NVDA

“the famous fact here is that an H100 costs more to rent now than it did 3 years ago, even though we have much superior technology. And we have much more compute in the world because as models get smarter, the opportunity cost of compute gets higher.”

“I have no positions except for an all-time high break long of Nvidia because I didn't know what else to do.”

“I traded Nvidia ATH break, but like kind of retarded, and just like let it close for a small loss.”

“some of these all-time high breaks feel like... I did it on NVIDIA. I bid $212... if Nvidia comes back for $216, really if it comes back for $212, we're in there. We are in there, brother.”

“a beat, fuck it, rip it higher”

“we need NVIDIA to keep growing. And I love the cheap, it's cheap propaganda because it is cheap. It is cheap.”

“White House takes whatever stake they need to guarantee their contracts and guarantee their CapEx and force these stocks to go up. It's probably best case scenario.”

“NVIDIA remains the foundation of our infrastructure, our training fleet, and the majority of our inference stack. Can you continue to run on NVIDIA GPUs. We are now generating $2 billion in revenue per month”

“So working with NVIDIA on, we just announced at GTC, Jensen announced it, this Space Rubin chip. So it's like a slightly mass-optimized, slightly radiation-tolerant, and then slightly thermally optimized chip which can be used for space infrared.”

“Both companies are releasing new versions of their AI models at fast— at a faster cadence than ever before, while pouring more resources into the training runs that create them. The arms race is showing no signs of slowing.”

“and you know Brad Gerstner had to come in with the 100.”

“What do you think the cybersecurity analogy is for visual reasoning? Like cybersecurity is so equipped for text-based models. Coding agents, the entire cybersecurity threat can sort of be understood as a big string of text, more or less. Where do you think visual reasoning goes in terms of applications?”

“But recently, the biggest cost center for AI labs flipped from researcher time, more or less, to compute capacity. And this creates a much larger economic incentive to figure out a way to drive down the cost of chips used to train AI models.”

“They got everybody. It's a murderer's row over there. At the same time, Orbital Estate said that it has obtained strategic credit lines totaling $8.4 billion from 12 major financial institutions, including the Bank of China. Are they going to build rockets with this? I mean, I don't— I don't understand why this would be so capital intensive if they're not going to fab the chips and they're not going to launch the rockets.”

“Yeah. And even if the models are exactly the same, if I'm able to put 10 agents securing my bank account against your 1 agent trying to break into it, I will have 10 times the amount of solutions. And so I should win that battle almost all the time. Uh, and you, you, it, it does, it does seem like we're shifting towards this, uh, the, the incredible value of inference and capacity, which of course makes the whole data center slowdown ban so much more complicated.”

“And but financial performance has actually been strong even in sort of legacy areas such as search. Google Search is growing, e-commerce sales. Amazon core business is growing. Enterprise software seats Microsoft 365. We're going to have more news on all of that. And”

“And so we're starting to see the rumblings of this idea of like the neural computer. There was this, people have been talking about this for years since the AI boom began, but the basic idea is like, you would have a computer that basically has no software whatsoever on it. It would just have an LLM or just an AI model, just inference capability or potentially connection to cloud inference that would generate whatever you want, whatever you need on demand, on the fly.”

“Yeah. And I think trying to share what that means, what an exponential demand means, and that we're still so early and yet the demand for AI compute is overwhelming. I think sharing that was interesting and I think helpful in educating the financial community. The other thing is that there are lots of ways to do this. That the GPU isn't the only way. You've got a TPU, you've got Trainium, you've got us.”

“clearly we're in a shortage and ironically in a shortage it's not the best company who wins. I mean, you can look at Nvidia's stock chart and that tells you it's the second, third, fourth best companies where the demand overflows.”

“the stock is like almost as cheap as it's ever been. Like it's trading at 19 times forward, below the S&P 500, which is growing at like 10%. NVIDIA is growing at 80%. So this dichotomy where NVIDIA is becoming more and more undervalued”

“NVIDIA is trading about 13 times earnings for 70% growth for the thing that is the most essential thing in AI.”

“Jensen just announced a data center that bolts onto the side of your house.”

“There's this big, there's this big trend right now in just the LLM, the frontier AI market where companies will, you know, use the most versatile tool for the job, the frontier model, they'll token max, it'll be really expensive. And then pretty quickly they'll figure out, okay, for this workflow we can actually use an open source model on commodity hardware, we can use the legacy model, we can cut the cost by 85% pretty quickly.”

“the frustration that customers had around, you know, not wanting to be locked into one vendor, not giving all of their margin to Jensen, forced them and they were willing to look at other options that were available on the market.”

“When there's no cost to spreadsheeting... We just came up with so much more spreadsheeting that you could do. And it's bigger than ever.”

“we started doing data centers back in 2017 with, with Airtrunk... which is actually just got bought by Blackstone for, I think, $16 billion or something like that.”

“I would approach this, this next super period, uh, in the same way, which would make me very wary of, uh, what now seems to be the cat's meow, which is huge expenditures in areas in which there is no known revenue,”

“OpenAI slash its partner spends $50 billion on the data center... $35 goes into NVIDIA. NVIDIA's gross margin is 75%.”

“Today, I think that the US still has a decisive lead on most aspects of artificial intelligence, mostly because it has the compute resources here. It has the leading NVIDIA chips and China does not.”

“And then when Rubin comes out, we'll know the gap is going to expand significantly.”

“but it's still the same thing, right? Like, it's not that it like played in gaming, it played in Bitcoin, it plays in AI, but it's still NVIDIA. Like, it's not he never thought he had to change the name of the company. He's gotten better over the years.”

“Overhyped, overfrothed. And I believe this AI CapEx bubble will end badly. In a way, I just think people are it's like the telecom bubble all over again.”

“As demand skyrockets, prices are going up for everything on the supply side. Whether it be the NGPUs, their prices are going up. In addition, their useful life is extending.”

“we use 3 different chip platforms. So we are customers of Amazon's Trainium chip, Google's TPUs, and NVIDIA's GPUs. You know, we use these chips fungibly.”

“something like NVIDIA, which is still... in early April is essentially as cheap as it gets relative to the market, like in the last 10 or 12 years... It's very hard to square that valuation with something like GE Vernova's valuation because it builds in like an unfathomable amount of share loss for Nvidia.”

“this is going to be amazing for the useful lives of GPUs and may single-handedly save private credit.”

“I mean, you can still buy NVIDIA. Maybe the multiple's slightly higher right now. I mean, there's such a catch-up trade in NVIDIA at 15 times 27, 12 times 28 for the most dominant, fastest, I don't know if it's the fastest growing, but very fast growing company at its size.”

“we're already sold out of all the there's not enough compute in the world. So Anthropic has half of what they need right now, and that's before this huge takeup. So Marc Andreessen said in the next 4 years, one thing he's sure of is there's not going to be enough compute.”

“Nvidia data center revenue has gone from like, you know, a few billion a quarter to like, you know, $20, $25 billion a quarter now. And, you know, continuing to go up, like, you know, big tech CapEx is skyrocketing.”

“the thing that is the biggest competitor for any new accelerator is kind of even the previous generation of NVIDIA, right? I mean, in a fleet, what I'm going to look at is the overall TCO.”

“At some point though, pre-training will run out of data. The data is very clearly finite.”

“If the capabilities were actually at AGI level, people would be willing to spend trillions of dollars a year buying tokens that these models produce. Knowledge workers across the world cumulatively earn tens of trillions of dollars a year in wages. And the reason that labs are orders of magnitude off this figure right now is that the models are nowhere near as capable as human knowledge workers.”

“the most valuable companies currently by market cap their output is digital. So NVIDIA's output is FTPing files to Taiwan.”

“A thing I've been trying to fight for is export controls on chips to China. And that's in the national security interests of the US... And yet it doesn't happen. And we sell the chips because there's so much money, there's so much money riding on it.”

“who holds the card is the infra providers for charging margin... as you go upstream to, oh, well, who has access to all the memory and logic capacity? Well, it's, it's NVIDIA for the most part. They've signed a lot of long-term contracts.”

“on any given problem an AI tool has a success rate of maybe 1% or 2%”

“Well, in the end, something has to transform electrons to tokens. That transformation, there's no— the transformation of electrons to tokens and making those tokens more valuable over time, I think that that's hard to hard to completely commoditize.”

“Deploying in larger scale-up domains is a huge unlock.”

“in the fullness of time, compute kind of is the single most important determinant on how things work... as you scale up energy and compute and parameters, intelligence will just fall out of that.”

“Maybe for the first time, this is no longer true, right? So the famous fact here is that an H100 costs more to rent now than it did 3 years ago, even though we have much superior technology and we have much more compute in the world. Um, because as models get smarter, the opportunity cost of compute gets higher.”