Investing is like a prediction market.

Investors bet money on the future they foresee, accumulating wealth or suffering losses based on the results. While corporate analysis and earnings reports are important, the essence of the battle lies in gauging what kind of future will unfold. However, we do not possess memories of a future we have actually experienced. Therefore, we have no choice but to imagine in our minds which future has the highest probability of realization.

Historically, the future most likely to be realized is the one desired by the most powerful people in the world. This is because they possess the power and will to actually shape the future they desire. Obstacles standing in the path of that future—no matter how difficult they seem to remove—are highly likely to be cleared eventually. Conversely, fields helpful to that future—no matter how difficult they seem—are highly likely to grow. This is why it is crucial to study what kind of future political powers like U.S. President Trump and the technological powers known as “Hyperscalers” desire.

The future spoken of by the world’s most powerful people all points to the AI era. For the United States, AI is no longer just technology; it is the core of the struggle for technological hegemony against China and a matter of national security. For the U.S., which aims for an overwhelming power gap, “overinvestment” is the default setting. For hyperscalers, AI is a matter of survival. Even if the current AI race turns out to be overinvestment later, all they lose is money. However, if AI is not a bubble, the gap resulting from timid investment might be impossible to close. This is why it is difficult to stop investing, no matter how many warnings of an AI bubble are raised.

The best way to monetize the future is to find the “bottlenecks” that will arise and the “misconceptions” that will be resolved in that future. You can achieve good results by investing in areas where demand explodes but supply is insufficient, creating a “bottleneck,” or where a bottleneck blocking growth is resolved, leading to rapid expansion. Additionally, significant profits can be found in the process of reevaluating sectors that were misunderstood due to conventional wisdom—points where reality has already changed, but perception has not yet caught up.

Until recently, the “bottleneck” of AI was the GPU. Securing GPUs determined the speed of development. Hyperscalers engaged in the AI race wanted to grow their models quickly but couldn’t expand due to a lack of GPUs, and Nvidia reached a point where it had no chips left to sell. Occupying the bottleneck of the AI wave, Nvidia grew more than tenfold since the launch of ChatGPT in October 2022, rising to the number one spot in global market capitalization.

However, as we entered 2025, the situation changed. People still held the “misconception” that GPUs were the bottleneck, but it turned out that the biggest bottleneck in the AI race was not the GPU, but the power grid required to run them. To handle the exponentially increasing demand for computer computation, data centers must be built, but the physical power supply could not keep up. A phenomenon even occurred where hard-won GPUs were stockpiled because there was no power to run the chips. Consequently, Bitcoin mining companies that had secured power cheaply in advance, as well as energy companies including nuclear power, are being reevaluated.

The area where growth bottlenecks are being resolved and rapid growth is occurring is the Physical AI market and the robot market. The development of AI has brought innovation to the speed of robot technology growth. Originally, every action of a robot had to be manually programmed by a developer. Because learning was not cumulative, just because a robot could pick up an apple didn’t mean it could pick up a handkerchief; thus, the speed of programming became the bottleneck for robot development. However, modern robots have become smart enough to apply untrained tasks after only a few dozen learning sessions via AI foundation models. The achievements made by AI in the past year exceed what humans achieved through direct programming over the past decade. There is no longer a need to manually program code. This is because “Vibe Coding”—conveying the desired program to AI verbally—is becoming the standard. There is even a saying that the hottest programming language recently is English, suggesting that the conquest of the computer programming field by AI is nearly complete. The mega-trends that will determine the future are already set.

However, the Korean market offers few ways to monetize this AI wave. This is because there are almost no companies possessing AI source technologies or economic moats competitive on the world stage. If we must pick, there are only Samsung Electronics and SK Hynix, which make the semiconductors used in AI chips. Last year, the KOSPI index recorded a return of 76%, and Samsung Electronics and SK Hynix accounted for 48.8% of this rise. Currently, the weight of these two companies in the KOSPI market cap is 37%. Aside from these two, the U.S. stock market is currently almost the only avenue for individual investors to directly access AI investments.

From an investor’s perspective, things the Korean government told people not to do often ironically turned out to be the best investments. The designation of “speculative overheating zones” for real estate in 2017 and “land transaction permit zones” in 2020 ended up signaling to the market exactly where the “Government Officially Certified Promising Real Estate Investment Areas” were. Bitcoin, which the government hinted at banning trading for in early 2018, dropped briefly but then rose more than tenfold, remaining an investment asset with the highest cumulative returns to this day. There is no one who doesn’t know that the government telling you not to do something is a signal for a “good investment.” That same government has recently brought up the phrase, “Do not invest in U.S. stocks.”

How you accept those words is up to you.