Big Tech's Slowdown Remarks Questioned; Bottleneck-Solving Companies Are Key

Professor Kim Hak-ju pointed out that these companies have been playing a winner-takes-all game, and analyzed that if they slow investment, many startups would struggle.

This would ultimately create an opportunity for Big Tech companies to acquire good startups through mergers and acquisitions (M&A) at bargain prices. Professor Kim said, "Haven't those Big Techs been playing a winner-takes-all game? How sincere is this statement of theirs, and can they really slow down?" He added, "If they could slow down, what would the likely scenario be? They would hold back, and then a lot of good startups would go under. After all, U.S. IT companies grow by acquiring such firms through M&A. But now they could buy them at bargain prices."

Professor Kim Hak-ju analyzed that Trump, having given up on the midterm elections, wants a third term and is preparing a major event to achieve it.

Professor Kim argued that Trump wants a "Greater America route" and that the natural resources beneath it could cut a considerable portion of U.S. debt. He explained that securing Greenland is essential for this. The view is that Trump's moves may be groundwork for long-term goals of resource and energy hegemony rather than mere political calculation.

This analysis is linked to moves by the U.S. to strengthen energy security and secure new resource supply chains amid instability in the Middle East. In particular, Greenland is an important hub on the Arctic shipping route and holds vast natural resources, so whether it is secured is expected to have large repercussions for international affairs.

Professor Kim Hak-ju stressed that if Buffett had judged that AI development would be meaningfully delayed, he would not have made such a large investment, since he does not take on risks of that kind.

Professor Kim said, "If AI were actually delayed meaningfully, as we worry it might be, Buffett would not have bought. Buffett is not someone who takes on this kind of risk," interpreting the investment as reflecting Buffett's judgment that AI technology development will not be significantly delayed. Amid growing concern about a slowdown in AI investment, this could be an important signal to the market.

Analysis suggests that Warren Buffett tends to invest only in companies that hold the upper hand in supply chain bargaining power. Professor Kim Hak-ju cited the timing of Buffett's 2016 investment in Apple as an example, explaining that he does not invest in a Big Tech company when it is being pushed around by its supply chain.

Buffett buying Google is interpreted as a judgment that Google has the capacity to reduce its supply-chain dependence and strengthen its bargaining power, for example by developing solutions that use less memory or by making its own chips. Professor Kim said, "Warren bought Apple in 2016. After that played out, if a Big Tech is being miserably pushed around by its supply chain, then he won't buy the stock." He added, "That he bought Google suggests Google will likely be able to release a solution that uses less memory, and since Google is also making its own chips, it will become more and more so."

The explanation is that if these split-up data centers are connected by communication networks to link their computations, they can perform large calculations without concentrating high-capacity data centers in one place. Professor Kim said, "There are many places where small modular data centers could be built. What if you build small data centers and connect their computations?" He said this would allow many data centers to be built in many locations. However, Kim stressed, "That connection is the most important part," emphasizing the importance of communication infrastructure that efficiently links the distributed data centers.

Professor Kim Hak-ju explained that Starship will shorten the cycle for reusing launched rockets and can place far more satellites into orbit per launch than the Falcon 9.

While the Falcon 9 currently launches about 20 satellites at a time, Starship can launch 50 to 60, allowing it to carry two to three times as many low-orbit satellites. Professor Kim said, "We're talking about reusing rockets that were already launched, right? The plan is to shorten the reuse cycle, and when they launch, the Falcon 9 carries about 20 now, while this one launches 50 or 60, so each launch can carry two to three times more low-orbit satellites." This is expected to play an important role in expanding Starlink, SpaceX's satellite internet business, amid a global race to build high-speed internet networks.

Professor Kim Hak-ju explained that Corning is not only the world's No. 1 company in optical fiber but also produces connectors.

He also named Amphenol and Connectivity as companies vying for the top two spots in connectors worldwide. Professor Kim said, "Corning is the number one in optical fiber. It doesn't only make fiber but also makes connectors. And for connectors worldwide, the number one and number two are Amphenol and Connectivity."

Recalling an earlier case in which he stressed the importance of cables and connectors to solve wiring-break problems in humanoid robot joints, Kim predicted that the roles of these three companies will become even more prominent in the countless connections and communications among data centers. As AI-based high-performance computing environments spread, demand for optical communication and connector technologies for stable, fast data transmission is expected to keep growing.

In conclusion, these three companies play a central role in building essential AI-era infrastructure and are analyzed as likely to benefit significantly in the future. In the complex connection networks of AI data centers, the technological strength of these companies will matter even more.

High-voltage direct current (HVDC) transmission technology is attracting attention as one way to resolve the chronic power shortage in the artificial intelligence (AI) industry. Professor Kim Hak-ju explained that using direct current transmission instead of the alternating current now commonly used could increase transmission capacity two to three times.

This minimizes power loss and allows more electricity to be transmitted efficiently, which could greatly help ease the power shortage for high-consumption facilities such as data centers. Professor Kim said, "If we could send transmission as direct current instead of alternating current, we could increase transmission capacity two or three times over what it is now."

Direct current transmission has already reached the commercialization stage, and about 5% of the world's transmission grid currently operates on this method. Professor Kim added, "There seem to be quite a few attempts to send grid power as direct current. Yes, so it has already started to spread. About 5% is already done this way."

This trend is expected to accelerate as electricity demand surges with AI advances. The spread of direct current transmission technology is expected to play a key role not merely in raising power efficiency but in overcoming the fundamental limits of AI infrastructure buildout.

As distributed modular data centers emerge as an alternative, concern over hacking from their interconnection is growing, and the importance of cybersecurity is being stressed more. Professor Kim Hak-ju pointed out that the risk of hacking rises along the intermediate paths when distributed data centers are linked through networks.

Professor Kim said, "The problem is that when these modular data centers are connected, there is a concern about hacking, in between. So cybersecurity becomes very important." In the AI era, the importance of information security is becoming more prominent alongside the physical spread of data centers.

As physical investment in artificial intelligence (AI) advances rapidly, a warning has been raised that investment in cybersecurity is urgently needed. Professor Kim Hak-ju stressed that if AI robots and the like malfunction due to hacking, the consequences could be catastrophic, so cybersecurity investment is essential.

Professor Kim said, "We have invested heavily in AI and now built a lot of physical AI. But if cybersecurity doesn't work, you can't use it. What if a robot starts behaving strangely? Wouldn't that be a disaster?" He stressed that rather than slowing physical AI investment, cybersecurity should be prioritized.

He particularly rated Palo Alto Networks and CrowdStrike as the best in cybersecurity. Professor Kim said, "Palo Alto Networks and CrowdStrike are the best. Both are successful cybersecurity companies and will become more important going forward. They aren't making money yet, but from now on they will explode."

As AI technology grows more advanced, hacking may become easier and its impact greater, so the role of cybersecurity technology is expected to become even more important. These companies are assessed as having the potential to grow while protecting the core infrastructure of the AI era.

Professor Kim said, "Both of these still lack technical maturity. The reason is that the government needs to invest. Only then can it settle through trial and error." Comparing this with China's government aggressively pushing the technology, he expressed regret that former President Trump seems to be standing by too much. Such a lack of investment has raised concerns that it could negatively affect U.S. competitiveness in advanced technology over the long term.

Professor Kim Hak-ju analyzed that the choice of gas power for immediate supply reflects a judgment that building infrastructure would take too long.

Professor Kim said, "Small reactors and the hydrogen economy are good alternatives that could solve both power and water at once, so why not do that? Lately they're just putting gas power next to data centers. And its economics are very weak." This means it is not sustainable in the long run and is also problematic in terms of efficiency.

Professor Kim stressed that because the urgent need seems to make infrastructure a long wait, they are going this route just to win for now, but the ultimate solution requires the government to lay this infrastructure so that it can be developed further, stressing the importance of active government investment and infrastructure buildout. SMRs and the hydrogen economy are expected to become key infrastructure for sustainable development in the AI era, as innovative alternatives that can supply power and water stably at the same time.

Professor Kim Hak-ju expressed skepticism about the possibility that Emerson, a company, anticipated the data center era and concentrated its core capabilities accordingly. Professor Kim analyzed that Emerson's current position is more likely an accidental result than a deliberate strategy.

Professor Kim said, "Did this company Emerson concentrate its capabilities because it predicted that the era of data centers was coming? I don't think so. It just happened as they went along." He added that if Emerson had accurately predicted the arrival of the data center era, it would not have sold some business units to a private equity fund in 2016. This suggests that the success of many companies may come from adapting to market changes and chance opportunities as well as strategic foresight.

Professor Kim Hak-ju explained that through this acquisition, Vertiv has secured a dominant position in the data center hardware market.

Professor Kim said, "With the acquisition of EN Engineering as well, Vertiv has now truly become the number one in data center hardware." He stressed that Vertiv's solutions will be nearly essential for newly built AI data centers. Professor Kim added, "There's no way they wouldn't use it," highly praising Vertiv's technological strength and market dominance.

This suggests that as building AI data centers becomes essential, Vertiv's hardware solutions will hold a strong competitive advantage in the market. Vertiv, which has core infrastructure technologies such as data center cooling and power management, is expected to emerge as a direct beneficiary of AI industry growth.

This is drawing attention as an innovative attempt to solve power and cooling problems on Earth. Professor Kim Hak-ju pointed to the potential of space-based solar power, explaining that sunlight that has not passed through the atmosphere contains much high-wavelength energy, meaning short wavelengths and high energy.

The idea is that capturing this high energy with materials that have a large band gap, able to withstand high voltage, could produce energy far more efficiently than on Earth. Professor Kim said, "Take the U.S. or SpaceX, for example. Let's build data centers in space. Sunlight in space hasn't passed through the atmosphere, so it has a lot of high-wavelength energy: short wavelengths and high energy. If we use materials with a large band gap that can withstand high voltage to capture that high energy, we can use much better energy. Solar power generation seems like it would work well. Solar in space is good."

Such space data center plans could also help solve cooling problems, since using the low temperatures of space could significantly reduce cooling costs. However, this is still an early-stage concept, and many technical and economic challenges remain to be solved.

Professor Kim Hak-ju noted that the construction and operating costs of space data centers are currently very high, raising doubts about feasibility.

Professor Kim said, "As you said, the economics still don't work out, but if the solutions advance a bit more, it could become possible." This means that while space data centers are promising in the long term, securing economic viability through technological progress and cost reduction is the prerequisite. Various technological innovations, such as improving solar power efficiency and reducing rocket launch costs, will likely need to accompany this before space data centers can become a reality.

While the U.S. and SpaceX are considering building data centers in space, China is exploring building them in the deep sea, a contrast in approaches between the two countries. Professor Kim Hak-ju cited one reason China wants to build deep-sea data centers near the coast as using electricity produced by offshore wind power.

Professor Kim said, "The U.S. and SpaceX want to put data centers in space. China, on the other hand, might just build them in the deep sea near the coast?" He explained China's intent to supply data centers with electricity that would otherwise be wasted from offshore wind power and to use abundant seawater for cooling. However, Kim expressed skepticism, pointing out technical challenges such as corrosion in deep-sea environments, barnacle attachment, sealing problems, and difficulty in maintenance when parts fail.

To address these problems, China appears to be considering floating offshore structures, in the form of artificial islands, for its data centers. However, Professor Kim added that even this could face technical problems, such as the risk of damage to subsea cables when drawing power from the shore. Ultimately, both space and the deep sea have clear pros and cons as data center sites, and their technical and economic feasibility requires further verification.

Professor Kim Hak-ju stressed that companies overcoming the limits of AI infrastructure may be more attractive to investors than companies like OpenAI at the frontline of AI.

Professor Kim said, "Solving the bottleneck is the most important thing, rather than something like OpenAI at the very end of AI." He disclosed that he is investing in companies like NuScale Power and Plug Power. These companies offer solutions that reliably supply the power and water AI data centers need, through small modular reactor (SMR) and hydrogen energy technologies.

Professor Kim analyzed that considering individual investors do not have long investment horizons, the stocks of companies that resolve the bottlenecks now holding back AI computing capacity are likely to move first. This means that rather than waiting for the final stage of long-term technological development, companies solving the fundamental problems now holding back the AI industry will offer short-term investment opportunities.