Tae Kim on NVIDIA earnings: 85% revenue growth, buyback catalyst, and why memory stocks are still cheap

who knows. Uh well congratulations on the progress for coming on the show.

Thanks for keeping us all safe

and we'll talk to you soon.

Cool. Thanks guys.

Have a good one. Goodbye.

Up next we have Tay Kim from Key Context, the hottest substack on the internet. We have Tay Kim in the waiting room. Let's bring him in to the TV room. Tay, how you doing?

Hey, guys.

Here we go.

Kicking it off strong. Good. Good fit. That You know what we need to do? We need to figure out how to uh do a uh like real time real time recognition of the face, face tracking, and then land the glasses on the correct on the correct face.

Your next appearance for next video.

Yes. Uh and m maybe before we get into uh AI and and uh different companies uh I'd love to just know how key context is going uh how like what what it's been like we since we last uh checked in. Uh how is life?

Uh it's much better. It's going great. I think the last time we talked was March 30th and the market was right at the bottom.

Yeah, that's right.

Weary. Everything was going terrible.

Yeah.

You never worried though.

I did not. I tripled quadruple down on here and that's CPU shortage by CPU by memory stops.

Yeah,

all those uh ideas are up 50 to 150%. So I uh I said uh I think 13 out of 14 ideas are solidly in the green. I I joke that like show Otani numbers with like three

it's not going to last but uh it it worked out. I mean we trusted in the actual

show would love memory stocks. Yeah, right.

I think he would. I'm sure he's his uh his financial advisors long HBM stocks. I'm sure uh

Yeah, he has.

Yeah, I read some I read some report that in in South Korea there's folks who are uh liquidating insurance plans to go long SK highex. It's a huge huge moment over there in South Korea

and everyone's making fun of these uh Koreans for being long memory stocks, but I mean we're talking singledigit P multiples probably going to lots two three two to three more years

uh triple digit growth. I mean I I quoted this Michael Dell thing where he spoke to a Wall Street conference. Uh he says 2525 that's the thing you need to know about memory stocks. uh AI accelerators gonna have 25 times more like AI GPUs from Nvidia two years from now going to have 25 times more memory per GPU

and there's going to be 20 a need for 25 times more GPUs. Wow. So he said multiply 25 by 25 you get 625 more revenue etc.

And you layer on to the fact that four years ago all these memory companies saw their uh revenue you know get cut in half. So they didn't expand capacity. It takes three to four years to expand capacity. So we're gonna see like mega pricing power. Like we haven't seen anything yet. These stocks are going to go keep going higher and their revenue uh revenue rates are going to be

astronomical. So everyone makes fun of these Koreans for piling into singledigit PE stocks growing at triple digits. And this cycle is different because there are only three companies that can make make the HBM memory.

Yeah.

So don't make fun of the Koreans. Don't don't uh is the the the pullback that you pushed to uh that you mentioned like the when they were beaten up was that the post uh crypto slump when uh semiconductor equipment and and GPUs were being used to mine crypto and then there was a pullback there. Is that what you're referring to? Like what is the prehistory? Why were these stocks beaten up?

So in 2022 like we had a kind of like a tech recession like everyone's all these tech companies were laying off people. Even Intel and Nvidia had these 50% draw downs. Yeah.

Uh the whole you know everything was it's almost like the the post uh co uh kind of overhang where people bought too many digital computers, electronics and uh there was too much over capacity and so that really is different.

What was Nvidia doing uh during that 50% draw down? Were they buying the stock back?

Uh not not really. Nvidia did they buy a little bit of stock back but not not really. And back then this is a different Nvidia where like data center is 10% of the revenue and gaming is 80 90%. Now now it's split right

where data center is nearly everything. Yeah. Uh what happened with Nvidia back then is uh everyone's mining their GPUs to mine Ethereum

and then Ethereum did the whole proof ofstake proof of work thing. So demand for GPUs collapsed and uh Nvidia had like a terrible quarter. They where they missed by like two billion which was a big number back then for that. Now

that's a rounding error. back then it was uh 20% of their revenue and the stock uh got hit hard.

Yeah. So what are you looking forward to from Nvidia this quarter, this year? What are you watching?

So we had like two unbelievable quarters uh the last two quarters and I expect another great amazing quarter. I mean Jensen is out there saying GPU consumption is through the roof. On Monday he said AI demand is far exceeding supply and capacity. So I think the numbers are going to be great and this is without even China. So I mean the absolute scale of these numbers are mind-blowing. We're talking like 80% growth on an $80 billion number 79 or 80 billion. Just think about in three months, right? Yeah.

So the absolute scale and the stock is like almost as cheap as it's ever been. Like it's trading at 19 times forward below the uh S&P 500 which is growing at like 10%. Nvidia is growing at 80%. So this this dichconomy where Nvidia is becoming more and more undervalued and we could talk about why. Uh

yeah yeah I'd love to know is that just because they're the largest company 5 trillion it's hard to wrap your mind around a10 trillion company.

So every year we go through this cycle um the last three years during this whole upswing up everyone says peak is here right Nvidia can't grow anymore and Nvidia keeps growing at some ridiculous uh growth rate. So the skepticism, the only reason why it's trading at below market multiple is that the AI skeptics believe that the peak year is near, right? There's going to be a no.

Is it AI? Is it AI skeptics or is it uh TPU and tranium enjoyers?

That too. They they believe competition is coming and they're going to gain market share. I kind of kind of laugh at that. And the numbers that people, you know, put out there for the competition, it's it's like a rounding error. It's like a small single low singledigit number compared to what Nvidia is going to grow in the next few years.

And what people keep forgetting is like if you actually look at the numbers,

I mean Nvidia has a trillion dollars in orders, right?

They're the ones that have locked down all the memory like I met someone guy a guy at GTC. He he he's the optical startup and Nvidia locked up all the capacity for lasers and optical, right?

So Nvidia has the supply components with memory, wafers, optical. So they're they're pretty much the only game in town if you want to actually buy AIG GPU capacity. Like TPU gets headlines, but I mean $5 billion like is nothing when you're look comparing it to a trillion dollars, right? So they had the volume, they have uh the great power, you know, power per watt uh numbers and metrics that um basically for if you're doing inference, Nvidia GPUs, even though they cost a lot more upfront, uh when you actually do the inference, performance per watt is excellent when these are uh four.

Yeah. How did you grapple with the question of whether or not Nvidia is a car? There are competitors. You said low singledigit percentage of the market, but AMD is getting its act together. Intel's maybe back in the game in a few years. You have tranium and TPU. Competition is rarely a good thing. Uh, is Nvidia a car or not?

It Okay, obviously it's not a car talking about how, you know,

I would have loved to I would have loved to be in the room with you while you saw that segment. You just like taking your computer smashing. I mean I I think I was tweeting about it. I think I wrote a piece on it. I wrote like, you know, four or 500 words about that podcast.

I don't know if we want to go into that on but uh they they you know they're going to be 80 90% of the market maybe they'll lose 10% of the market share in two three years but 80 90% of a market that is growing you know 50 70% a year and

the hypers scale or capex numbers went up huge. There's like 780 billion going to a trillion dollars next year. Yep. Like the market believes with Nvidia's valuation that Nvidia is gonna stop growing within the next year or two.

But the overall market I think is going to could keep growing at 50%. At least if it grows 50% in the next two or three years and is going to grow at even if they like maintain market or and I don't believe that's true. I think they're they're actually going to maintain or even increase market share. But even if market share goes down 10%. I mean the numbers are just insane. And I don't think Nvidia is going to trade at like 10 12 times earnings when they're growing at 50%. I mean that that doesn't make sense to me. So I think it's going to reray higher. And then the key thing here is the stock stock buybacks because that's like if you guys remember during the whole iPhone like when when the iPhone uh

five and six came out when they actually had the large screen iPhone Apple was trading at PES7

like Xcache like people thought Android was going almost like the same thing like Android's come in and destroy the iPhone market share when iPhone went on this generational run uh with the large screen iPhones right

but a big thing that Apple did was they started buying back stock and that's when their multiple started going from single digits to 15 to 20 to 30 times and I I think what and they already kind of hinted at it but I think Nvidia is going to start buying stock buyback and size they they said 50% of their free cash flow in the next 12 months they're a little bit cryptic on if that was after before prepaid uh the prepayments they have for the inventory and suppliers but I think we're going to get some more clarity tonight and if they if they actually put numbers maybe it's this quarter or next quarter if they actually put the actual you we're going to buy x hundreds of billions of stock you know in the next 12 months I think the the the PE is going to rate much higher

so you know

as soon as we get visibility that next year is going to keep growing at 40 50% or higher I think it's going to be higher and the year after that you know stock that stock should go on top of the rerating from the capital returns

uh Gavin Baker was on uh our friend Patrick's show this morning uh he is making the point that it's possible that TSMC TSMC could uh prevent a a bubble just given that they are not investing in in capex as aggressively as let's say Jensen might want them to. He said, you know, basically said what you're saying, which is uh Nvidia could sell a trillion dollars of GPUs, you know, pretty much immediately if they if they had the supply. Uh what's your what's your read on that? Do you think that TSMC is

I think well so I think I've been I mean I've been following chip the industry for you know 30 40 years so you know I I very good at reading

I'm very good at reading the tea leaves and I've noticed that Jensen and what TSMC what the CEOs have been kind of hinting at that their tone has changed in the last two three months. Yeah. So TSMC just reported their Q1 earnings and you know if you read the transcept it's pretty obvious that they're just going to raise capex on on a different order of magnitude uh for the next year and the year after. They said their capex is going to be much higher over the next three years than it was the last three years. and basically and all the stuff they're saying, you know, they're saying and Jensen's been going out there during these speeches and Q&As's he's confidently saying the supply issues in two three years are going to be on a different level. Like even with memory, he's saying all these component issues that we have now in two three years, I think he even said that on the Dores podcast interview,

two three years this all will be like less of an issue. So I'm confident that TSMC um memory might be a little problematic but TSMC wafer wise it you know it's going to be dramatically higher over the next two three years

but even if they raise capex it feels like there might be like a I don't know a speed bump essentially because if they if they triple capacity and that takes two years uh but everyone's sort of pricing in or expecting 10x a year or or half an oom a year there's a mismatch there that's could be fine Right now, right now the Nvidia valuation is trading like it's growing at 10. It's going to grow at 10% the next year or two, right? If they grow at 50% the next two years,

I mean, that's definitely, you know, like TSFC is going to raise our capex the next. So, so it's like a the numbers are way off like

Yeah. Yeah. I'm more just thinking from a market perspective.

The 10x is not going to happen. Yeah. Yeah.

But you know 50 to 75% the next three years. I think that can happen.

Yeah.

And and stocks will go much higher if that happens.

Yeah. Yeah. No, that makes sense.

10x is you know

Yeah.

It's not software. I mean this, you know, you have to actually move physical and adds.

Yeah, that makes sense. Uh how do you how do you think Nvidia will be positioning uh China exposure, China opportunity? Uh that that was an interesting point of debate on the Doc Cash episode.

Yeah, I mean I I've given up on China. I mean at GTC it Jensen said you know he got approvals on both sides and then weeks later like that you know fell through again like

it it's forget China it's probably isn't going to happen. There's also another report that gaming GPUs were incrementally banned in China. Like I I just think like you can't rely temp your expectations. I don't think China's going to happen.

Not relevant to the business case or the valuation, but potentially important to like the geopolitical story.

Um I I think it's just going to be the status quo where Nvidia is not going to be able to sell to China. And anything that happens is incremental upside, but it's not going to be a huge number anymore. Even the stuff that is allowed is is a limited number. The way the US has, you know, approved the licenses, it's China is the one that's saying, "No, we're not going to let you sell right now." Like, it's been going back and forth between China and the US, you know, who bans what and whatever. Yeah. But even the incremental numbers, even if China starts approving, it's it's not a huge number the way it was in the past. So, I would temper expectations on China. It's just it's been so, you know, so back and forth and unreliable the last last couple years.

Yeah.

Uh what was your reaction to Google IO? Uh, I'm actually pretty negative on Google right now. And they're, you know, they're basically non-existent in the whole coding agent stuff, which is like the one that's going to that's growing exponentially right now in terms of coding agents and everyone's paying tens of billions of dollars to anthropic and now codeex for open AI, you know, they're adding a million users every few weeks. Um, Google anti- anti-gravity is is nowhere to be found. you know, no one's talking about Twitter. Every anyone that talks about it says they don't like it.

So the hottest thing which is coding agents that's gonna, you know, that's going to automate things and do actual work for an enterprise is the biggest market that's taking off like a rocket. Google isn't there. So I mean Google Cloud is doing great, you know. Yeah. Growth,

etc. Did did uh seeing yesterday did it start to make more sense to you why Google had been committing so much so many resources to anthropic like do you felt do you feel like

yeah so I think they at the time at a lot of at you know at the time uh everyone was like wow I wonder how you know Demis feels feels about this of course Demis is an angel in anthropic and and uh obviously you know thinks they're a great company and they have a good relationship But at the time it was like wow what like they have this new model incoming they have this massive comput advantage they have this massive distribution advantage they have massive data advantage like they have every single advantage they have this huge head start uh vend it in everywhere on all these platforms from Google workspace which is like a massively you know underrated uh distribution point to the Google uh you know docs and and sheets and basically everywhere like every single possible advantage and an incredibly talented team and exponential demand for tokens and it was still you know somewhat surprising to see them you know invest uh invest that much in in a competitive lab and they're benefiting you know with Google cloud and get getting that business but I think long-term strategically I think Google has to be very careful and I think Microsoft experienced this with open AI I is once the value acrews to anthropic and open AI um they're going to be the power players you know four or five years from now and be able to push people around. I mean we saw that when uh Yahoo literally used Google as their search engine right Google was nobody and Yahoo put it on their on their homepage and powered their search engine and then Google became the power player in a few years. Same thing with Netscape and Yahoo. I mean this is like history repeating itself. Netscape was everyone used Netscape. I don't know if you guys were around then, but Netscape was the browser, right? And the reason Yahoo became powerful is because you clicked on that little thing on the right on the homepage and it drove people to Yahoo the portal.

Then Google, you know, took off because they started using Google the Google search engine inside yahoo.com. Yeah.

So I I think all these companies have to be be careful like if you if you help open AI anthropic uh become super powerful and they have the best models and you know this thing is all a flywheel right the more people that use it more people that use uh cloud code you know anthropics is going to get smarter and make the better version of cloud code. Same thing with chat and if you let Gemini kind of like lose the market right now.

Yeah. like and Anthropic gets all the business and all the revenue like Google starts losing power in the whole tech ecosystem. So

yeah, it's almost like the whole industry is on a merrygoround but there you know it's like a Spider-Man meme with like you know guns. Everyone's kind of just like going around.

Yeah.

Very very

What are your expectations for the SpaceX IPO? What does that do to the market? They sort of jump straight into the Mag 7 in terms of valuation. Are they going to suck up liquidity? Well, it's really it's really tough to tell without the the S1 and we'll see versus you know all we have is like these spot leaks and yeah,

you know, Star Darlink and now now with the you know selling the the the AI compute a colossus to anthropic

um I mean

the valuation is extremely extremely high. Even even uh uh the D1 Capital guy who who's up, you know, up. Yeah. He he was interviewed on on on Oshan's show and you could tell if you just read his body language, he's like, "Wow, this is a really high multiple." I mean, go back and listen to it. Even he even though he's going to benefit from it, like he's like, "Wow, this is really high."

Yeah. Uh so

I mean a lot of a lot of SpaceX investors were happy with you know uh investing at 60 billion but it was doing 10 billion in revenue and had great margins and it's a completely different business both on the valuation

goes to them they saw stuff that no I mean like it you know the launch uh costs going down and it's and darling coming out of nowhere and adding billions of you know unbelievable telecom business total market great

but like 1.5 1.75 trillion in off those Starlink revenue.

He's earned the what was this from Gavin Baker? He was saying like uh Elon has just always delivered for shareholders and that means that he can marshall unlimited capital at all times and I'm just you know I'm just hesitant like you know at those extreme valuations that's less upside for a retail investor.

Oh sure. Yeah. Yeah. Yeah. Hard to imagine like the Tesla scenario.

Yeah. the passive indexes if they go in at that price that valuation.

Yeah.

Like

on the next downturn, it's just I I just I just wish Anthropic went out first.

Sure.

Right.

Cuz the numbers are dramatically real. I mean, Open AI probably is accelerating now with Codeex 2. Yeah.

And

you know that that will help things a lot better than you know,

you know, SpaceX.

Oh, sure, sure, sure. might interesting thing right now looking at the valuations of the MAG 7 and then you add SpaceX in there and then you add you know Enthropic and OpenAI and SpaceX will look potentially make Anthropic and OpenAI look cheap on just like multiple basis but then it still looks very expensive versus let's say like a meta right

um sitting at one and a half trillion with with one of the best businesses of all time. I did I did want to ask you about Meta, your kind of updated mental model. They obviously had a big uh painful round of of layoffs today to be able to afford uh their inference bill. Um uh I'm I'm joking there, but it does seem like, you know, they're effectively uh whatever they're saving on on uh comp, they're just sort of reinvesting into

AI

AI uh both capex, but also uh their their inference bills. Um, how are you?

I think you asked asked me the same question, you know, six, seven weeks ago and, you know, they they reported an unbelievable quart. I mean, they're growing at 33%. Like crazy amounts of profits like um there's market skepticism that they could do well in AI models, but you know, there's also the plan B like you could say the same thing about what happened with Elon and XAI. like they were able to there's so much demand in this mega trend hypers cycle

that even if they falter a bit they could sell compute capacity to to the highest bidder uh that's out there so I I wouldn't count meta out I actually think like I said um their core advertising dig advertising business is more durable than the Google search engine right now which if you go you it's filled with with ads everywhere it's not it's not a great experience and a lot of that a lot of that traffic is going to shift over to uh AI chatbots. So I if you actually look at the numbers, Meta is is their business core business is very durable, has competitive advantage, it's growing, still growing like a weed. Uh I I just wouldn't count them out.

Yeah. Basically, as soon as they either have an AI breakthrough uh or they pull back on AI, the stock has to rewrite, right?

Yeah. Yeah, I mean that's the same thing happened with with with the the metaverse stuff. As they start cutting back spending, you re rer rate higher because

all all those losses don't don't hit the income statement.

They look tiny now.

Yeah. It' be so it'd be so wild to think about what Meta's stock would be doing right now if the metaverse debacle had never happened because the metaverse was like this high conviction bet on a category that was just way too early and yet AI doesn't feel And yet they feel like late on AI. I think Mark felt late or at least clearly felt behind enough to to to sort of, you know, 10x the seriousness. So um

I think yeah, I think they're a little more skeptical because of what happened with the metaverse where they spent tens of billions of dollars

that has pretty much gone.

Yeah. Yeah. Becoming a Neocloud is a weird outcome for a hyperscaler that could totally have a cloud platform but has not historically. And uh you would imagine that there's a whole bunch of hard one lessons that Azure, GCP, AWS went through, organizational decisions, structures, just key people in place. It feels hard to spin up immediately. At the same time, there's neoclouds that are scaling revenue extremely quickly every day. So uh I I I agree with you that it is it is a possibility but a whole set of new challenges if that's the direction.

That's the worst case scenario right like

they they they want to use the AI compute to make their internal businesses better with all the ad personalization and monetization and

create this stuff and actually I've been uh playing uh playing around with OpenClaw you know for the first time a few weeks ago and uh yeah I've been using WhatsApp and that's the future everyone's going to have a digital assistant. It's like so amazing. I ask it to give me the top 20 stories. It's powered by the latest version of Codeex. Uh it goes out there, searches every day for me, and emails me the top 20 stories I should read, and it's really smart.

Um you guys should try it. It's actually amazing. And and the thing I'm thinking is like, why can't Meta do this? I mean, they have this Meta AI that's not powerful, but you know, having this little channel for my little T-bot, I call it T-bot. Um, and and all I do is WhatsApp it and say, "Do this. Do this. Find this for me. Find this for everyone's gonna have this. This is a basic thing." And I'm using

I mean, the big the big head underline token number from Google was that they 7x token production across all of their services. That actually seemed a little low to me given how broad the the surfaces are and how crazy the models have gotten in the last year. Um, but I would expect to start seeing that out of Meta because uh now when you go to Instagram and you search, it gives you a little AI overview and obviously they have the Meta AI app and I would imagine that there's more AI hydration happening across the ecosystem anyway

and this digital assistant thing is going to be like OpenCL like I was skeptical until I tried it. I'm like oh my gosh.

So Meta should be all over this because it's on their

and probably with the Ray-B bands they'll do it. Uh well thank you so much for coming on the show. Always great hanging out.

Great to see you Tay. What do you do for Nvidia earnings? Do you do you just like turn off all the lights in the room and and uh like just like sit there

case of beer, nachos, big screen TV?

Super Bowl. Is it more like a Super Bowl or like a

No, I I I wish like we should bring back those bar crawls like they had it. They did it once and then it stopped.

Like we should bring that back.

Too much of a top signal.

It's going to be the number one most valuable company for a long time. Yeah. Like I'm sorry Alphabet. It's not going to happen.

Yeah. So, we should bring those those bar earnings parties back.

That sounds great. Well, we will talk to you soon. Have a great day.

See you.

All right. Thanks, guys. Good to see you again.

Uh, before we bring in our next guest, I want to actually pull up this Wall Street Journal article. See how SpaceX is about to eclipse every other blockbuster IPO. Scroll down on this because they have a beautiful visualization of IPOs over time. You can see uh the first graph. The these are IPOs uh over the last seven years uh ranging from 0 billion raised probably 10 million hundred million all the way up to two billion. Continue to scroll down and you'll see what happens as the Blockbuster IPO starts stacking up. You get Airbnb, who we're talking to the founder of just in a few minutes, uh ARM, Cerebrus, uh Uber and Porsche. And then as you keep going, Saudi Ramco comes in at 25 billion if you keep scrolling down. And then eventually uh SpaceX comes in to dwarf them all at 80 billion. It makes everything else look like a flat line on the chart. Uh