저자: DanIves및진행자 · 외국인
요약 보기
AI 인프라 신흥 클라우드(CoreWeave·Nebius 등)의 부채 부담과 채권시장 공포를 다룬 영상. 결론: 채권시장의 공포는 과도하며, Dan Ives는 최근 하락을 수요 문제가 아닌 자금 조달 비용 문제로 보고 매수 기회로 봄. AI 인프라 구조: 하이퍼스케일러(마이크로소프트·아마존·구글)와 신흥 클라우드(CoreWeave·Nebius·IREN)가 데이터센터 GPU 용량을 임대하며, 신흥 클라우드는 차입으로 데이터센터를 짓는 구조. 채권시장 우려: 금리 상승으로 차입비용이 올라, CoreWeave의 CDS 프리미엄이 사상 최고치를 기록. 부도 보험 비용 기준 약 60% 이상 부도 확률 반영. CoreWeave 부채 약 500억 달러, 일부 채권은 액면가 80센트 수준으로 하락. 주가 영향: 10월 14일 CoreWeave 약 15%, IREN 약 12%, Nebius 약 10% 하락. 가트너는 2030년까지 신흥 클라우드 80% 이상이 독립 운영을 중단할 것으로 전망. 엔비디아 연결고리: 신흥 클라우드는 엔비디아 주요 고객이며 엔비디아가 일부 지분 보유(CoreWeave 포함). 이들이 자금을 못 구하면 GPU 구매가 줄어 엔비디아 매출에 타격. Dan Ives 시각: 시장이 '수요'와 '자금 비용'을 혼동하고 있음. 한국 방문 후 메모리 칩 수요가 공급의 약 13배라고 주장. 자금 비용 상승은 '과속 방지턱'이지 종착점이 아님. 젠슨 황도 신흥 클라우드 수요가 급증 중이라고 발언. 리스크 및 조건: 신흥 클라우드마다 재무 건전성 차이 존재. Morningstar에 따르면 Nebius는 순부채가 적은 반면 CoreWeave는 부채 의존도가 큼. 모든 회사가 같은 리스크를 지는 것은 아님. 출연자 전망: Dan Ives는 '건설 중인 라스베이거스 스트립' 비유를 들어 장기 낙관. 빅테크는 자리를 지키기 위해 비용을 감수할 것으로 봄. 관찰 포인트: 금리·채권수익률, 신흥 클라우드의 추가 자금 조달 능력(CDS 스프레드), 빅테크 실적과 설비투자 지속 여부.
다듬은 글
Do you remember about Neo Clouds? What is Neo CoreWeave, Nebius? Of course. They They, actually, what do they actually do? What do these companies do? You can think of them as renting space from a data center perspective. So, okay. So companies that eventually need to build their own data centers are turning to hyperscalers.
And hyperscalers can also turn to Neo Clouds. They are essentially no different, they rent space in data storage in these data centers. Of course. And then, so you have... you have...hyperscalers, what kind of role do they play? Yes, let's remember hyperscalers: Microsoft , Amazon, Google. These are well-known technology names, but what do they do?
Hyperscalers are, in essence, the foundation of infrastructure for all companies. So when they move to the cloud, when they get on the AI path, yeah. they do it through a hyperscaler anyway. So Neo Cloud is a rental... Tell me again. Neo Clouds are, in fact, part of the overall capacity. When you think about GPU power, data centers for companies that need to get data center space.
Yes. who ultimately need to ensure their workflows that they are building in the field of AI are good. they can turn to hyperscalers or Neo Clouds Yes. names like CoreWeave, Nebius. The same. They perform a similar function. perform a similar function in terms of capacity. Of course. Now here's the thing: if you look at last week, the big big problem was that CoreWeave Well, big these companies are also taking on a fair amount of debt Of course.
to build data centers. Of course. Oh, so CoreWeave is involved in the construction of data centers. And so if you look to CoreWeave to borrow money to build data centers, you're taking on debt. Debt is now more expensive than we expected, which is why it is under great pressure. But why were they singled out?
Is it because they had bad earnings reports? No, because CoreWeave and Nebius are two of the biggest players in Neo Cloud. Good. So, there is an opinion: well, it's good if they have a hundred billion order portfolio. Yes. and they take on the demand for GPUs on GPUs. And if they take on five billion, ten billion, fifteen billion, investors are essentially saying, " Look, they're taking on debt now.
The income they're getting here, yes. Investors are essentially saying, on the risk, yes. between that, yes. And there and there, I want to... I don't feel... I don't feel comfortable with this capital expenditure versus monetization. It's a story of capital expenditure versus monetization again. No doubt about it.
And they, they... Okay, but why are the shareholders of the company so impatient with this capital expenditure versus monetization? Actually, it's less the shareholders, more the bond market or the " bond avengers," or whatever you want to call them. Almost making shareholders nervous. Right. Because of what's happening with CDS spreads and bond yields .
That is, with what's happened with the 10-year, and because yields are going up, not down, attracting debt is getting more expensive. Raising debt, yes. And now, let's say, if you think about the percentage of debt issuance that has been historically recorded. Right. Historically, let's say 1%. Right.
Now it's 10%. Right. Okay. So, by definition, part of this expansion is debt-based. Right. And to me, it's... but the answer to that is we'll get from the hyperscalers and the tech companies reporting to confirm what's going on. Feelings are not facts. The general fear around all of this is not necessarily based on facts.
It's just a lot, it's just... But, but, but I was in Korea last week and I see that demand is increasing significantly. See. Okay. So to me, that's more important than... Well, you see, you see the real numbers. You see the real demand. So all this talk is just talk. Okay? Look, the point is that the demand for GPUs, the demand for building data centers is real.
The demand for GPUs is real . The fact that it's going to cost a little more money is a fact. The fact that it might take a little longer for people who invested in these companies to get their money back is a fact. It might take a little longer. Maybe it's true that capex versus monetization is a fact.
But investors are underestimating the scale and scope of the AI revolution. And it all comes down to this: it's like building the Las Vegas Strip in 1955. Nobody believed it. Right? When Steve Wynn came along, he had a [ __ ] bingo hall and he said, "I'm going to build this." "It's going to be a multi-billion dollar project." "Nobody believed it." But I go back to 1955.
In 1956, 1957, were there any problems with certain buildings? Maybe... Right. Right. Yes, of course there were. But the main thing is that there's only one strip. Last week, at its peak, if you wanted to insure $ 1 million of CoreWeave debt , it would cost you about $119,000 a year. That's not a typo.
The bond market was pricing in a 60%+ chance that one of Nvidia's biggest AI customers would default within five years. And if that sounds scary, it is. So, for the next few minutes, I'm going to explain all of this in plain language. What the neo-cloud is, why the bond market is panicking, and what it means if you own Nvidia stock or any AI stock.
And finally, I'm going to talk about one thing that Dan Ives says almost everyone overlooks. Once you see it, you can't " unsee" it. Let's start with something simple. To make AI work, you need graphics processing units ( GPUs), mostly from Nvidia. And GPUs need house—giant data centers, equipped with power, cooling, and networking.
So who’s building these houses? First, there are the hyperscalers: Microsoft, Amazon, and Google—the cloud giants. Most companies launch their technology on top of one of them. Then there are the new players—the neo-clouds, like Coreweave, Nebius, and IREN. They do one thing. They fill data centers with GPUs and rent that computing power to AI labs, big tech companies, and even the hyperscalers themselves.
Think of them as landlords, but instead of apartments, they rent out AI computing power. Okay, so far so good . So what’s the problem? The problem is how they pay for it. These data centers cost a fortune to build, and many neo-clouds pay with borrowed money. They borrow first and then slowly pay back over the years of the lease.
This works great when the loans cheap, but when interest rates go up, debt gets expensive. And when lenders get nervous, it gets even worse. That's what happened. Here's the twist that most people missed. Dan says this panic did n't start with investors in the stock market. He thinks it started in the bond market.
The people who lent these companies money started demanding more protection. And you can see it in something called CDS spreads. CDS stands for credit default swap. Fancy name , simple idea. It's essentially insurance against a company not paying back its debt. When that insurance goes up, lenders get scared.
And Coreweave's insurance has hit an all-time high. According to Reuters, Coreweave owed nearly $50 billion. Some of its bonds fell to 80 cents on the dollar, and on October 14, the stock fell about 15% in one day. IREN fell about 12%. Nebius fell about 10. Then Gartner’s research made matters worse.
It predicts that by 2030, more than 80% of neo- cloud providers will cease to operate independently. Oh, and it’s not just the small players. Reuters reports that default insurance for Amazon, Alphabet, and Meta also hit record highs, although it grew much more slowly. But wait, why should you care if you don’t own Coreweave stock?
Because of who Coreweave is buying from. Neo- clouds are some of Nvidia’s biggest customers. Nvidia even owns stakes in a few of them, including Coreweave. So here’s the fear in one sentence. If neo-clouds can’t raise money, they buy fewer GPUs, and if they buy fewer GPUs, Nvidia sells fewer. That’s why this one company’s bond has the entire AI market nervous.
Okay, now I want you to make a choice. This is the beginning of the AI meltdown, as 2008 for data centers, or is it just a temporary setback? Pick one. Record your choice. Keep that answer in mind, because here's what Dan Ives said. He called the selloff in Coreweave , IREN, and Nebius a buying opportunity, not a warning sign.
And that's the distinction I promised to show you. Dan says the market is confusing two very different things. One is the demand for AI computing. The other is the cost of paying for it. Demand, he says, is still growing rapidly. He just got back from a tour of Korea and says that demand for memory chips is about 13 times higher than supply .
It's not the demand that's changed. It's the price of money that's changed. Construction is getting more expensive, and it may take longer for investors to recoup their money. That's a reality, but he thinks it's a " speed bump ," not a dead end. In other words, feelings are not facts . So if you've chosen AI crash, you're not crazy.
That's what the headlines were screaming . But Dan thinks the headlines have confused a financial problem with a demand problem, and Jensen Huang seems to agree. He told CNBC that demand for neo-clouds is simply skyrocketing. He acknowledges that there is a capital shortage, but says that will be addressed.
Now, I don't want to sugarcoat anything. Even optimists will tell you that the AI boom is now more leveraged than it used to be. That's a real risk, and not every neo- cloud is built the same. Morningstar, for example, says that Nebius has a much stronger balance sheet with very little net debt. Coreweave relies much more heavily on debt.
So when the music stops, some of these companies may not be able to find a chair. But there's an old joke that optimists like to tell . Pessimists predicted 10 of the last two market downturns. Every scare in this market is called the end AI, and the construction is still ongoing. Which brings me to Dan's idea about the Las Vegas Strip.
He told Fortune that we'll remember this period as the construction of the Las Vegas Strip in 1955. His point is that back then, few people could have imagined what the Strip would become. There were problems along the way, but the Strip was one, and those who built early got the best spots. Dan sees AI the same way.
The tech giants will endure the costs now because they don't want to lose their place. So what's really worth watching? Three things. One: interest rates and bond yields. That's the price of all this construction. Two: can the neo-clouds keep raising money. Watch CDS spreads. If they go up, worry. If they stabilize, breathe.
Three: the earnings of big tech companies. If Microsoft, Amazon, and Google confirm that they're still spending money, the fear usually goes away. So, Let’s recap. What is a “neo-cloud”? An AI landlord that rents out GPUs is often built on borrowed money. Why did the bond market panic? Because the cost of CoreWeave’s debt insurance hit a record high, and lenders got scared.
Why should Nvidia investors worry? Because neo-clouds are big buyers of GPUs, and Nvidia owns stakes in some of them. And what’s the difference that Dan says everyone misses? Demand hasn’t fallen. The value of money has fallen. This is not financial advice. Do your own research. But here’s the main takeaway in one line.
Debt concerns are real, but they’re a reason to watch closely, not panic. If this helped you better understand the AI trade, subscribe. I’ll continue to dissect every twist and turn of this story. And next time, I’ll show you what the hyperscalers just revealed about their spending, and whether it will save or ruin neo-clouds. them.
Comments 0