FULL INTERVIEW: Why I Think Nvidia Is Perfectly Positioned In The AI Race

0:00Tae Kim, how are you doing?

0:02Tae Kim:Thank you so much for taking the time to come chat with us. And congratulations on the launch of your business. Thank you. I mean, it's been really gratifying. That first day, you never know who's going to show up. I was thinking maybe 15 subscribers or 20 subscribers, but like hundreds of people showed up, tons of billionaires and tech founders. It's insanely gratifying. Yeah, it's great. So is it over for NVIDIA? They're down 21%. We just read since the 52-week high. Is it doom and gloom? Is it over? No. I think I was on last December, and the stock is, semis and chips have gone up, and now they're back down to where they were in December.

0:48The chip sector is flat on the year. NVIDIA is down 10%. And it reminds me a lot of about a year ago. Do you guys remember? Everyone was freaking out about DeepSeek, the super efficient models. We're going to destroy AI compute. There will be a huge compute glut. And then everyone freaked out about Trump's tariff wars, the preparation day. And this year seems very similar to that. Almost it's like Groundhog Day. We have fears over AI CapEx. People think that it might be the peak. And then we have the Iraq war. and one of these things is Iran.

1:24Tae Kim:Oil up here. Iran. Easy to get them mixed up. They happen so much. Feels like the same thing over. Yeah. But, um, I'm sorry to distract. We wanted to throw, we wanted to, we wanted to show respect. We wanted to show respect. That's true. To a real podcaster. I mean, it's very similar to Iraq. That's, that's why I love it. These are great. But, you know, $100 oil, this stuff is unsustainable and, oh, probably subside. Okay, so because I like the deep seek analogy, and I feel like the market half digested the agentic coding narrative, and the Citrini article, whether you thought it went too far or was too hypothetical, like, clearly, the markets did react and a lot of names sold off.

2:09But in a world where you believe that narrative, you would think that NVIDIA would be going up. But you're saying that there are other factors at play that are sort of tamping down the excitement in the market broadly? I mean, there's no doubt. Just like tariffs a year ago, NVIDIA had 30 % drawdown when their business was actually flying, the actual fund of the business. I think the same thing is happening here with the Iran war. Things will eventually subside. oil can't be$100 forever and Trump will probably backpedal in the next few weeks ahead of the Trump So let's recap a few of the key stories around NVIDIA.

2:46We just came off of GTC and there's a lot going on at the company. I mean, it's a huge company. Maybe it'd be good to start with just next generation chips, changes to strategy, what people are actually buying. Maybe that means great CPU standalone sales or the development with the Grok partnership. What's sticking out just on the actual AI product side to you that you're most excited about? Well, inference demand is exploding, driven by the AI agents and coding assistants. I met with Ian Buck. I met with dozens of engineers at Meta, Google, NVIDIA, and all of them are seeing crazy inference demand and AI compute shortages.

3:32So across the board, people are in crazy clamoring need for AI. And we're, I mean, yeah, you're seeing that from talking to engineering leaders at big tech companies, but we're also seeing it from vibe coders who are just on X and Twitter and talking about how they're hitting rate limits and they're subsidized and they have multiple plans and they actually shift around from one model provider to another just to make sure that they're getting the tokens they need to build whatever they're building. And you see the tweets. People are building bots to pick up any kind of B200 GPU that can. They're waiting weeks and months or whatever.

4:09Tae Kim:Basically like sneaker bots, but for NeoClouds. That's crazy. Exactly. I can't believe that. And the great thing is Jensen, he's very prescient. He probably saw this demand months away. He locked up all the supply agreements for memory, co-authors, connectors ahead of time. He saw this inference demand. And to take advantage of this coding system boom, it's almost like a gold rush. You see OpenAI pivoting toward it. Anthropic obviously is thriving on it. Billions of ARR every few weeks. And Jensen acquired Grok, acquired the assets of Grok and the people of Grok. And the combination of integrating Grok's technology together with Verirubin lets NVIDIA serve this tremendous wave of compute demand economically.

5:00And Ian Buck talked about it. Jensen talked about it. So NVIDIA is positioned perfectly to thrive on this coding agent wave that we're seeing right now. On the Grok deal, Jensen did a fantastic interview with Ben Thompson and was sort of asked the same question two years in a row about ASICs, the threat of ASICs, the idea that the GPU, the general architectures can truly satisfy 100 % of demand. it feels like there's a shift in nvidia's strategy there do you see that it feels like the right move but do you do you see it as a shift in the philosophy of the company or the strategy or or is this just something that the gears have been turning for a long time and this is maybe just an unveiling of a strategy that makes a lot of sense and has made a lot of sense for a while i think what jensen does he sees where the market is shifting and where the economic value is with Mellanox, he did this in 2019.

5:59He saw the world shifting to, it's a networking chip, but he saw the world shifting to these 10 ,000, 100 ,000 GPU clusters, and Mellanox is on the need for that. In the same manner, he saw AI agents and the inference behind that taking off, and he said, oh, this Grok thing will work perfectly with Vererubens. It doesn't replace everything. And Justin has talked about 25 % of the inference demand would be, Grok would work on that. But them working together where 75 % of the inference is very Rubin, 25 % is a Grok low latency stuff. It's like the perfect combination to take advantage of this. And the other thing is, we're just in this great liftoff of AI innovation.

6:45We've talked about Anthropic Mythos, the blog post that leaked out. So we're going to have this step-up function. They told Fortune there's going to be a huge step-up change. OpenAI is coming out with their model soon. And then when I went to GTC, the biggest takeaway I had was this session between Jeff Dean and Bill Daly, both chief scientists of Google and NVIDIA. And it's online. I highly recommend people watch it. And he talked about, Jeff Dean talked about the context, have context window innovations where they could focus on the 10 ,000 documents that work well with your request and query.

7:20So we're going to have this context window innovation. Both two scientists talked about stacking memory right on top of the GPU or TPU. And that's going to be a huge innovation in the coming months or years. And then Jeff Dean talked about synthetic data for audio and video. There's this huge runway that data is not over, and then they're going to be able to take advantage of all this data that people don't realize yet. So you have all these vectors where AI models, you can just keep getting better and better. Yeah. How are you processing the idea that NVIDIA will be investing in an open-source frontier lab capability?

8:01that feels like potentially competitive with some customers. NVIDIA has never really been in that market before. But at the same time, I've been the biggest supporter of open source American AI models. I loved when Meta was doing it. I want more of it. I loved when OpenAI, Open Source, GPT, OSS. It feels really, really important, really great. But it does feel like a strategic shift. How did you process that announcement? It's not a huge, I think it's like 25 billion over the next few years, which doesn't really compete with what OpenAI and Anthropoc are doing. But these smaller models are going to be helpful for people running smaller use cases.

8:45So GPUs, as long as they're utilized even locally or in the cloud, NVIDIA benefits. And saw the top people at Quinn left, and we don't know where they left to. Quinn is an amazing model. It's kind of like what Deep Deep is, what people thought Deep Deep should be. Quen works well locally. Quen kind of subsides because all the people left.

9:08Tae Kim:Wait, what's your theory on where they all went? Another Chinese lab? I asked all the engineers when I was at GTC. No one really knew. But people are trying to say NVIDIA should actually hire them. Because the more capable open source model. NVIDIA doesn't care if you're using GT to run open source or not. They just want more AI adoption across the world. Yeah, and NVIDIA has probably more levers to pull if it turns into a negotiation with China. Like, we're tracking, like, the Manus story with Meta, and there isn't that much that Meta can give to China in exchange. If there's, like, hey, look the other way on this particular deal.

9:50Like, let this one flow through. We'll trade this. Meta not really doing any business there, but NVIDIA, of course, is going to be selling black wells at some point in the near future. And there's probably some level of pricing. It can be part of a larger discussion, which makes a lot of sense. And one thing that kind of went under the radar, Jensen literally said at GTC, they got license approvals on both the US and China side. So we're going to see billions of dollars of H200 orders. Okay. So yeah, I mean, it seems like there's a path on the demand side that's very, very clear. You've mapped it out a few times.

10:25It's a huge number. It's already massive revenues, just an incredible growth. But what is the supply side looking like? Because it feels like TSMC is not ramping CapEx nearly fast enough over the next few years. And if we see another 10x increase in compute demand, we could be really constrained on on the leading edge fab side. So how do you think NVIDIA is going to process that? Well, NVIDIA is in the driver's seat because Jensen goes there five, six times a year, and he's best friends at TSMC, and speaks at their employee day. So they are getting a higher allocation to wafers and co-wass and all that stuff.

11:06So NVIDIA will benefit. But I agree with you that industry-wide, Google is dying to get more TPU wafer tests. Sure. Sure. All the hyperscalers that have ASICs are trying to get more wafer capacity. So there is going to be an AI compute shortage in the years to come, just like you said. And maybe it just benefits because they're the biggest dog in the house and they can prepay tens of billions of dollars to get the allocations they need. Yeah. I mean, maybe there's some offtake in ASICs that can potentially be fabbed somewhere else at some point. I know that a lot of the ASIC companies wind up fabbing at TSMC, but it feels like if you're already doing some sort of re-architecture, maybe there's a way that you can squeeze something a little bit out of an Intel deal or something else.

11:58I'm not exactly sure. It's Samsung and Intel are their only other fabs that could possibly do it. That's the bookcase on Intel. Yes. Yeah. Is that is that at some point the labs and Google like across TPU, extra GPU capacity, NVIDIA, the new R. Like there's just so many buyers of lab capacity of fab capacity now that you could imagine everyone coming to the table potentially in Washington, D.C. or Mar-a-Lago since the U.S. government owns a slice now and everyone's saying, okay, let's hold hands and jump across this and say that if the supply comes online, we will buy it at this price because we have really, really solid use cases that will justify the investment for us and for Intel.

12:47So that would be a really, really good case. But again, even if the money is there, how long does it take to get to good production numbers? I mean, I suspect like Apple and NVIDIA are considering either Intel or Samsung for their lower end stuff. Whether it be like a mid-range iPhone or NVIDIA side, definitely their consumer gaming GPUs. They may go back to Samsung and maybe even Intel. Yeah, I have one more, but go for it. I wanted to know how you're processing the ARM CPU announcement. It's an interesting dynamic because they're sort of frenemies with NVIDIA now. they're competing in many ways to break the x86 monopoly because they both are selling ARM CPUs, but then they're also competing.

13:34And so I'm wondering how you think that plays out, what that means for NVIDIA and just the rest of the semiconductor supply chain. I think ARM is, their CPU opportunity is a longer term, you know, even they said 2030, 2031. It's a longer term opportunity. I don't really expect the major hyperscalers like Amazon to switch to ARM's product offering, they have their own. And same with NVIDIA, they have their own ARM CPU that they're gonna incorporate and sell. So it's not that big of a, I don't think Amazon or NVIDIA are really worried that ARM is gonna take any big share. It's probably gonna be on the margin for companies that can't develop their own ARM CPU, the more the mid-tier hyperscalers or enterprises that use these things.

14:22But I think the ARM thing is very important because it kind of confirms what the biggest underlying thing that's not really consensus yet is this massive CPU shortage that we're seeing. Just over the last few months, we have Dell, AMD, Intel, CFO talked about, they're talking about three to five year locked in supply contracts from hyperscalers. So this is a major trend that's going to go over the next few years. And the reason why is AI agents need more CPUs. The ARM CEO talked about four times more CPU cores versus last year's kind of AI infrastructure model. So we're going to see this massive demand for CPUs that people aren't really understanding yet.

15:09Because AI agents, the whole thing, requires orchestration, tool calls, database queries, web searches. And that's all handled by the CPU.

15:21Tae Kim:Give me your bull and bear case for TerraFab. TerraFab, I'm not that optimistic. I mean, it's so hard to build that. Do your absolute best to give me the bull case. Because TSMC is so short that Elon needs to find. But even then, how are they going to buy semi-cap equipment from ESML and AMAT? Like, there's just no capacity there. So I'm not optimistic on that. And this is stuff that takes decades. Chip fabs is almost like cooking, and it's not like something you could just follow a manual. It's almost like cooking where it takes a lot of trial and error accumulated over decades, TSMC and even Intel.

16:15So it's not something you could just jump right in and do. Yeah, it's something that goes back to the XAI debate about, like, do they need AI researchers or should everyone be an AI engineer? Like, are we in a research period or a, you know, the Ilya Sutsukov age of research versus the Elon Musk age of engineering? Where are we in semiconductor production? it feels very engineering, like an engineering process. But what we've seen from ASML and TSMC is that it does feel like there's a little bit of research and artistry to it. And the cooking analogy holds. I've been doing a lot of research in the space.

16:57And it's a lot of trial and error. Yeah. It's almost like cooking a recipe. And it also feels like, at least with XAI, if all the researchers are in San Francisco, you can sort of just like walk across to the coffee shop, poach someone. But if the best semiconductor engineers or technicians are in Taiwan and they see it as a national urgency to bring stability to the country, both economically and geopolitically, then you have a very different calculation. It's like, oh, yeah, I could make five times as much if I left my home country to be abandoned. That's a very different calculation. And everything that I've heard about the culture at TSMC is that the folks who work there are extremely dedicated beyond the economics.

17:49They are true missionaries, not necessarily mercenaries. And so it does feel like it's even harder to do a talent raid in the leading-edge fab world than even the AI world, which is extremely competitive and there are still tons of missionaries.

18:04Tae Kim:I guess another question I have is would you expect XAI slash SpaceX at any point to basically just open up a shop as like a neocloud? Because the thing that was probably one of the least compelling aspects of the TerraFab pitch was him just saying we need all of this compute. We need to do this because we're going to be so chip constrained. We're going to be so supply constrained. But there was no explanation of where the demand was going to come from. Is it going to come from Optimus or Croc or Twitter? Yeah, it was just very unclear. It was a lot. But there's even the question right now is, should XAI be kind of renting GPUs?

18:51Tae Kim:I don't know. Renting out GPUs? Renting out the GPUs. Because the biggest win has been Colossus 2. The infrastructure is good. Yeah, Colossus 2, which was built very fast. I think Elon's pitch with the SpaceX IPO, and we'll see it in the coming months, is the AI compute. There's going to be so much demand over the next 5, 10 years that you're going to have to use these SpaceX satellites that have GPUs in them to serve that to them. And maybe, I mean, even though Tesla's been vertically integrated to the point of being a consumer product, SpaceX has not. It's been a railroad. And there is a world where you fab the chips, you put them on satellites, on Starlinks, in space, and then you let other companies do whatever they want with those GPUs.

19:34Think what Elon did with Starlink. That's a telecom infrastructure play, and this will be an AI computing infrastructure play. Yeah, yeah, yeah. That fits that model. There's a world there. I'm not going to bet against Elon. It might just take long.

19:46Tae Kim:Yeah, yeah. What about, what's going on with helium? What are you tracking there? There's chatter about helium shortages, potentially. Jensen has talked about this. This is a risk, but there is probably like six months, six to nine months of inventory in the channel. Bernstein has talked about it's not a risk in the short term. So if this thing, if this Iran stuff lasts in two, three, four, five months, then becomes a problem. But if it gets solved or opens up with a toll or whatever final negotiation they come up with over the next few weeks, I don't think it's going to be a problem. Yeah. I do think that, like most of these materials, there are extra deposits.

20:31They're just not economical to mine. I don't think that all the helium exists in the Middle East. It's similar to the railroad thing, just like you said. Yeah, where in a supply-constrained scenario, it becomes more economical to mine American helium. Let me put it this way. If helium becomes an issue, we're going to have bigger problems in our hands. Okay. There's going to be world starvation. Let's hope not. Let's hope not. That'll be the least of our problems if helium becomes a problem. Take me through depreciation gate. How did you process that, and where do we stand now with the fear that GPUs will depreciate precipitously and H100s will be worthless in 6 to 12 months.

21:09It's totally not a problem right now. CoreWeave has talked about these things are lasting 5 to 6 years and they're getting almost 90, 95 % of the pricing. So it could potentially be a problem if this is a bubble. I don't think it's a bubble. But if it's a bubble 2, 3 years from now and there's a compute glut, then the stocks are going to go down because there's a compute glut. But as of now, it's the opposite. Like all the GPU rental prices, even for stuff that's six years old, is still being sold out. And the AI compute demand outpacing supply is so large that this is not an issue right now.

21:47Tae Kim:Do you have any theories on where the next step change in token demand could come from? Because right now we're seeing it in CodeGen, and there's a lot of optimism around these types of workflows being applied to other forms of work. But we were talking about this on Friday. Like even if AI can just one-shot beautiful financial models, it won't necessarily even make a real dent in token demand, at least compared to CodeGen, because no company needs to just constantly be generating models at the rate that, let's say, Gary Tan generates code. And so I'm like kind of been trying to wrap my head around where could these incremental use cases come from?

22:35I actually think CodeGen is still just early innings.

22:39Tae Kim:Yeah, and I don't disagree with that. 10, 20 agents, and they're kind of overseeing them. But then we have this other stuff where these models, the mythos and open AI, they're just going to get better, where you could automate all these work process flows. Companies are going to use them for every single vertical customer service, research simulating chip design where they can verify drug discovery where they verify drug molecules can do so we're just getting started at this stuff so you can you're going to see vertical ai agents on every single category and i think logan's coming on he wrote this great post on x that he says this the ai agent wave is is going to kind of uh attack this six trillion dollar knowledge economy right it's not just about programming anymore they're coming for us yes i don't think say i'm actually they're attacking the key context economy and the tbpn economy no i think it's it's like a calculator a spreadsheet you know 30 40 50 years ago we had like you know 50 accountants doing doing the spreadsheet manually right and now after a spreadsheet came, it didn't get rid of all of knowledge work.

23:54It just enabled people to think at a higher level and get more done. And I'm very optimistic about that. I mean, one way that you 10x token demand around a financial model without 10xing the number of financial models that you're building is having the agent go and collect 10 times as much data. And so there's a lot of situations where, I mean, you look at like hedge funds that want to understand the price of Walmart stock. There are hedge funds that will task satellites to take pictures of Walmart parking lots, estimate the number of people on a day-by-day basis that are going into the Walmart to shop, and then using that as a proxy to project revenue, and then flow that through to cash flow, and then flow that through to the DCF and the actual evaluation of the company.

24:46And if you think about all the different financial models and all the different businesses where you could go and say, well, for this company, I need to go to every single local, like, I want to know the price of Squarespace. Let me go to every single website that's powered by Squarespace and estimate the revenue that they're bringing in and their willingness to pay for their hosting service, something like that. And all of a sudden, like, it's just one spreadsheet. It's just one number at the end of the day. But it's like a thousand times more work went into it. Let me give you this great example.

25:14Every year I do this same store sales for these fast casual companies. So like Chipotle, Cava. And I put out this tweet. It goes viral. A year ago, when I do it, I would have to manually go to every IR website for these six fast casual restaurants. It would take me like an hour or two. I would try to use a chatbot. They would get it wrong. I did it like a few weeks ago. And all the chatbots got perfect. So it just saved me two, three hours of tedious manual labor. So that's only going to get better and better. Yeah. Yeah, it's only going to take you one. Like this year is the year that you do it with multiple chatbots and you fact check it yourself.

25:56And then forever, it's going to be just one prompt. And it got it right. And it got it right. A year ago, it wouldn't get it right. But now in one, two minutes, I put, give me the same sort of sales for these six restaurants. I put in Gemini. put in chat GPT and just to make sure they're right and they're right so that all the tedious labor all the manual labor all the data entry that you know all of us are used to um that stuff is going away and we could think higher level so i could look at the same store sales and say oh the economy is at risk and whatever but all the grunt work all the tedious work is going to be taken care of uh by these ai agents i agree completely i agree completely uh we got a lot