저자: LarryFink및진행자Dave/Marley등패널 · 외국인
요약 보기
AI 데이터센터 투자 규모(GW당 500~600억달러)로 전력 부족이 병목이며, 하이퍼스케일러 대신 하드웨어·네트워킹·냉각·보안 등 공급자 선호를 제시. Oracle은 레버리지 확대·현금 악화·웹트래픽 감소로 약세이며 SaaS 취약성 지적. Intel은 안정적 이익 확인 전까지 관망. 엔비디아의 자금 지원은 수요 약화가 아닌 강한 수요의 반영이라는 주장과, AI 저비용화가 오히려 컴퓨팅 수요를 키운다는 논리가 제시됨.
다듬은 글
testing your networks is fast enough. We don't have enough electricity, and that's what's causing some problems at the state level, where there's concern about electricity prices for consumers, which is a valid concern. But other states that have more energy and the ability to supply it, without increasing the price for the consumer, will be the main winners in economic growth.
Yes. So, last quarter we contracted about 1 gigawatt of capacity in Pennsylvania. OK. We are funding another large data center in another state, and in my conversations with all the hyperscalers, the demand is outstripping supply. And yes, you are right. These large companies were not previously burdened with large assets on their balance sheets, and now their business has changed, and that is the role of capital markets.
I promise you that this will be a great investment for individuals if we can, but you have to understand that the basic business model that we will eventually arrive at will generate more than enough revenue and profit to pay for all the data centers. They are not cheap. Yes. That's $50–60 billion per gigawatt.
What worries me, Dave, is not the demand. I worry about whether all of society will be able to benefit from AI, because computing is very expensive right now. My question to everyone in the hyperscaler business is not whether they have enough demand, but how quickly they can bring the cost of computing down.
I'm not worried about BlackRock paying for AI, and I think you know we've been a huge investor in that, and our systemic equities team has really benefited from the flows and the returns. You know, AI development is much more than just data centers and chips, so we took into account everything inside this data center.
This is hardware, networking, cybersecurity, cooling, as mentioned earlier. I think this is the best way to deal with what is happening in the market. I think hyperscalers will spend this money. I just don't know if they're going to make the profit that everyone expects, or what people would call a reasonable profit.
So in this environment, I would probably redirect funds from hyperscalers to those who provide all the hardware that will make AI a reality in the future. And, you know, most people don't notice a lot of things . For example, cybersecurity is very important. These things should be real cyber fortresses.
Regarding cooling: almost half of the energy costs in a data center go to cooling systems, because if you've ever charged your iPhone, you know that it's a little warm after charging. You know, imagine a building three football fields long, filled with racks of servers, and how much heat it all generates.
So they have to keep that equipment cool, and I don't think anyone has really realized that part of the problem, so to speak. Therefore, we prefer to own those who will be the recipients in this transaction. We would probably take money from those who spend and let them prove later whether they can actually monetize those large expenditures.
We've talked a lot about check writers and check cashers and whether they'll start to converge, but it does n't look like that's going to happen this quarter. Michael, looking at Intel, you both made it pretty clear that you are not actively investing in it right now and are not putting more money there.
What does Intel have to prove to change your mind to a more constructive one? I would like to see more stable profits . I mean, this quarter, as you mentioned at the beginning — by the way, it was a great summary, Marley. It's only one quarter, we missed this trend. I don't think anyone could have predicted how sharply they would go up, say , from March 30 or April 1.
I would like to see more stable profits. Remember, there are no dividends. So why should I buy these stocks? Our question is always : why buy these stocks now? This is not the best time, as I already mentioned, but I would like to see more stable profits . Basically, there has been no growth in revenue lately and the number of site visits has fallen.
The shares fell even more. Tell us a little more about that. Yes, I'll put it this way: we're talking about web demand, and when we say "web demand," we mean their corporate website. So, we measure the interest of corporate clients, which obviously correlates quite well with revenues. And you know, Oracle is a crazy story.
If you go back to September, I think the stock went from 240 to 340 on the earnings reports, and since then it's just been in free fall, down about two-thirds; We are now at annual lows, as you said, 116.18. Um, and this is the same data I'm talking about: enterprise-level website traffic , the yellow line, is down about 11% year-over-year.
Um, and stocks have obviously fallen much more in the year. They definitely have difficulties. But you know, Oracle is playing a leveraged game , and everyone is probably a little worried that they're going all in. Um , and you know, they're getting a little tight with cash. Debt is rising , credit ratings are falling.
Um, they're, you know, putting everything on the line, and they have a lot of future commitments. They have significant future revenues, ahem, and significant future expenses waiting for them. This all looks like a total gamble on their part. And I don't think the market is taking it too well, frankly.
Um, you know, if you look at companies like Oracle, a significant portion of their business is software as a service, not all of it, but about 21%, and that's all extremely vulnerable . I mean, we're talking about HR, supply chain, CRM, planning, finance, project management, all of which are very vulnerable to the development of AI software.
The discussion around artificial intelligence is starting to change. For most of the past two years, investors have been asking one simple question. How much will the world's largest technology companies spend on artificial intelligence? Now another question comes to the fore. Will the industry be able to continue to finance such an extraordinary level of investment?
This discussion was sparked by the announcement that Nvidia could help fund one of OpenAI's massive data center projects in the US. At first glance, the headline was disturbing. Some investors immediately wondered whether the AI ecosystem had become dependent on chipmakers helping customers finance infrastructure.
But others saw the situation quite differently. Their argument is that Nvidia is not intervening because demand is weak. She is intervening because demand remains extremely strong. The AI industry has reached a point where infrastructure projects have become so large that traditional funding is not always enough.
This does not necessarily mean a slowdown in demand. Instead, it reflects the sheer scale of what the companies are trying to build. Artificial intelligence is no longer measured in millions of dollars. It is measured in hundreds of billions, and over time, financing becomes part of a competitive strategy.
Nvidia has an initiative to ensure the further movement of the largest infrastructure projects in the field of AI. Every new data center means future demand for GPUs, networking equipment, software, and artificial intelligence systems. Helping customers secure financing today can ultimately support equipment deployments for years to come.
From Nvidia's perspective, protecting the ecosystem can also protect its own long-term development. This does not mean that investors should ignore risks. Questions about return on investment remain relevant. Capital-intensive industries must ultimately prove that huge expenditures create equally impressive economic value.
But for now, the largest cloud providers continue to consider artificial intelligence their highest strategic priority. Instead of reducing investment, they continue to expand infrastructure. Demand for cloud services remains high, and enterprise adoption continues to accelerate. Developers continue to create larger and more powerful AI applications, and these trends indicate that demand remains stable.
Another noteworthy development is the rapid growth of open artificial intelligence. Over the past few months, Chinese developers have introduced increasingly powerful AI models at significantly lower costs. Initially, many investors perceived these releases as a threat to companies like Nvidia. They say that if AI becomes cheaper, then fewer graphics processors will be needed.
But this assumption may oversimplify what is actually happening. Cheaper models do not eliminate the demand for computing power. They are expanding access to artificial intelligence. Instead of a few companies using expensive cutting-edge models, thousands of organizations can start implementing AI in everyday applications.
Each additional user creates new requests for processing. Each new application generates additional load. Each new workload requires computing resources. Cheaper intelligence often leads to greater intelligence consumption. Technology has reportedly always followed this pattern. Cheaper access to the Internet has created more Internet users.
Cheaper cloud computing has created more software, and more affordable smartphones have expanded the scope of mobile applications. Artificial intelligence seems to be following the same path. Competition at the software level can be
Comments 0