The problem is that as the line gets too thin, the electrons start to get out of control.
30. FinFET process comes out.
31. Existing flat-panel FETs were in contact with one side, but FinFETs raised the bottom so that they were in contact with three sides instead of one side.
32. Stability is increased by touching the line in three places, the line width is not reduced, but the performance is improved.
33. The line width is not reduced to 10 nanometers, but it is said that 7 nano grade performance comes out when it is made with the FinFET process.
34. There is a better process.
35. GAAFET.
36. FinFET is tangent to 3 sides, but GAAFET is through the middle and is tangent to 4 sides.
37. Theoretically, GAA fat that touches 4 sides is more stable and has better performance than FinFET that touches 3 sides.
38. Samsung Electronics has a patent for MBCFET, an upgraded version of GAAFET.
39. Samsung Electronics' 3nm, which is scheduled for mass production in June of this year, is MBCFET method.
40. 30% better performance and 35% smaller size than the existing FinFET.
41. There is a bigger performance improvement than this.
42. Consumes only half of electricity compared to the existing Finpet.
43. That means it can be used for a longer period of time.
44. TSMC is also challenging 3nm.
45. TSMC was scheduled to mass-produce 3nm in February this year, but the schedule was pushed back to July, and if it continues, the lead will be given to Samsung Electronics.
46. The difference between TSMC and Samsung Electronics is that TSMC is challenging 3nm by replacing the existing FinFET method, while Samsung Electronics is challenging the MBCFET method.
47. Samsung Electronics' argument so far is "I don't know 7nm with the existing FinFET method, but 3nm would be difficult." lim.
48. TSMC's position is that "FinFET method can be used up to 3nm, and GAA method can be done with 2nm, which is scheduled for 2024."
49. TSMC postponed the 3-nano mass production scheduled for February this year to July without any special explanation, highlighting what Samsung Electronics said again.
50. Samsung Electronics said, "Isn't it the limit of the FinFET process?" The question has been raised.
53. If TSMC and Samsung Electronics succeed in mass production, the next thing will be yield competition.
54. TSMC's 3nm is already the No. 1 supplier to Apple, and Apple takes it and the remaining ones are Intel and Qualcomm, and the reservation ends. It means that once you make it, you have decided where to sell it.
55. Samsung Electronics' first external customer is still blank.
56. Qualcomm, which was expected to be the first 3nm external customer, chose TSMC instead of Samsung Electronics.
57. The fact that the yield of 4nm, which Qualcomm entrusted to Samsung Electronics, was only 35% is a big influence.
58. The 35% yield means that if 100 are made, 65 are defective. Qualcomm is struggling with 4nm, but it is decided that it cannot be entrusted to Samsung Electronics for 3nm.
59. When the 4nm yield was too low, Samsung Electronics began to cover the responsible material, such as an internal audit.
60. As Samsung Electronics failed consecutively in 7nm and 4nm, 3nm was unconditionally successful.
61. Fortunately, TSMC does not seem to be able to mass-produce 3nm.
62. TSMC postponed mass production in February to July and then postponed mass production in July again. Not only did he act, but he could not clearly reveal the date of mass production.
63. Samsung Electronics started preparing for mass production at the Pyeongtaek 3 plant, and around the summer of this year, it is scheduled to start testing equipment in earnest and test production in the second half of the year.
64. Once the mass production is successful and the yield is at the expected level, AMD seems to be the first customer, which missed the lead of TSMC to Apple, Intel, and Qualcomm.
65. In this situation, Intel appeared out of the blue.
66. What the United States, who is approaching semiconductors for security, wants is a high-tech semiconductor factory built in the United States.
67. Samsung Electronics built a state-of-the-art semiconductor factory in the United States, and the United States provided national support, creating a win-win atmosphere in which semiconductors are managed in the United States.
68. TSMC responded that it would invest 100 billion dollars to build factories not only in Taiwan but also in the United States.
69 . In the midst of this, a variable arose when Intel, which had unexpectedly failed 7nm, jumped into the semiconductor competition.
70. Unlike Samsung and TSMC, Intel is a completely American company.
71. If Intel, a US company, takes on the challenge of cutting-edge semiconductors, the US government and Congress have agreed that it should support Intel, not Samsung or TSMC.
72. The priority of support in the United States has become Intel, and the US Congress has proposed to provide semiconductor subsidies only to companies whose headquarters are located in the United States.
73. This means that the headquarters will give semiconductor subsidies to Intel in the United States, rather than Samchun and TSMC, whose headquarters are in Korea and Taiwan.
74. Intel, behind the US government, skips 3 nm and declares that it will challenge 2 nm by 2024.
75. It is not surprising that Intel, which failed even with 7nm, will go straight to 2nm instead of 3nm.
76. Intel has no choice but to do so.
77. Samsung Electronics and TSMC are on the verge of starting mass production of 3 nanometers soon, but it seems a bit like a request for government support for a challenge to 3 nanometers.
78. The US government will also be vague.
79. Samsung and TSMC are almost certain and near reality, and Intel has an uncertain future, and there is a possibility that it will be a prank.
80. First of all, Samsung and TSMC will compete in the 3nm market by 2024.
81. It remains to be seen whether the 2nm competition in 2024 will be a two-way battle between Samsung Electronics and TSMC as it is now, or a three-way battle with Intel.
