The political situation is confusing, but anyway,

The political situation is confusing, but anyway, the situation of semiconductors is going fast regardless of the situation in Korea, so please comment for a moment

When Korea experienced sudden social turmoil due to martial law in early December 2024, CEO Pat Gelsinger quietly announced his retirement and stepped down as CEO of Intel in the United States. Galsinger has a higher technical understanding than his predecessor, Mr. K, and above all, he was a 40-year Intel man who has rolled through all areas of design-processing since the 386s of the 1980s, so there was some expectation both inside and outside of Galsinger. Although Galsinger was unusually a design engineer (for example, one of the real players in the 80486 design through the 80386), he was also a rare process engineer who participated in foundry process design and knew how to touch equipment. It is really difficult to find an engineer who knows both design and process in semiconductors.

However, after Galsinger, Intel's CTO, briefly left Intel and was transferred back to CEO, the situation at Intel was already like Titanic, which was tilting a lot. It's not that Intel didn't do much to respond to the mobile market as early as the mid-2000s, but in fact, the market did not respond well to the ambitious Atom series AP in 2008. It was because the structure was not suitable for a mobile environment in the first place. Perhaps that's why Intel's Atom was not successful, and Intel eventually withdrew from the mobile AP business in 2016, which had only suffered a deficit. In 2010, Intel's ambitious GPU business was also closed when Lalavi, Intel's ambitious GPGPU, was canceled. Above all, the evolution and stability of foundry process technology, which Intel has traditionally had strengths in, led to a situation where TSMC and Samsung were given technological leadership throughout the 2010s.

In early 2021, when Galsinger took office as CEO, Intel's situation was still lacking much room for improvement, but Intel's stock price rose by 8% anyway, glistening to see if Galsinger, a former CTO, had any expectations of doing something. The market seems to have been really looking forward to him completing his responsibility for the improvement of Intel's constitution, but the fundamental problem with Intel was that it was difficult for Intel's 40-year IDM history to continue any longer. Moreover, since it is a structure where design-manufacturing-OEM is done by itself, there is no concept of a proper fabless customer, and the fact that it has forced the absence of a semiconductor ecosystem expansion strategy until too late made it difficult to adapt to the changing era of Intel's dinosaur.

Throughout the 2010s, Intel's market was questioning why the chips were not made properly and why it was competitive to create products designed by oneself in their own foundry. So, even before Galsinger took office, Intel shareholders and customers demanded that the foundry be separated. Of course, this was not allowed by the pride of Intel, the founder of IDM's business. It would have been difficult to accept requests from Intel to build the CPU he designed, go to TSMC, and turn the foundry into a legacy and just a stable process that makes money. It would have been like a meaning of abandoning Intel's soul.

In any case, Galsinger, who became the new CEO with high expectations and responsibilities, gradually changed the composition of Intel's business division to overcome this internal and external pressure. For example, during his first three months in office, he divided Intel's data platform group into two divisions, aiming for the rapidly changing AI data center market. One division was divided into Data Center + AI Group, and another division was divided into network platforms, IoT, and connected sectors, to create a forward base for AI semiconductors. However, the advance base for going to AI semiconductors is not just about creating departments, but it's also determined whether you actually had the ability to build such hardware properly.

In fact, Nvidia was already dominating AI semiconductors, or GPGPU, in 2021, but Intel had to catch up with No. 2 AMD, which came close to its chin, not No. 1 Nvidia. As of 2020, for Intel, AMD would have been like Hynix to Samsung 10 years ago. Galsinger had a huge task to keep AMD in check and catch up with Nvidia at the same time, but his strategy began to creak from the check against AMD, which was competing for second place right away. He had an interview in which he expressed confidence that Intel's manufacturing capabilities had returned as he neared his first year in office, and that he could sufficiently overwhelm AMD in the same generation.*
*"AMD has done a solid job over the last couple of years. We won’t dismiss them of the good work that they’ve done, but that’s over with Alder Lake and Sapphire Rapids."

Of course, I think this was based on a strategy to show Intel's confidence to Intel investors, rather than Galsinger misjudging the market situation. However, Galsinger underestimated AMD a little bit and Intel overestimated the other way around. The 12th generation Elderlake, which Intel released in the second half of 2021, was manufactured by the Intel 7 process, showing benchmark performance figures similar to or lower than AMD's Ryzen 7600 series, which was thought to be lower. According to the Intel 7 process specification, the benchmark score should be higher than the Ryzen 7900, but it was lower. There was also a fair yield issue here, but there was also a design issue. Intel made changes that optimized the Hyperthread (HT) architecture more from the 12th generation, which would have helped improve the performance of logic chips if it was designed well, but it would have been a miss in the design