저자: JeremySiegel/영상전반AI인프라관련다수출연자—소속미상포함 · 외국인
요약 보기
결론: 금리 상승과 AI 수익화 의문이 맞물려 있으며, 기술주 중심 랠리는 금리 부담으로 둔화될 수 있다는 취지. 채권·금리 시장: 물가연동 국채(TIPS) 30년물 금리가 3.35%로 상승해 과거 20~30년 내 보기 드문 수준이라고 언급. 주식시장 PE 20 기준 실질 수익률 약 5%로 경쟁력이 있다고 평가. 연준 전망: 파월 연준이 연내 최소 두 차례 추가 인상이 필요하다는 견해. 다만 중간선거 6일 전이라 12월 50bp 인하 가능성도 거론. 2년물이 적정 연준금리 지표라는 주장. 기술주 vs 전통산업: 기술 섹터는 마진율 50~70%로 금리 인상 충격을 흡수하나, 마진 7~10%인 실물 산업은 금리 인상 시 이익이 크게 줄어든다는 주장. 상반기 기대된 순환매(로테이션)는 금리 인상으로 중단됐다고 평가. 팔란티어: 카프 CEO가 데이터 주권·보안 중심 전략 강조, Athena OS 중심 기업 운영체제 제시. 주가 연초 대비 65%, 1년 70% 상승 언급. 토큰 비용은 하락 전망이며 고정가격 모델 제시. 마이크론: 전년 이익 약 432% 증가, 순현금 약 683억 달러 언급. AI PC 교체 수요와 메모리 공급 부족으로 가격 결정력 확대 기대. 다만 주가는 이미 성장 기대가 반영됐다는 경고. 엔비디아: 피지컬 AI(로봇·자동화)로 수요 확장 가능성. 애널리스트 목표가 300달러 언급(장기, 투자 논리이지 보장 아님). 마이크로소프트 / 생태계: 클라우드·기업 소프트웨어 강점으로 AI 도입 수혜 가능하나, 인프라 투자 대비 고객 지불 의사 입증 필요. 사이버보안·네트워킹·메모리 업체 수혜 가능성 언급. SpaceX: 약 140억 달러 자금 조달 시도 보도, AI 인프라로의 사업 확장 가능성 언급. 리스크: 하이퍼스케일러 대규모 투자가 공급사 이익을 먼저 키우는 구조라 AI 기업의 실제 수익성 입증 여부가 핵심 변수.
다듬은 글
Nice to see you, Karl. Did you feel the start of a rally in the bond market today? Well, I was relieved that it happened because I just couldn't believe the returns I was seeing. In particular, I like to look at the yield on inflation-protected Treasury bonds, and the 30- year has reached 3 and one-third percent, 3.35%, and I don't think we've seen anything like that in the last 20, 25, 30 years.
And now, you know , is it competitive? When you get almost three and a half percent after inflation, a stock market with a P/E of 20 gives you 5% after inflation. So, the supply is still there, but it is dwindling. And I think one of the reasons why the tech sector is doing so well is that you look at their margins, the G7 is at 50%, 60%, 70%.
So, there is some increase in interest rates. But outside of technology, in the real world, where they're operating at 7, 8, 9, 10% margins, when you start raising those rates, it cuts into profits significantly. We all thought there was a great rotation going on in the first half of the year , and it really looked active, but I think this rate hike has put a stop to it.
Yes. Are the " bulls" right to take inspiration from Jefferson's words about needing more time, Carl, based on what we heard from Waller ? Yes, of course, Powell is now in a very difficult situation before the next meeting . This is 6 days before the midterm elections. Frankly, I think we need two more hikes this year, but he could say to the committee, " Listen, guys, let's wait, maybe do 50 basis points in December." Let's not politicize this by giving raises 6 days before the election.
Do you know what a difference a month or two makes for 25 basis points? So I think he'll have to maneuver skillfully again. But I mean, when you look at the futures market and some of the other indicators, it's the strength of the economy, and the rising yields just tell you that this economy can withstand rising short-term interest rates.
Yes, yes. A good article by Greg Ip in the magazine today, where he argues that nominal GDP of 6.3 with a regression to 10-year bonds at 5.5% is not a problem at all . Yes. Yes, and we typically think that 10- year bonds should be 80 basis points above the Fed rate. Well , you know, if you 're at 5, 5.25 or 5.5%, that's significantly higher than where we are now in terms of the normal slope of the curve.
Some say that 2-year bonds are a good indicator of where the Fed rate should be, and that's impressive. So it also...So, given all these term structure issues, it argues in favor of higher interest rates. with their data, Liz, and we take this very seriously. You know, Alex Karp was at our conference yesterday.
Palantir is one of our largest global partners , and we talked almost exclusively about data sovereignty and security. This is what we focus on above all else . When you look at Athena OS, it's really the first full-fledged enterprise operating system that uses data and artificial intelligence as a foundation so that the next decisions become smarter because of the previous ones.
How does inventory drive marketing? Then marketing manages the logistics. Logistics manages production in a way that makes everything real-time and makes every single decision smarter. You know, you made Alex Karp take off his Palantir branded clothes and put on Zeta Global clothes yesterday at your event.
Palantir and Zeta. When he came out, there were a thousand of these caps in the crowd. And it was a wonderful moment. You know, listen, we had the most amazing people in the world, but what really matters is whether we can turn this event into a huge boost to business and continue to grow our business very quickly and more profitably than before.
John, one of the things that businesses and companies are starting to get very smart and frugal about is the cost of computing. And they see that the computational cost to get some of these services is prohibitive. And when we say “computing,” we mean tokens. So how does Athena OS handle this situation and perhaps smooth out some of the inequities and costs?
Well, I think it's important to remember that the value of tokens is only a temporary phenomenon, but with the increasing number of alternatives that allow for the use of open technologies, the price of tokens will decrease significantly. So I don't think anyone should worry about token prices. The market will take care of everything.
We will get a token price that suits the market, and since there are strong open source alternatives, I believe the token issue will not become a big problem for the industry. I have to say, I know what you're going to say, David: that this new OS actually requires 80% fewer tokens. This is great. That makes it cheaper, right?
Well, actually in Athena OS we will have no token costs at all. We will use our own inference models. We will set fixed prices. So our clients will know exactly what their costs will be. John, as always, is absolutely right. Token costs will decrease significantly, but we believe that ultimately it should be a fixed amount, not a variable amount, so that customers can see what their costs will be.
David, John, keep going. I'll do it quickly. Yes, I just wanted to say that about 2 months ago, Jensen Huang said that the real internet economy is about to begin, and that's where we belong at Zeta Global. This means that the level of output that sits above these advanced models and the innovation that takes place in Zeta Global Athena is a big part of it.
And the huge advantage we have with our database, which gives us data on customer intent that no other company in the world has, and what David does with it strategically, makes a big impression on me as a co-founder of the company ; I watch his leadership, his achievements, and his work with leading companies like Palantir.
And this impresses investors. The stock has risen 65% year-to-date and 70% over the past year. Nice to see you both. Thank you very much. Thank you, Liz. Thank you, Liz. Today, AI has a money problem not because companies have stopped investing, nor because the demand for computing has disappeared. And because Wall Street is starting to ask a much more difficult question.
When will all these expenses turn into long-term, full-fledged profits ? And this question could determine the next major step for Nvidia , Micron, Microsoft, and the entire AI infrastructure market. Let's start with a company that investors may still be undervaluing. Micron, a memory chip maker, has already shown extraordinary growth: according to market surveys, profits last year amounted to approximately 432%.
And even after such a jump, some investors believe that the stock has the potential for further growth . Their argument is simple. , data centers require increasingly powerful computing systems supported by large amounts of advanced memory. When demand outstrips supply, producers gain a pricing advantage.
This can lead to increased revenue and potentially much higher profits. Micron could also benefit from another development—the next generation of AI-enabled personal computers. If consumers and businesses start replacing older machines with systems designed for AI workloads, it will create a new source of demand for memory.
The optimistic outlook also points to the company's financial position with approximately $68.3 billion in net cash, according to discussions. But there is a nuancould fall sharply even if the underlying business remains strong. Investors should distinguish between a great company and a stock whose future growth is already reflected in its price.
This discussion brings us to Nvidia. The company remains a central player in building AI infrastructure, and some analysts believe its future capabilities extend far beyond today's data center market. Consider physical AI. Robotics, automated machines, and other systems capable of interacting with the real world could create additional demand for advanced computing.
If these markets develop at scale, Nvidia's potential opportunities could expand significantly. One analyst discussed a price target of $300, suggesting it could be higher in the long term. This is an investment thesis, not a guaranteed result. Importantly , investors are increasingly valuing Nvidia based on what AI could become in the future, not just the product it is today.
And this is both a great opportunity and a risk. Now let's look at Microsoft. Its advantage comes from another part of the AI economy. Microsoft already has relationships with businesses around the world. It operates a large cloud platform, sells enterprise software, and has a multitude of ways to integrate AI into existing products.
As companies move from experimenting with AI to implementing it into their operations, Microsoft's cloud business could benefit. But investors ultimately need evidence that customers are willing to pay enough for these services to justify the infrastructure costs. This is the same challenge that every major AI company faces.
And this task becomes even more difficult when you consider the broader ecosystem. Cybersecurity firms stand to benefit as organizations deploy AI agents that access sensitive information and perform tasks. Enterprise software vendors can help businesses structure data and integrate AI into existing systems.
Networking and memory companies can benefit as computing infrastructure becomes more complex. The opportunities extend across the industry. However, not every company will receive the same share of the benefits. Some will have critical technologies, while some will supply interchangeable components.
Others may find it difficult to convert growing demand into sustainable profits. The next phase of this market could reward companies according to their true place in the AI economy, rather than simply for being in line with the trend. And then there's SpaceX. Market analysis reports point to an attempt to raise approximately $14 billion in funding tied to its broader ambitions.
The most interesting thing is the potential evolution of SpaceX beyond traditional business . AI infrastructure requires a huge amount of computing power, and companies with the capital, technology, and strategic resources to build it may find themselves in a completely different market. This does not mean that every ambitious project will succeed.
This means that the AI investment cyclnology sector. But the more capital is poured into AI, the more important one question becomes. Will the return justify these costs? Imagine a hyperscaler investing billions in a new data center. This investment creates immediate demand for chip manufacturers, memory, network vendors, and infrastructure providers.
Their profits may increase even before the investing company fully recoups its costs. This is one reason why AI vendors can report exponential growth while investors remain uncertain about the long-term profitability of the companies that buy their products.
Comments 0