원문 보기
Jim, you know, consistency will continue to be a key issue, as will how to achieve it. Jensen seems very confident. Yes. in that these models can be effectively aligned, so to speak, with what we call human behavior, or at least with what we all consider to be typical human behavior. I...You know, okay.
We'll see how the laboratories react. I think there's clearly no weakening in the idea that computers will continue to be an incredibly valuable asset that will be built and paid for. Even if you slow down or limit the training rate of advanced models, the inference of results will continue at an incredible speed.
So when it comes to the ecosystem, it looks like there's definitely no impact yet . Although Sarah Fryer did talk, you know, in your conversation with her, about the need for security and potential slowdown, right? Yes. I mean, listen, I think they think they can slow it down. They don't really talk about it .
This word probably doesn't quite fit. They can...They...Why do n't we listen to what you...I don't want to twist her words. We really need to take security seriously. And if that means we have to limit cutting-edge developments and slow down, we will definitely listen to our researchers and do that.
But as CFO, I have to make the right business decisions. And now, I think we at OpenAI are doing what we do well. We push these boundaries and take security and consistency very seriously. You mentioned that as a CFO you have to calculate resources. Do you think I know this is your boss...Call him boss.
Is it advisable to hold an IPO now? And I understand that. No wonder . But what would you cut costs on? What would you spend less on if you wanted to slow down processes? So let's talk about this slowdown, because what really fascinates me is that even if we stopped today, the amount of intelligence available in the world is enormous.
So, Astra, as you know, is now the most advanced, most intelligent and most coherent model. And it sells the fastest. Let me ask, even though it costs $100. And these are the fastest sales we've definitely seen. Of everything. Yes. And what does that mean? This means that intelligence is available to institutions like Boston Children's Hospital.
40 undiagnosed, unsolved diseases have now been diagnosed by clinicians there. If you are the parent of a child with this condition, it is life-changing. Um, if you're in the investment business, Valley Asset Management has a central bank analyst who can quickly take all the input from people like our dear friend Kevin Warsh and reduce a 2-hour synthesis to minutes.
So these things are changing the lives of people everywhere. And this is what is available to us today. So, going back to your question, certainly if we need to accelerate things, we will always make investment decisions based on high ROI . But from where I am now , there are so many opportunities to drive growth that I'm still very focused on getting more computing power to keep that flywheel turning.
Yeah, look, I feel like people think like there's a huge gap between what Jensen says and, say, what Sarah says. Um, I think there 's a little more consistency, so to speak, in the idea that look, we have to be careful, we have to have safeguards. We have cybersecurity. Yesterday I had George Kurtz from CrowdStrike.
I had Todd McKinnon from Okta, who does identity. Me, and I think I have Nikesh Arora from Palo Alto. I try to talk to each of them. I feel more at ease knowing that we have it under control when it comes to people who know what they are doing. I think Jensen made a very interesting point where he says, "Look, the people who are speaking out about this, they're not...Some of them are not researchers.
They don't really know what they're doing." David, there's a belief that maybe the situation is out of control , but then you have... You know, you have a swarm of 1,200. A swarm of 1,200. I don't know, Jim. I mean, listen, I shouldn't disagree with Jensen Huang, who has obviously forgotten more about this than I will ever know, but in doing so, the models are becoming more powerful.
We know this. I know that. And they're getting more powerful at an even faster rate, and we're getting closer to recursive self-improvement, which means that models can essentially create their own next generation. And when that happens, do you know how much control people will have? So these are important questions, and this is an important point.
Jim, I wanted to quickly jump to the story related to OpenAI regarding this potential new capital raise. I know, you guys seem to have recorded your interview a little before any news broke about it, or maybe it's not even news. I heard about it from OpenAI investors, actually, late last week, early this week.
I had the information and decided not to report it because it didn't meet my criteria, because it seems like OpenAI is not doing anything right now . Does this mean that their investors, some of the big ones, wouldn't like to see a higher valuation? Of course, we would like to. Is it possible that once OpenAI has a realistic timeline for an IPO and estimates how long it will take to get there , they will want to raise capital at a higher valuation, perhaps 1.2, or maybe even 1.5 if they can?
Yes, but that doesn't mean, Jim, that they're actively doing it right now. So others have reported this possibility, I chose not to, but I certainly wanted to provide some context. By the way, Before we start, subscribe to the channel. And if you want to ge, join our membership and get access to our premium analysis.
Now let 's get down to business. Micron is about to release its earnings report, and it could be one of the most important reports in the history of AI memory. Why? Because the amount of memory required for next-generation AI systems may be much larger than many investors currently realize. Some analysts even compare the memory needs of advanced robots to autonomous vehicles, arguing that robots may need significantly more memory as AI becomes increasingly complex.
And if physical AI becomes a major growth market, it could create another huge wave of demand for memory. So Micron's story is no longer just about smartphones, PCs, or traditional servers. This increasingly applies to data centers, robots , autonomous systems, and ultimately, machines operating in the physical world.
And investors believe Micron's capabilities may still be undervalued. But the obvious question arises: has the market already factored all of this into the stock price? Micron has already experienced massive growth. So even if the long-term outlook remains strong, investors should still ask how much future growth is already reflected in today's valuation.
And that's where the debate begins. Some investors believe that Micron could eventually become one of the major infrastructure companies of the AI era, occupying a place in memory that is somewhat comparable to Nvidia's role in accelerated computing. Whether this comparison will ultimately prove to be correct remains to be seen.
But the underlying reason is becoming increasingly apparent. AI requires memory, and as AI systems become more complex, their memory requirements continue to grow . Now let's move on to Nvidia. One of the most interesting developments is Nvidia's growing involvement in AI security. The company is moving towards a security approach that spans multiple levels of the AI stack.
Hardware, software, services—and this is important because AI security is not necessarily something that one company can solve on its own. If AI is to become increasingly automated, safeguards must exist throughout the system. Models need control, and the software environment needs control. Infrastructure needs to be monitored, and security systems must have the ability to detect and respond when AI systems behave unexpectedly.
This becomes especially important as concerns grow about increasingly powerful advanced models operating beyond their intended boundaries. And if the industry wants governments, businesses, and consumers to trust increasingly powerful systems, security infrastructure could become as important as "pure" performance.
And it looks like Nvidia is positioning itself directly within this ecosystem. But Nvidia's story today is not just about technology. An important story of capital distribution is also developing. Nvidia is generating a huge amount of cash, and these funds can now be used in several ways. The company can invest in its own growth.
It can invest in well-known companies in the AI ecosystem and can return capital to shareholders. This is very different from the situation in the initial stage of the AI war. The big question then was whether Nvidia could handle the explosive demand. Now the company is so profitable that investors are asking a completely different question.
So what should Nvidia actually do with all this cash ? And one of the answers is a share buyback. Nvidia announced a large-scale additional share buyback approval. The important part is not just the size of the permit. This is the message behind it . Management effectively demonstrates confidence that Nvidia can continue to fund its expansion while returning sustainable capital to shareholders.
And the most interesting thing is that Nvidia doesn't necessarily have to choose between growth and share buybacks. She tries to do both . The company continues to invest in AI infrastructure companies, suppliers, and strategic partners. At the same time, it can buy back its own shares. This added another layer to Nvidia's investment story.
It's no longer just a high- growth semiconductor company. It is turning into a platform that generates huge cash with multiple ways to allocate capital. Free cash flow is especially important here. If Nvidia continues to grow free cash flow at a very high rate, the amount of capital available for investment and shareholder payouts could become enormous over the next few years. And this changes the indicators being discussed.
Comments 0