원문 보기
Nice to see you. You know, first of all, Jim, thank you for coming here. Welcome to Boise, home of Micron, the center of memory innovation for America, and soon, as you can see in the background, the center of advanced memory manufacturing. Honestly, it's impossible not to notice. This is amazing. My colleague David Faber recalled that as a child we heard about this wonderful thing called Bell Labs.
That's where all the innovation happened. And who knows what might have appeared? Maybe even a videophone. I'm talking to the person who's opening Micron Research Labs right now. These are innovations with a long-term perspective. Tell me, Sanjay, why are you willing to spend $10 billion on what is essentially basic research and hope?
And this will be critically important. I am very excited about the creation of Micron Research Labs . We are an initiator and leader, uniting the ecosystem to make breakthroughs in the next generation of the AI era. This is critically important because memory is the strategic infrastructure of AI.
Memory definitely needs higher performance. For AI to become smarter, it needs more memory, faster memory, and more energy-efficient memory. Memory is intelligence. So we're bringing here, through Micron Research Labs, $10 billion in investment from Micron, bringing together our customers, universities, startups, ecosystem partners, and of course Micron researchers and scientists, to drive research in materials, semiconductors, memory, computing, how to bring them together, and drive innovation for a new era.
I think we need to help people at home understand what memory is. Because they will ask, "What is this ?" Cerebral cortex ? Where is it? No, when we talk about memory, we mean a special type of chip that was frankly a " stepchild" of both the CPU and the GPU. CPUs are good old Intel and AMD. And GPU is, of course, Jensen Huang.
Why has memory become so important now? Why are we so obsessed with memory compared to other life forms that we have always considered superior? You know, memory is no longer just a component of the system. Memory is a strategic infrastructure for AI. This is no longer a consumer product. This is a high value.
This creates value for our customers. Without memory, it is impossible to make AI smarter. You can't make AI faster. You can't scale AI. You know, AI is evolving. A lot of context is generated . All of this context must be stored and processed through memory. So, artificial intelligence dictates a whole hierarchy of memory requirements: from high-bandwidth memory to DRAM and SSD.
This is an exciting era for the memory industry and a critical need for all artificial intelligence systems. Also, tell them that this is an important period for the United States because you never gave up. You always believed that we could be great. We have to compete with Japan, we have to compete with Korea, and by the way, we know that SK Hynix and Samsung are incredible.
Why has n't your company abandoned this country? It seemed like everyone else had moved abroad. You know, Micron was born right here in Boise, almost 50 years ago. You know, we are innovators. Today we have over 62,000 patents. And, of course, we believed in our vision of memory development. We say our mission is to accelerate intelligence to improve life for all.
You are absolutely right: even during a deep recession in 2023, Micron invested in cutting-edge research. Today you see the results. We have the best technology and product portfolio in the industry. We have invested $10 billion in our R&D and capital expenditures. Well, I wouldn't be surprised if it was a $10 billion Micron research lab, unlike anything we've had before, and it would be located here.
This research lab will be here. It will have branches around the world, where Micron also conducts R&D operations, and next year we will definitely start construction. Micron's research lab will truly strengthen America's leadership in semiconductor and memory technologies. I believe in it. That's why we're here, and Sanjay Mehrotra, thank you very much.
We're coming back to you, Paul. A rate hike is becoming somewhat more likely. I think the probability of this has increased from 60% to about 74%. So, that would be negative for our risk assets. But yeah, they both rolled back a bit. Bitcoin is around 77k, Ethereum seems to be a little over 2400. We've seen some pretty strong inflows this month, which is very nice.
About 700 million for Bitcoin products, about 150 million for Ethereum. However, just yesterday we saw outflows from Bitcoin worth about 120 million, while these Ethereum-based products received an inflow of about 35 million. It was quite interesting , a little different than what we usually see. We usually see more flows directed into Bitcoin.
But this could be a sign of growing institutional interest in some of Ethereum's new staking products. But this is something we will keep an eye on. Listen, the Genius Act , or as I call it, the Clarity Act, actually laid the groundwork for the federal structure, and now everyone is waiting for this law, which has probably been brought back to life to some extent by September 15th.
Your thoughts? Is any of this built into the price of cryptocurrencies? People in the crypto space want to know. Yes, I would say that the adoption of a transparency law is not included in the price. We considered this when the probability of approval was even higher, around 40-50%, and we still didn't think it was already factored into the price.
We are here now. Next week , on the 15th, we will have a procedural vote. We'll see what this leads to. We need 60 votes, so we still need to win over a few Democrats. When comparing Nvidia to Apple today, the valuation gap is hard to ignore. Nvidia trades at a multiple of around 14-15 to earnings. Apple is closer to 35% in profit.
But the bigger difference is growth. now the problem is not weak demand. The problem is a shortage of supply. There simply isn't enough AI infrastructure to satisfy everyone who needs computing power. This has created a very unusual situation for Nvidia: the company is growing at an extraordinary pace while demand continues to outstrip supply.
But there is another question that investors should ask. How fast is Nvidia's infrastructure actually aging? For a long time, some investors assumed that graphics processing units (GPUs) would have an extremely short economic lifespan. A new generation emerges, companies update equipment , and the old becomes almost worthless.
But real-world usage calls this assumption into question . Let's take the Nvidia A100. The architecture dates from around 2020. And yet, some of these GPUs are still contracted for workloads until 2029. This is important because newer GPUs are not needed for every task. The latest equipment is extremely valuable for training advanced artificial intelligence models.
But older GPUs can still handle inference, batch computing, physical AI, enterprise workloads, and other applications. So, all infrastructure doesn't necessarily disappear. It can move to another part of the workload stack. This extends the useful life. And if an expensive GPU can continue to generate revenue for a few more years, the economics of AI infrastructure become much stronger.
Now let's look at the Nvidia appearance, where I use the system . Some of the first systems in the NVL 72 configuration have already reached customers. But a GPU alone does not create ready-made calculations for AI. A complete system is required : GPU, racks, network, power, data center, and software.
Everything has to work together , which is why cloud providers are becoming such an important part of the AI ecosystem. They take all these components and turn them into real computing power for customers. And the demand remains huge. Some providers say that almost every GPU received can be distributed among several potential customers.
That's how tight the market remains. Eventually, supply will catch up with demand. More data centers will be built and more electricity will be generated. More GPUs will be produced, and more cloud capacity will come online. But the important question is: how long will demand outstrip supply? Nvidia's forecast suggests that this imbalance could remain significant for years to come.
At the same time, we see money flowing into the semiconductor market. AMD, Micron, SK Hynix, Marvell— many companies related to AI semiconductors are gaining momentum together. And this is important because the market does not move in a straight line. Capital is rotating. When trading becomes too saturated, momentum slows down and investors move to other assets. And this is exactly what we are seeing more and more often.
Comments 0