원문 보기
We never want to view something as temporary in terms of inflationary pressures and then be caught off guard. Uh , you're absolutely right. I mean, Powell's "temporary", you know, to describe inflation...that word really became so he said, "let's take it out." So this fear...but there is a right moment.
I mean , you know, let's hope this war ends. Oil will return to a lower price. about the oil glut going back to 60. So that's honestly a much better story about " temporality" than anything during Covid, when the government was spending trillions of dollars and the Fed was, I think, recklessly just expanding the money supply.
So, there are some reasons. And if you take what's called long-term expectations—the difference between inflation-protected bonds and regular bonds—we didn't see such a significant gap. What the Fed calls the “5- year forward 5 years from now,” that is, the forecast after 5 years, which is cyclical and distant, did not show a significant increase in inflation expectations.
Now, if Warsh wants to continue to hold positions and argue this, I'm sure he'll mention it. But there are a lot of people who see inflation in the fundamentals, other than energy, and I think they'll say, " Listen, I think it's time to raise the rate by at least a quarter of a point." So Wednesday is going to be a really important day, in addition to all these earnings reports that, of course, come out every day.
How much do you... F recently rose to a 52- week high, right? In March it was $69, now it's $121—this is a 52-week high. A remarkable leap: 37% growth in 3 months, 60% in 6 months. So this ETF HACK, representing cybersecurity, that you just talked about, the names of cybersecurity companies, I know you mentioned a few, and I want to continue the conversation and look at some of the IPOs that we're expecting.
But do you like all these cybersecurity companies? Some, and I don't mean stocks, but your opinion as a person in technology: do you think there are clear leaders out there? They, they have clear leaders. They also have transparency, and that's what we saw at OpenAI, about a week after Hugging Face thought, " It looks like there's an agent here, " and they said, " Yes, there was." I also heard a podcast about a week after OpenAI where Sam said, " Hey, this is what happened." He used the word "visceral" to describe his reaction, and I really applaud that, because that's what a leader like that should feel like—to be moved by it and truly ready to take action.
I see this with Palo Alto Networks Crowd. Strike were very transparent a few years ago when they had a huge problem. Vector AI is also very transparent, and they help companies understand the paradox I was talking about: AI for good, AI for bad. It must be the same . But I don't think these statements will derail the IPO .
However, I believe they have changed the course of the conversation, the due diligence process in this conversation. Because now Anthropic and OpenAI investors will ask not only about revenue growth or computing costs. They have to ask: how do you restrain your agents, right? How quickly can you detect and stop the behavior you are responsible for if something goes wrong?
Because, as we see, agents are gaining more and more autonomy. Security becomes a matter of business and evaluation. One serious public failure is, you know, a blow to reputation and brand. Therefore, control in companies engaged in advanced AI should become, in my opinion, a key indicator for investment.
Yes. I wanted to ask you about IPOs, but first let me give you one statistic from a few months ago. I'm sure these numbers have since been updated, but they show a trend: AI-enabled hacks increased by 89%, and cloud attacks jumped by 266%. And this focused on many companies in the cybersecurity space.
That's why this article was written back in March. But it really shows that these threats continue and how important cybersecurity is. What about the Anthropic and OpenAI IPOs? OpenAI may even be delayed until 2027. Does this affect how they can conduct their roadshows, or is anything changing in the IPO space for such companies?
I think it does, and I think it should be that way. I think, especially for sensitive industries like banking, as I said, the word that Sam Altman used to describe his reaction last month was "visceral," and as I said, I applaud that , because we need our leaders to be upfront with us. AI is everywhere.
It cannot be avoided. We use it every day. But we need to know that companies, whether they are preparing for an IPO or not, are real. They are transparent, they are accountable, and they declare this to potential investors, shareholders, and citizens. This message, I think , should be as fast as the agents' capabilities , as well as the control over them.
It must be delivered. It must be clear and truthful. And that should change when we see these results in the future. I don't think these companies have a choice. Yes, of course. I mean, we just saw a consortium of six major institutional investors raise $500 billion for Nvidia. Now Micron is also trying to raise funds.
What are your thoughts on some of these big tech companies in general and the tech trend? One last quick thought, I see three things here. First, AI capabilities are evolving much faster than deterrence, because, as I said, agents don't just answer questions. They can receive data. They can operate with minimal human involvement.
Second, the word "outcast" in relation to people is the movie "Star Wars." It doesn't mean " malicious" as you said. Um, and Nvidia is becoming the bank for the AI boom, and the latest moves are changing how investors should think about the company, because Nvidia doesn't just sell GPUs. It is increasingly helping to fund the infrastructure that will use these GPUs, and the latest example is huge: it is reportedly ready to provide up to $15 billion in support for a massive AI data center campus in Ohio that will be used by OpenAI.
The project could eventually reach a capacity of around 8 gigawatts, with the first phase expected to be operational in 2028. But here's the part that really matters, and simply supplying chips for the project helps solve one of the AI industry's biggest problems. How to finance infrastructure development that costs hundreds of billions of dollars?
Think about everything you need. Land, energy, data transfer, infrastructure, cooling, construction, networks and , ultimately, a huge number of GPUs. The need for capital is staggering, which is why Nvidia's strategy is becoming increasingly interesting. Nvidia previously announced partnerships with major financial firms including BlackRock, Blackstone, Goldman Sachs, KKR, Apollo, and Brookfield to help raise more than $500 billion in capital for AI infrastructure.
Now you see that the same funding strategy has appeared at the project level. Nvidia helped make this infrastructure possible. This infrastructure runs on Nvidia computing systems, and these systems generate revenue for operators. This creates a powerful economic flywheel. More funding, more data centers, more GPUs, more computing power, and more AI users.
But this is where investors should slow down, because the same mechanism that can accelerate the AI boom could ultimately accelerate the AI crash. Imagine billions of dollars pouring into data centers because everyone expects demand to continue to skyrocket. And then imagine that demand grows more slowly than expected.
Suddenly you find yourself with expensive GPU power sitting idle. And it's a concept that some investors are calling "dark GPUs." This has a historical parallel to the internet era, when a huge amount of fiber-optic infrastructure was built, which ultimately far exceeded short-term demand. The technology was real, the infrastructure was real, but the timing for the investment was wrong.
This is a risk that investors in Nvidia should understand. The question is not whether there will be a demand for AI. He obviously exists. The question is whether future demand for AI will be large enough to justify all the infrastructure being funded today, and there are several indicators to keep an eye on .
GPU rental prices, data center utilization, hyperscaler capital expenditures , AI model deployments, and most importantly, the return on every new dollar invested in AI infrastructure. If GPU rental prices remain high while new capacity continues to emerge, this is a very encouraging signal. If rental prices start to fall sharply and capacity continues to increase , investors should ask themselves whether supply is catching up with demand too quickly. There is also another huge bottleneck , and that is electricity.
Comments 0