저자: JeremySiegel/Brett,Sam,Joe등다수출연 · 외국인
요약 보기
AI 설비투자는 여전히 긍정적이며 하이퍼스케일러 지출이 수익성 높은 기업으로 이어질 것이라는 낙관론. 컴퓨팅 수요는 2030년까지 수조 달러 규모로 예상되고, 로보택시·휴머노이드 확산으로 경제활동이 크게 늘 것이라는 주장. 마이크론은 100억 달러 규모 메모리 연구 투자로 메모리를 전략 인프라로 격상시키려 하며, 메모리 사이클 종료 여부는 투자자가 신중히 봐야 한다는 논조.
다듬은 글
I mean, we have a long history that shows that companies that overinvest or simply invest more capital than others don't have the same stock returns as those that allocate their resources very carefully. But again , as I said, the hyperscalers, all the providers and disk manufacturers are making money now, but the next revolution is all those who will be able to do what used to cost so much for a fraction of the cost.
This is the group that, in my opinion, can achieve rapid growth in profits and margins. And I think we're seeing this rotation already now. You know, non-tech stocks , so to speak, have held up exceptionally well. And as you said at the beginning, it's really amazing what's happening in the Middle East and oil isn't worth 120.
Well-off, I think, if we just let some of these ships go and something else. And given the prospects of AI, I am still very optimistic about the market. But I think there is a reassessment of how much capital investment, the so-called " arms race," will benefit companies that are spending so many billions.
Professor Siegel, you are right that capital spending became a toxic term last week, right after we heard Alphabet's report and saw the stock fall due to their spending volume. But I started thinking about what would happen if some of these hyperscalers actually said, " You know, we're not going to spend that much money." I think it would be much more disastrous for the market at this point if they all said, "We're not going to spend anywhere near what we were talking about before." Well, it depends on how they phrase it.
Either because they see a decline in demand, or because they have reached such a level of efficiency that they can ensure an increase in the supply of computing power, for which there is such high demand. This is positive. If they say they see a drop in demand, that's negative. So, I think it all depends on how they justify the reason for the investment.
I don't know. Google said they are spending more because their demand is endless. Yes, well, that's positive. They were punished anyway. This is still a positive point. So the question is, can we make a breakthrough in chips or technology so that we don't have to spend so much money to provide the computing power we need?
I think this is a huge plus for society and for everyone in general. What, what, what if AI replaces all lawyers? I mean, what will this do for the stock market and society in general, Professor? Can we double down Joe, I've heard your comments about all this advertising everywhere. It worries me as much as it does you.
That is, you know, if you can't earn money through hard work, then we'll just sue someone to make money. And I've seen very few articles about it. What, what explains this explosion in the way this country's wealth is distributed. 21% or a little more is the same penetration rate as the Internet in 1996.
This was 2 years before Google was even founded and registered. And 1–5% penetration is where we were in 1990 during the Internet boom . That is, before people even knew about the browser. So we believe we are at a very early stage in this cycle. And as more users connect to AI and intelligent workers learn to use agents for their needs, we will see a huge demand for computing resources.
In fact, we expect trillions of dollars in demand by 2030. And this will stimulate demand for Nvidia and AMD chips, as well as AWS and CoreWeave services. I think people are really underestimating the scale of what's going to happen. Brett, since we're getting inflation data tomorrow, and Sam and I have been discussing inflation all day, especially in the context of the Fed, I'd like to hear your thoughts: do you see AI and AI productivity as a deflationary or inflationary factor ?
Of course. I mean , in intellectual work, people are obviously using AI to do more. I think in the realm of physical implementation, for example, robo-taxis and humanoid robots, it will clearly reduce the cost of doing certain things. Just moving around the city. We believe that after implementation, in monetary terms, the cost of providing robotaxi services will be lower than the marginal cost of driving your own car, which you have already purchased and paid for in full.
So, you will be willing to pay more because you save time, but this means that the cost of moving from place to place will decrease significantly. You might think, "Hey, this should curb demand." But that's not true, because when the cost of something goes down, especially in technology, people start using it much more often.
Think of situations when your child wants to go to the movies or visit a friend, and you say, “I can’t take you that far .” I'm not going to do that." If you can put your child in a robotaxi for a small price per mile , you'll be happy to do it. Therefore, we believe that there will be an explosive growth in economic activity associated with this reduction in costs.
Brett, thank you very much . The AI boom is causing unexpected changes in the semiconductor industry. For years, the industry has considered memory one of the most cyclical parts of the chip market. Now, that narrative is being challenged, and Micron is making a huge bet that memory will not just support the AI revolution, but could become one of its most important foundations.
Micron is investing $10 billion in a new research initiative in Boise focused on the future of memory and AI. The goal is to bring together Micron researchers, customers, universities, startups, and other technology partners to work on the next generation of memory technologies. And time is very important because AI creates memory problems that the industry has underestimated.
Why has memory suddenly become so important? When people think of AI hardware, they usually think of Nvidia GPUs. But GPUs don't work in isolation. AI systems need memory to store and process the vast amounts of information generated by increasingly complex models . As AI becomes more powerful, the requirements are constantly changing.
We need more memory, faster memory, higher bandwidth, lower power consumption, and more storage. This includes technologies such as HBM, DRAM, and SSD. And that's why Micron's CEO makes a much stronger argument. Memory is not just another component in a holistic AI system. It is becoming a strategic infrastructure for AI.
Micron's $10 billion research investment is particularly interesting because it's not just about increasing production next quarter. It's about long-term innovation. The company plans to focus research on areas such as materials, semiconductors, memory, and computing , bringing together different parts of the ecosystem.
This is important because some of the greatest technological breakthroughs in history have come from long-term research programs. Micron is essentially trying to create a dedicated research engine for the next generation of memory technologies. And this is happening in the United States. Why Boise is so important.
Micron has deep roots in Boise. The company began operations there nearly five decades ago and has continued to invest in memory research even during severe downturns. This is important because semiconductor production has become a strategic issue for the United States. America needs more domestically produced semiconductor products.
Yes, more computing power is needed, and they require huge amounts of memory. So Micron's investment is not just a story of corporate expansion. It's also part of a broader race to build America's semiconductor infrastructure. But is the memory market really no longer cyclical? This is where investors should be careful.
Micron CEO claims that memory has evolved from a commodity component to strategic infrastructure. There are strong reasons behind this argument. Our AI servers need more memory, and data centers are getting bigger. AI workloads are becoming more complex.
Comments 0