These days, whenever I meet with economic ministry officials, one topic invariably comes up at the table: "When do you think the semiconductor super cycle will peak?" Considering that just two or three years ago, "worrying about Samsung" was practically a national pastime, the change feels like a different era. High Bandwidth Memory (HBM), which once sounded like an alien term, has now become so commonplace that snacks named after it are sold in convenience stores.
Until now, it was considered common knowledge that semiconductor market conditions repeatedly expanded and contracted with economic cycles. This was due to the nature of semiconductors—whether made by Samsung or Micron, products with the same specifications show no difference in quality. Because of this characteristic, factory yield rates determined a company's operating profit margins. This is why semiconductors were viewed as a commodity.
However, as we entered the artificial intelligence (AI) era, the rules of the game changed. How many AI chips a company can secure and how powerful its computing and inference capabilities become has now become the criterion that determines a company's survival. The companies betting their lives on this are the so-called Big Tech firms like Google, Meta, Amazon, and Microsoft. As they pour hundreds of trillions of won annually into facility expansion, semiconductors have become precious commodities. In this process, memory bottlenecks emerged, giving rise to the current semiconductor mega cycle.
Those who view this phenomenon as irrational frenzy—the AI bubble theorists—argue the following: ①Capital expenditure is growing to levels that even the mighty Big Tech companies cannot sustain. ②The revenue being generated is far too meager compared to the scale of spending. ③AI's productivity improvement effects may have already reached their limits.
The opposing side argues: ①This is different from the dot-com bubble, when companies that couldn't generate a single penny in profit were racing ahead. Compare the price-to-earnings ratios (PER) of companies that collapsed back then with the PERs of Nvidia or Google. ②While there may be some delays, AI will become infrastructure like power grids or the internet. ③Companies have already invested so much that no one can stop now (Too Big to Stop).
In a world with volatility as extreme as a roller coaster, it's difficult to predict what will happen. However, for at least the next one to two years, the AI optimists' arguments appear more likely to have stronger momentum in driving the global economy forward. This is why market forecasts suggest Samsung Electronics could earn 200 trillion won annually and SK Hynix could bring in 100 billion dollars per year.
From this perspective, our questions going forward should change. Rather than calculating when the semiconductor cycle will end, it's time to seriously consider how to leverage this thunderbolt of a blessing.
First, we must pursue bold investments to escape our single-legged growth structure dependent on semiconductors. Of South Korea's $709.7 billion in exports last year, semiconductor exports accounted for $173.4 billion, representing 24.4% of the total. Based on cumulative figures from February 1-20 this year, that share has reached 35%. If semiconductors catch a cold, the national economy will come down with a fever. This is why strong development measures for struggling industries like batteries and solar power must emerge within the next one to two years while Samsung and SK are raking in hundreds of trillions of won. If left as is, these industries are highly likely to be devoured by China.
Semiconductor companies also need to fundamentally rethink their approaches. For instance, in the case of Samsung's Yongin semiconductor complex, there should be serious discussion about building fabs for production capacity beyond 2040 in regional areas. This is the only solution that simultaneously addresses South Korea's energy problems and regional depopulation crisis as an island-like nation.
Of course, such regional relocation should not become coercive "arm-twisting" pressure. What moves economic actors is not punishment but incentives. For example, what if the government made a special agreement to provide the entire 20 trillion won promised to the Integrated Special City exclusively to investing companies? Even for Samsung, 20 trillion won is an amount that cannot be ignored. I hope our economy takes this opportunity to form a united public-private team and properly devise ways to leverage this bubble.