원문 보기
Wisdom comes, and I consider you wise. You and I have seen a lot in our careers, and we have both seen significantly higher interest rates. So I'll start with what I understand: in relative terms, when things change from low levels, it's a much larger percentage move than from high levels. But does this really threaten to stop the AI boom at this stage due to the inability to finance at current rates for 10, 2, or 30 years?
Isn't there too much concern about this right now ? I can't believe that the Ministry of Finance decided to take this step. It's almost like, "Wow, we noticed this and we're worried." I would ignore it at those levels. First of all, it's nice to see all three of you together again. Oh, exactly. Thank you.
Yes. We called it the "wild three," and we don't put anything bad into it . But I completely agree: I am not a fan of " operation twist". I did not consider the markets to be unmanageable or in need of additional liquidity. And I don't think I completely agree with you, Joe. We were already at 4 and 3/4 on the 10-year bond before.
This is not a critical point. I'm not saying we can't get to it in the future. Of course, in the future, but this is not the time to panic about funding. As you said, years from now, I would say 40 or 30 years ago, we would have exclaimed, "My God, how low interest rates are now compared to what we've been through." But I think there's a performance scheduled for Friday at 10:00.
I believe this is a very important event. I really think the market wants the Fed to not give forecasts. They are ready to accept it. They want to know what criteria you look at? Do you look at 5- year inflation expectations? Are you looking at the futures market for the Fed rate? What criteria are important to you?
And I believe that this is the proper function of the Fed—to reveal this to the market. And if Warsh does that, I think we could be in for a big boost on Friday. I think, for example, on Wednesday we'll get the PCE report, and I think it's going to be good. In fact, I think the figures will be at or even below expectations.
So, you know, I don't think there will be any significant pressure. Um, let's face it , from a political perspective. I don't see them raising rates. Is it true, or is it just an urban legend, that the market by definition tests every new Fed chairman? And if I did that, if I were the market and I was being sneaky, I would continue to grow to see if...Is this how I...Is this what we're seeing?
Is the market testing Kevin Warsh right now, creating Well, I think that...I think if Kevin disappoints on Friday and gives some, you know, bland, uninformative speech, the market might start testing again. Again , there is always a chance for it to correct itself at the Fed meeting on September 16th, which is important because of SCP forecasts and whatnot; but there could be a bit of a test if he disappoints on Friday.
So, yes, to some extent, and maybe you could even say there was a tiny mini-test, you know, when the 30-year bond rose above the 5% mark. So, I think for, again , for this Friday, uh, the speech is what we're watching. Hey Jeremy, when you say "if he disappoints the market," how exactly can he disappoint the market?
I think if he, you know, doesn't give any criteria as to what he's considering, um, and just says, you know, "we're having great discussions" and " we'll let the market"... I think the important thing is whether he comments on Bessant and the intervention in the 30-year bond. As we all know, he was a big proponent of letting the market speak for itself.
And now, uh, we find out, you know, he...and there were a lot of questions about that. What was his reaction? What is his real reaction to Scott Bessant? Did they communicate? Was this a surprise? So, I...you know, I'm not sure if he's going to include it or not. And I... if I remember correctly, there is no question and answer session after the Jackson Hole speech.
So, um, but that would probably be the main question they would ask. The answer may vary, professor. I need clean bets. Well, you know, will I get net rates if the Fed changes its term structure at this particular moment? But I think what the market really needs to know is: you're not raising rates, you voted against it, okay.
Just what criteria do you take into account? Ahem. Therefore. Um, and there are a whole range of indicators that could justify that, and they would feel more confident. Here are the benchmarks we're looking at, and you're looking at, we can also keep an eye on this , and it won't... Nvidia is entering a new phase of the AI race, and specialized chips are becoming more common.
Specialized data output systems are evolving, and Nvidia's answer is surprisingly simple. Surpass them. Jensen Huang made it clear that Nvidia already has the largest infrastructure for AI. Its platform is used for closed models, open models, cloud services, and data centers. So, instead of trying to eliminate all specialized chips, Nvidia wants these chips to be plugged into its ecosystem.
This is exactly what we observe with D Matrix. The startup has developed a specialized output processor designed for ultra-low latency AI workloads. And now it is partnering with Nvidia to integrate its technology into the Nvidia platform via NVLink fusion. Why is this important? Because Nvidia already has a huge installed base.
Its equipment is deployed in cloud services and data centers around the world. For startups like D Matrix, rebuilding a data infrastructure from scratch would take years. With Nvidia, they can tap into an ecosystem that already exists. D Matrix has spent over 7 years developing memory- centric output technology .
And time is very important. AI agents and coding tools have dramatically increased the demand for rapid data inference. When you interact with an AI agent, low latency matters . The faster the response, the more useful the system becomes. This is where specialized output equipment can play an important role.
D Matrix's upcoming Raptor X Fuse takes a different approach to memory, instead of relying on HBM. The company is developing a 3D stack DRAM architecture that places memory much closer to the compute unit. The goal is to overcome one of the biggest limitations of AI: the memory barrier. AI systems constantly pump huge amounts of data between memory and the processor.
This creates issues with latency, power consumption, and cost. D-Matrix believes that their architectures will help solve these problems. The company expects its first integrated product to hit the market in about a year. The potential customer base is also quite broad. D-Matrix targets hyperscalers, advanced AI labs, government AI projects, inference clouds, and high-frequency trading firms.
So it's not necessarily about replacing Nvidia GPUs. It's about adding specialized computations alongside them. And this distinction is important. It seems that Nvidia understands well that AI will not always work on the same type of processor. Different workloads will require different architectures, and some will require Nvidia GPUs.
Others may benefit from specialized inference chips. So Nvidia's strategy seems to be that if specialized chips appear, they need to be connected to the Nvidia platform. This turns a potential competitive threat into an opportunity for the ecosystem. However, Nvidia's expansion into the specialized AI chip market is also attracting the attention of regulators. The Ministry of Justice has reportedly launched an investigation.
Comments 0