Following the previous announcement of the 10-year semiconductor industry development plan by former South Korean President Moon Jae-in, in May 2023, the Ministry of Science and ICT of South Korea released a 10-year R&D roadmap. The former plan focuses on the current development direction of the industry, especially system chips, with IC design companies and foundries being the two most important components. The plan is clearly inspired by Taiwan's example, indicating a desire to compete with Taiwan in this field.
As for the 10-year R&D roadmap, it combines the efforts of industry, government, and research institutions to develop emerging memories, logic chips, and advanced packaging, encompassing almost all future new technologies in the semiconductor industry. Is the policy too vague? No. This is not an industrial development plan but a forward-looking technological R&D initiative that has a broad scope, aiming to function as an insurance. For example, in the field of emerging memories, research projects cover FeRAM, MRAM, PCRAM, ReRAM, and more. If there is a technology that will eventually prevail, it will not be missed due to the wide scope of research projects.
The strategy of extensively covering forward-looking technologies naturally comes with financial and manpower challenges, but South Korea ranked 12th in GDP in 2022. With the nation's concerted effort, South Korea has the ability and made the right choice to vigorously compete in its most important industries.
There are two highlights in South Korea's plan that should concern Taiwan. One is in-memory computing, which involves performing calculations directly in memory. In the traditional von Neumann architecture of computer design, instruction and data are stored in the same memory. As the CPU and memory need to share the same pathway (bus) for data transfer, it leads to a potential bottleneck in the system's overall performance, especially in high-speed and large-scale computing. Therefore, performing computations directly in memory and locally storing them has become one of the solutions. This topic has received increasing attention in recent years at various semiconductor conferences.
Another highlight is neuromorphic chips, which simulate the structure of neurons and synapses in the human brain to perform functions such as learning, thinking, and memory. Current artificial intelligence (AI) computing is primarily based on GPU chips. Taiwan's semiconductor industry has experienced significant growth due to the rapid rise of ChatGPT, and in the future, it may be possible to execute AI computations using neuromorphic chips. Intel has already introduced two generations of products in this field.
Both of these are well-known future trends in the industry, and the key point is that both are based on emerging memories. While Taiwan's foundries will also cover the development of embedded emerging memories, they may not be as focused as independent memory manufacturers. Although Taiwan's memory manufacturers used to have the second-highest production volume in the world, the individual scale of the companies was relatively small, making it difficult to afford NAND development costs. They also suffered from the impact of the 2009 financial crisis, causing them to fall behind. Without sufficient participation from domestic memory manufacturers, Taiwan's development in these areas is relatively lacking.
Even in advanced packaging, Taiwan faces similar issues. Techniques such as WoW (Wafer-on-Wafer) and CoW (Chip-on-Wafer) involve incorporating two or more chips, some of which include memory chips like DRAM. Generally, these are designed and manufactured by specialized memory manufacturers. Taiwan lacks domestic memory chip support, which will inevitably pose challenges in future competitions.
In summary, South Korea's 10-year R&D roadmap covers various aspects of the future of semiconductors, with the nation pooling its efforts to push it forward. The plan fully utilizes South Korea's established absolute advantage in the memory sector to project into the future development of technologies. In my opinion, this is a reasonable plan.
As for Taiwan's policy, what did the past "5+2+2+1" initiative (which later included semiconductors) and the recent six projects under the previous cabinet say about semiconductors? What have they actually accomplished? Or to be more upfront, does Taiwan have a semiconductor national strategy?
