为什么 AI 交易才刚刚开始…… - Bubble Boi

Yo, yo, yo.

Mr. Bubble Boi, welcome to the stream, man. How are you?

I'm doing great. How are you?

I'm good, dude. I appreciate you coming on. Congrats on the TPN feature as well. You're on a little bit of a media tour right now, huh?

Yes. I'm doing the runs. When you get a trophy, you have to milk it while you can.

You have to. Dude, what's the L? What does Bubble Boi mean? Where does that come from?

I've told the story a couple of times, but maybe not publicly. I've been trading on and off since, I want to say, 2016. I was in college at the time. There was a time when I think Silicon Valley Bank was failing, and I had my little portfolio or whatever. I was working a job—I’m not going to say where—but it was in the financial space.

And by the way, I should mention I'm a hardware engineer, electrical engineer by education.

You have a job? You work?

Yes. Believe it or not.

Okay.

Believe it or not, I'm overemployed.

I work hard.

More or less, the long story short is that I threw up at work, and I was like, “No, this is over. I'm going to lose everything.” Then I made a trade that ended up working out amazingly well. I told my coworkers, and they called me Bubble Boi.

What was it?

It's technical, but it had to do with interest rates. In derivatives, the most liquid future on Earth is this thing that used to be called Eurodollar, but is now called the Secured Overnight Financing Rate. It kind of sets the interest rates for most of the world.

That day, because a bank was failing, people thought interest rates were going to go down. It was the biggest move in interest rates at the beginning of the day, and toward the end of the day it was double what even that was.

I ended up being kind of lucky. Ever since then, I was like, “Yeah, this is a funny name to have,” and everyone gave me credit. They were like, “Okay, well, you definitely called that one.”

Whoa. Give me—wait, what do you do as an electrical engineer? I guess you maybe can't share that much, but what do you do? What does that even mean?

A hardware engineer.

Yeah.

The way I would describe it is chip design. I write this thing called RTL, or register transfer logic, but it basically defines how the chip works—its architecture.

I'm very much in the chip space. I mentioned this earlier, but someone asked me, “Why don't you just quit whatever you're doing and do this?” I started realizing that if I wasn't in the technical caves and mines, trying to solve problems and seeing people struggle, I probably wouldn't have as much alpha.

I don't think I could have called Intel if I didn't really understand its advantages—its technical advantages—early on. Even though I might have worked at a financial company at one point, I was doing coding and chip design for them. I'm not really a finance person. I don't really know how to read earnings. I mean, I can, but I wouldn't do it better than real finance professionals.

I'm more of a technologist, and I just bet on technologies.

Word. Okay, so you really understand the intricacies of how these chips work and the tech side of the AI trade in general.

I'd like to think so.

How many of these people, especially the finance KOLs or shiller guys on Twitter—you tweet this a bunch of times—are there? Is there, like, 100 people in the world who understand the chip trade? Or does no one understand how computers work?

How well understood is it, relative to how many people are talking about these stocks and names? Is the AI trade, or the chip trade—

Is the question how well understood it is on the institutional finance side, or more generally how well understood it is?

I think generally. How well understood is it, especially by the retail trading side that's piling into these names?

For retail, I would put it at 50/50. Where retail really gets burned is when they try to go into things that even I don't understand, like when we have to start talking to PhDs in electrical engineering to understand the differences. That's a bit too far.

There are a couple of people. I'll name-drop A Rational Analyst, who I think is really good and called a lot of things early. I've met A Rational Analyst, and he's really good in photonics. He's actually an expert in that field.

But I don't think telling a retail trader, “Hey, guys, you need to get in on luma,” would be particularly useful. They would have done fine, but there are so many niche companies in the supply chain that you cannot confidently know which one is going to win or lose.

Honestly, maybe you don't need to. Maybe you don't need to pick a winner in a sector. You should just go along with the sector.

I don't know if that answers the question, but I think it depends. I would say most of retail probably doesn't understand it, but there are a few people who I would call retail. Calling them retail is a joke. These are very educated, talented people.

Beautiful. I want to talk about the Intel trade. Obviously, we'll start there, then we'll go backward.

I was going through your Twitter—which, by the way, I thought I tweeted a lot, man. It took me 45 minutes to get to April 10 or something—but you kept referencing a GigaBull post from February 25. It was the dancing video about when Intel hits $100 or something like that.

What from your work history and prior knowledge, using that and whatever else you studied or learned, enabled you to identify Intel as the trade back then?

There are a couple of points, and I'll go through them quickly. One of them was that Intel was a turnaround story. They took on a lot of debt to expand their fabs, and they basically telegraphed to the market, saying, “Hey, free cash flow for this year and maybe next year is not good. It's negative.”

What you need to understand—and what I'm learning by talking to more finance guys—is that that's actually all they care about. Free cash flow and margins are the only reasons I actually want to own a company.

What I saw was a distressed company that was investing. I actually like that, counterintuitively to many other people. If I see someone investing in the right technology and the roadmap, they still have to execute, but that's a way better use of capital than just returning money to shareholders to do nothing with it.

Secondly, when I was looking at Intel, what Intel's big bet was adopting high-NA. There's a lot of controversy over whether that's good or not, but high-NA is a pretty big game changer when it comes to making chips.

You're getting way better resolution and making smaller transistors, right? The questions on high-NA were never about the technology. That was never the question. Everyone knows the physics are very basic—we all know this makes smaller patterns.

The question was whether it was profitable and whether the margins were good. TSMC famously didn't adopt it, and they're telegraphing that they won't adopt it until 2030.

I think Intel's early adoption of that—and I knew somewhat of the intricacies of how hard it is to adopt, that it's not just a simple plug-and-play thing—was a big deal to me. I thought, “Okay, well, if there's a 50% chance Intel gets this working—let's say even 20%—they have a technological edge.”

I was pretty bullish on that. The Intel story now is that they've actually executed very well there, by the way. I would say their yields on 18A are great, much better than they were originally. And I think 14A now—14A is kind of what comes after this current generation of chips—is looking like a leapfrog over TSMC.

You obviously have a much deeper understanding of how these chips work than I do, but at a high level, the public sentiment on Twitter feels like TSMC dominance. Is it not the case that most people think Intel is ahead the way you do?

I wouldn't just say Twitter. I actually talk to hedge funds and institutions, and they feel the exact same way. It's a very easy narrative to latch onto. TSMC is a big brand. Taiwan is a champion.

But I think what people aren't understanding is that TSMC has a very conservative culture when it comes to the leading edge. They don't take many risks. They're very careful with how much money they spend because semiconductors are capital-intensive.

What I saw was that Intel had a strategy that said, “Okay, I'm actually going to burn all this money to use that weakness against them.” I thought it was a nice strategy and a nice strategic move.

As far as right now—and I think this is more of a recent thing, because we could talk about the earnings—

Okay.

Node quality, or node density, is one thing, but increasingly, semis aren't driven by nodes at all.

It’s actually being driven by packaging. I’ve really been hammering that TSMC is not investing at all, and they’re really behind the 8-ball on packaging. This is why Intel is going to win. It’s not necessarily that the transistors are smaller on Intel’s node; it’s that they actually have this packaging solution no one else has, and it’s world-class—better.

And this is a big reason why Nvidia invested in.

For the chat—not for me, but for everybody else who isn’t an expert on Intel packaging and chip packaging—give me the Intel packaging thesis and why that matters. How does it improve margins, what does it do for the business, and what can’t TSMC do?

Think of it like this: Why are we doing packaging for a customer? If you look at an H100, that’s as big as a chip could go on normal technology. It’s about 800 square millimeters.

When we went to B200s, what do you do? You can’t shrink the node. How does Jensen deliver 50% performance year over year? Moore’s Law doesn’t work that way. So what did they do? They got one die—800 square millimeters—and got another one, and started stitching them together. They started having communication between them.

Many chip companies have been trying to do this. No one has done it quite like Nvidia, where it’s a whole mega-sized chip. They would use a smaller chip; that’s AMD’s strategy. So that’s packaging. When you think about it, you just took one chip, added another, and now they look like one chip. You just have more area to do stuff, right?

Packaging is huge. In fact, packaging is the only way Nvidia is driving performance on its roadmap. Now, here’s the problem: The way TSMC is packaging things is using a very well-researched design where you put what they call an interposer—a plate below the two dies.

Intel is very different. Intel will use a little ribbon to connect them. That sounds like a small difference, but it’s a world of difference. Imagine if your plate cracks. Imagine if any of the lines that go through the plate are messed up. Goodbye, chip. Start over. You need to rip it out and redo it.

Now imagine if a ribbon isn’t doing well. That’s actually easier to fix, and the ribbon is going to take up less area. So you can package things more densely.

Packaging is useful for different kinds of compute, but also for HBM and other technologies. People will mix networking I/O and photonics, which would have to be packaged because photonics would go on a different node than compute. They’re different classes of technology.

Given all that, if you look at Nvidia’s roadmap, they want to go from 2 dies to 4, and they couldn’t pull it off with Vera Rubin. Fineman wants to do 4 and wants to go to 8. Then they want to do even more HBM. I don’t have the number at the top of my head, but let’s say they have 16 or 18 HBM stacks and want to go to 24 or more. This is just more heat and more stuff for things to go wrong.

Intel is really the only company investing in a solution for this. Rumors have come out already that Fineman will be using this Intel EIB. EIB is the name, by the way, for their packaging technology, for their Fineman series of GPUs.

When that note was released, they said, “Intel is just packaging it, but TSMC still gets a bit of the packaging, right?” Maybe that’s believable. But do you think Jensen bought 5% of that company for no reason? If he’s using that technology on Fineman, do you think he’s not looking at it for the real thing—for making it integral to their manufacturing?

That’s really where I think the Intel thesis is now. Yes, the yields are good on 18A, and 14A looks really good so far, but packaging is screaming at you. This is obvious. I don’t know how to say it in other words. This is as obvious as it gets on the packaging side.

How much can packaging growth grow Intel’s revenue? How much of an impact can this have on the stock?

Great question.

I’ll talk about revenue. Let’s look only at Feynman to make it simple.

Cool. If you believe the rumor that came out—that Intel is going to package Feynman, TSMC packages the compute dies, and Intel is going to package the I/O dies and maybe other things like HBM—when I do the math, that’s an extra $4.8 billion in revenue for that year.

Just this quarter, they said they had $1 billion that isn’t even recognized. It’s a $1 billion backlog, meaning they have $1 billion in orders for packaging. No one talked about this. I saw everyone tweeting, “It’s a CPU shortage. Here’s a CPU shortage.” I’m like, “Whoa, do you not see this business line coming out of nowhere and growing?”

They’re not even recognizing that as revenue because it’s not this quarter. They haven’t fulfilled it. But they have $1 billion for this quarter.

Think of Fineman's 4.8 8 billion. Think of TPU. Think of any kind of ASIC that Broadcom is doing. Think of any kind of customer who can’t go on TSMC because TSMC is booked. They have no option.

If you imagine Intel gets the complete Feynman order, or they get the next generation, that's 48 to 60 billion in revenue just on packaging. It’s a big number.

On one GPU, on one product line?

It’s a big product line. Now imagine TPU. Imagine anything Broadcom is doing for its customers. Packaging could probably be a $100 billion business alone.

That’s a big number. Can you explain your trading style? When you did your due diligence on Intel and found an opportunity here in early 2025, did you just gas it? What is your trading style? How did you enter the trade, how did you size it, and how did you think about it relative to the rest of your portfolio?

If I told you what was going to happen next year, how would you size it? If I told you Intel’s revenue next year, you would just buy a bunch of stock and hold it, right? That’s basically what I did. I bought a bunch of stock, held it, and when I saw any kind of catalyst, that’s where you leverage it up. That’s where you might use options.

For a long time—almost all of 2025—Intel was 80% of everything I had.

80%?

80% of any money I had in my life was in Intel stock.

Not just your portfolio, but your own money?

No, in my life. My 401(k), savings, anything. In my trading account, I was maybe leveraged on Intel.

If you have a certain amount of money, a broker will give you what’s called portfolio margin. With any kind of margin, you could 2× your capital. You have $100,000; you can use $200,000. You can imagine how confident I was.

That’s not for everybody. When it comes to Intel, and when you make profits, you need to actually take some profits as you go. If Intel goes up and you have call options, they can go up 200–300% in a day. That’s a huge P&L swing, and you’re better off locking it in than just saying, “It’s going to keep going.” The math of option derivatives works that way.

If I’m extremely confident, I would start buying stock slowly over time. If I’m not as confident, I’ll take a position and wait. When I see any kind of catalyst, that’s when I would play with options.

You’re saying you have to actually realize profit. What percentage of your Intel position have you actually realized?

I’ve almost realized everything with options.

Got it.

The profits on the options are probably the size of the position, right? I can give you an example from these earnings. I’ve had maybe 40–50% in stock. The 20% I had in options went up 3× the value of the stock.

Anything with this kind of expiration, just cut it and take it. If the stock runs up, like on SanDisk, this is a better company, but if it ran up a lot and I wanted to keep holding it, you could just hedge and keep a floor. Buy puts that you think are cheaply priced and hold on to it.

My point is that if you see a trend or see something no one else is seeing, and you’re getting incremental signals that people are waking up, you have to go for it.

One of my favorite crypto traders of all time is this guy Jez. He has a quote about the art of the full port: “Why would you ever put money behind your second-best idea?” I love it.

How many times have you been north of 50–75% of your net worth in one asset, and how many times out of that have you been right?

I think I’ve done it 3 times just this year. But again—

On what?

SanDisk was one. SanDisk at the beginning of the year. When Iran happened, I was very, very full-port long oil, fertilizer, gas—the whole thing. And then, of course, Intel into these earnings. I was very confident.

But that being said, I kind of agree with that to an extent. Your biggest, most confident thing should be most of your returns. I only diversify when I’m scared—when I don’t know what’s going to happen, when I’m unsure. If I’m going full-port, there’s a reason; there has to be a catalyst. So think of it as returns over time: when the Iran situation first fired off, I might be up 300% on SanDisk, but is SanDisk going to go up 300% in the next month or so? Lock in the gains and go into this. There are more returns to happen in this short time in this asset class than there are in what you’re holding, right? So I think of it as almost time-weighted, right?

So, I also—I don’t know if you could see that; that’s a Phillips 66 American crude barrel back there. I also—

Nice.

I took the oil trade. It wasn’t a full-port trade. I understand that was a unique situation, but as it relates to SanDisk, as it relates to Intel, and as it relates to prior trades—a broad question—what does your DD process look like? What do you need, and what do you do to convince yourself that this is going to be a north-of-50% portfolio allocation? I’m going to go all in. I’m going to tweet about it. I’m going to get excited about it. I’m going to tell everybody about it. How do you mentally get there in the DD process?

I will never DD just a company. I’m not big on, “Pick the best company.” Intel’s a rare, rare example, but even Intel, I’m really long on it because of a technology, right?

I ask myself, what technologies are out there that are interesting and can help scale things? What is kind of obvious? What’s my view on that? My view might be that packaging is very important. My view might be that flash is very important. My view might be that this radical-sized die stuff is kind of cringe—it’s stupid, it’s not going to do well. Maybe wafer-scale is better.

However I see a technology, I try to research and understand the technology. I’m not really into, “Let’s find the best company on these earnings.” No, I don’t care about that. Then, once I feel like I’m very confident in this technology, I’m like, “What are all the names in the space?”

Got it. What are all the names in the space, and which ones do you think are going to benefit the most if this technology takes off?

If you’re trading, the Iran thing was a trade. I was not planning on just holding oil. This is more of a “What would be the impact if this happens?” situation.

If the impact is that I’m bankrupt, I have to get ahead of it and plan for that, right?

I think, on Iran, I didn’t play it as good as I should have, but it’s flaring up right now still, right? So there are better ways to play that.

It is. To be fair, oil was a tough expression of a long Iran war after the first week or two. You started to get—you know, you were PvPing Bessent.

Last Intel question, and then I have some fun stuff to talk about.

Price target. Intel price target.

Intel is cheaper than AMD, which is crazy. It’s crazy—like, that shouldn’t happen. It’s cheaper than AMD. GlobalFoundries, which is a no-name fab, has a better multiple than Intel to this day. What? Intel should be a trillion-dollar company. Minimum $600 billion by the end of the year in market cap.

So that puts you at, what, $100 to $150 a share? I think $200 a share is about a trillion. Again, it’s not going to happen overnight. They have to deliver, so the next earnings matter, but it should be at least $600 billion in market cap today, minimum.

Word. Thank you. To talk about some of the general stuff, the market has gotten pretty crazy. I’ve spent the last 4 years only paying attention to crypto, so I’m coming to this AI trade a little bit late here.

The state of the feed on Twitter over the last couple weeks—I have a good Alexander tweet. I’ll read it: “The CEO of Starbucks just quit and is joining OpenAI to serve coffee. Here are the 4 micro-cap South Korean companies I am buying on IBKR to escape the permanent underclass.”

The entire feed now is this bottleneck, this bottleneck, this AI bottleneck, photonics, this, this, that, that, that—micro-cap, micro-cap. Here’s a $10 billion, $5 billion, $1 billion, $500 million, $100 million company. Yeah.

Let’s see how far out we can go before everybody dies. Before I ask what the bottlenecks are, what do you think is the state of this AI trade? Where are we right now in the cycle? How much higher can this stuff go? How many sectors are left to pump? Are you afraid of the micro-cap shilling that’s going on?

Be careful of the micro-cap shilling. I would say that to anyone. Micro-caps are a whole different game, and that’s not how I would play it.

If you think something’s going to go up 1,000% but it could also be volatile, you don’t need to put all your money in that. If 1% is going to go up 1,000%, you’re fine either way, right? You’re going to make 10x your money, right?

Be careful with micro-caps unless you are an expert in some kind of technology field and can evaluate their IP and strategy better than people on Twitter. Then don’t listen to my advice.

Yeah.

As far as I see, AI bottlenecks—I made a joke about this yesterday—but there isn’t really much of a bottleneck. The only bottleneck is possibly TSMC and Intel fab capacity. You could think of TSMC—really TSMC at this point—as this warlord that kind of sets the price for everybody else.

At the end of the day, if you’re going to do AI compute, you need AI logic nodes, right? You need logic nodes. If you have AI logic nodes, you need HBM. If you have HBM, you need advanced packaging. If you have advanced packaging, you’re going to need these special PCBs.

TSMC is kind of like this warlord that just controls the whole semiconductor economy right now. Whatever they decide flows through to everybody else. You don’t need HBM if you don’t have logic capacity; there’s no point.

That’s how I think of it, and I really don’t think there’s any kind of crazy bottleneck. There are a few places, but they were priced accordingly. If there really was a bottleneck, NVIDIA would have bought them or bribed them to expand their capacity. A lot of this stuff is speculative, right?

I think where it’s most well known is in photonics, where people are like, “Okay, well, you need a really high-quality laser for photonics. Look at this little microcraft who makes this pretty decent laser. Maybe they’re going to get used by NVIDIA.” It’s like, maybe—but maybe not.

Yeah. What’s the take on photonics in general? Every day there are 5 new tickers on my feed. “This photonics company—this is the one.” What can you explain about photonics as a vertical broadly?

I am the wrong person. I was actually going to go to grad school for photonics, but I said, “No, that stuff doesn’t scale.” I was like, “That stuff is cooked.”

Oh.

Right. I was at the Intel Museum last week, and they had little photonic stuff that they did back in 2010. It didn’t hit the market, and I’m like, it didn’t hit the market because it didn’t work. It’s not because Intel didn’t want to bring it to market. Intel’s been investing in photonics for a long time. They still are.

Photonics is scary. The reason why photonics has exploded is literally that NVIDIA is making a networking switch that has co-packaged optics. Co-packaged means the lasers and all the optical stuff are next to the die. It’s next to the compute. It’s not like those plugs you might have seen in a data center, right? It’s not a plug; it’s in there. It’s in the package.

Let’s see if they get it to work. I really don’t know. My guess is, yeah, maybe a 40% to 50% chance they get it to work, but I don’t want to full-port anything, right?

Here’s the thing: I talked to someone who was more confident on this than me, and they have more expertise. Even if they don’t get it to work, all the people who supply lasers and fibers—and here’s the thing: photonics needs coolers. People don’t know this. Photonics needs special coolers, because if the temperature goes off even by a little bit, the thing doesn’t work.

Imagine your computer stopping—I mean, your computer will stop working if it gets too hot, but it’s like 103 degrees C, right? Photonics is very sensitive to temperature.

But he said, like, for the cooling company—

Yeah.

It might be called Fabernet. Let me see. They've already run up Fabernet. They’re called thermoelectric coolers. Here is F—FN.

It might be thermoelectric.

Yeah, but if you just talk to Claude or GPT and say, “What companies make thermoelectric coolers?” they’ll give you a few.

Hey, what do you think about this? By the way, there’s this concept—it’s called, as I’ve heard it, “hallucination yield,” which is basically the delta between the price an asset trades at and what ChatGPT thinks it should trade at.

And then, I guess more specifically, news drops, an announcement comes out, or a narrative is hot, and then everybody runs to Claude, Gemini, or GPT and says, “What do I slam?” They all get told the same thing. Momentum comes in, attention piles in, the narrative picks up, and the thing pumps. What do you think of everybody using the same AI to trade the same stuff?

It’s been happening for a long time. I don’t think it’s anything new. I think every big trading firm has their own little news—like, hey, some news breaks out: what do I trade? They’re doing it in milliseconds, right?

As far as how it works for the general retail person, it probably does make the move stronger in whatever direction the AI says. But I’ve been saying this for a long time: AI is only going to answer the questions you know how to ask, right? If you don’t know what to ask, you’re not going to get a good answer.

I saw you tweet this today. It’s a good tweet.

Yeah, it’s only going to answer what you know how to ask. If you don’t know what to ask, you’re not going to get a good answer, right?

People don’t know about thermoelectric coolers, so they don’t even know to ask about them. If you’re using AI for investment research, you still need to do research. The AI is just making you do it faster.

Instead of saying, “What photonics companies should I buy right now because of this news?” you should say, “Teach me photonics. What is it? What can I learn? What are all the pieces?” Then you might build some fluency in that discipline.

I’m not trying to stir the pot, I swear, but I want to ask you directly: on this low-cap stuff, what do you think of the Serenity account?

I think Serenity is probably pumping and dumping a lot of stocks. Definitely. I can’t say anything else. It’s definitely pump-and-dump.

I think it’s a she. People tell me it’s a he, but I like saying she because it looks like a girl. She did call some interesting companies. I called Intel a hacker, but that doesn’t mean everything I say is right.

When you’re posting a sub-$100 million, sometimes sub-$80 million company, and you already own 1% of the company because that’s possible at those scales, right? 1% of 80 million is $8 million. You can borrow and leverage it, right?

I would just say be careful. You come from the crypto side, right? A lot of people had WhatsApp groups and signal groups where they were like, “Yo, this is the next big thing.” You’re exit liquidity, bro. That’s what you are. Take everything you read with a grain of salt.

I’m purposely trying to muddy the waters and make it so people can’t understand what reality is anymore.

What the fuck was your tweet yesterday? Yesterday, the—was that this murder last night? What was that?

That was me basically saying what all these other people were saying: “Oh, there’s this new bottleneck.” It was like, yeah, but that was completely false, and people ate it up, right?

There was a kernel of truth, but I definitely spread it. In my eyes, I’m just trying to teach people a lesson here. You’ve got to be careful.

Trust me, guys: for everyone who’s listening, this AI trade is going to go on for at least 3 years. There’s enough money to be made. Everyone will make money. You just don’t need to full-port everything. Stay in the game. It’s about surviving.

If we’re on a rocket ship, don’t jump off. Don’t jump off and say, “I’m going to get to the top of the rocket ship. I don’t want to just do Coke. I want to do crack. I want to do the craziest thing I could do.” Everything within reason—be moderate.

To ask that follow-up directly, I’m kind of in the spot I think a lot of people who watch are in. I had a really good trading year, all things considered, but I got pretty bearish on AI. I was trading long oil, which naturally got me pretty bearish on equity markets broadly.

It was sort of the ultimate trap. Tip your cap to the market makers: just enough to shake everybody out and make you think it’s going to nuke Car Island[?], and then rip that thing back after 8 months of consolidation to a V-shaped ATH.

I think there are a lot of people who have actually done pretty well, but they have a lot of cash. They’re stuck on the sidelines, watching these semis, photonics, and AI moves go parabolic. How do you approach entering the market here? How do you evaluate names?

You said it’s going to go on for at least 3 years. How would you approach entering here—figuring out what to buy, when to buy it, and how to buy it?

Ask yourself which companies are actually adding value. Which companies are actually going to lower the cost per token? Which companies are innovating now? Do these companies have any kind of competitive advantage, or are they just selling a small commodity?

Size it appropriately. Intel and SanDisk are companies that make products for which there is demand, and they’re innovating. They’re also genuinely pretty large, so it’s harder for them to go down that much.

You see the tide is rising, and you just need to be on it. You do not need to go crazy. I think larger, more established companies are better for the most part. There are exceptions, but these are large companies for a reason. They’ve succeeded for a reason up to this point, right? You could be more aggressive with sizing those.

When you’re looking at all these niche photonics companies, what do you know about photonics? I don’t know anything either, right? Let’s have a little humility here. Know what you don’t know. That’s actually number 1.

With Iran happening, if you really think it’s a risk, you can hedge. You can take a small amount of your capital. Say you have $100,000. If semis sell off, you’re probably going to lose $30,000 or $40,000, approximately—just back-of-the-envelope math.

You could take $10,000, $15,000, or $20,000 in oil options and hedge that risk. If semis are really outperforming, even if you lost $10,000 or $15,000 on that, you’re probably still going to be fine.

It’s about risk-reward, right? Do you want to be all risk, or do you want to have less risk and good reward?

What are your narrative invalidations to us going up only for the next 3 years? How are you going to trade IPO season? SpaceX—we know it’s coming very soon.

Yeah.

Then you have OpenAI looming, Anthropic looming, and the rest looming. Market top—how do you think about that?

No, I don’t think it’s a market top. Why is there even a bull run at all? It’s because people are buying these products, right? They need to buy them because they need to serve AI, and serving AI is a great business.

You actually see Anthropic making tons of money off of this. Their revenue is—if you did a regression, the line is just changing. The slope is off the chart, right? There’s money to be made, and people need to buy these materials to make that money. That’s going to continue.

As far as IPO season, they need liquidity. They’re big enough that it matters. I think Anthropic and OpenAI will probably be fine.

If you think tokens matter—if tokens are like electricity or tokens are like the internet—you have to pay that monthly bill or something. These are new businesses, and I can’t work without tokens at this point. I’m stuck. I have to pay for it. I’m sure there are many other people in the same position.

Again, we’re just at the start. We just got the commitments from these big hyperscalers and big tech companies that they want to spend $680 billion this year.

We just got that this year, right?

Yeah. Next year maybe it’s not $680 billion. Maybe it’s $300 billion or $400 billion. The year after, maybe it’s $200 billion. But the trend is still there, right?

As far as SpaceX, I’m not going to say anything too bad about SpaceX, but they definitely need some liquidity, and it’s priced a little aggressively. Let’s just say that.

Yeah. A follow-up on that: I’m curious what you would look at and pay attention to as far as metrics go, if at all. Is it a qualitative thing, or is it metrics?

You get this Wall Street Journal article today that really has no net new information. It’s about OpenAI—nothing new, right? We’ve been seeing the Brad Gerstner clip, which is 8 months old, but OpenAI isn’t making enough money to fund these contracts or pay Oracle for these data-center buildouts. Same story.

Are there metrics that you should pay attention to? Is it just general adoption of AI growth? What do you actually look at to see whether this is accelerating or decelerating?

Capex spend.

Yeah, this is the best question I've ever been asked. By the way, kudos to you. This is a great question. You're asking what's the signal, right?

I think capex spend is a signal, but we all know it's going to level off eventually, right? But it's still pretty high. Leading labs' revenue—I mean, don't look at just one company; look at the aggregate. Look at revenue, but also look at tokens: how many more tokens are they shipping? Look at that.

Then I would say you absolutely have to look at leading-edge fab capacity, because if that's expanding—like I said, TSMC is a conservative company. They're not going to expand that if they didn't know there were customers. I know for certain—it's actually been telegraphed even by TSMC—that AI accelerators on leading-edge nodes for TSMC today are 35% of that capacity, right? So, however many chips they make, 35% are AI accelerators. That might be going to 90% in a couple of years.

Wow.

In a couple of years. So that's all of it. That's basically all of it, right? So these are the things I would look at. I would look at shipments, like data centers, or whether construction is breaking ground. But that's not as important as these other things. Construction has all kinds of other issues. But that's why I would look at those main things. Tokens are going up.

Yeah, I like that, actually. How concerned are you that—even consumer businesses, every engineer is using tokens, and it's only trending up and to the right forever. I don't know why that would slow down—but at some point the gap is basically even between Chinese AI labs and United States AI labs, and all these massive enterprise companies just get switched over to Kimi or whatever for one-tenth of the cost? Is that an actual concern?

Yeah. No, that's actually—again, you're asking the best questions on Earth right now. For the consumer side, it definitely is a concern. For startups, it's a concern, right? Those people have lower switching costs.

A big multinational like JPMorgan or something—I mean, even today, I think we got an announcement from Amazon that OpenAI is on Amazon Bedrock. I don't even know what Amazon Bedrock is, but it's on AWS, right? It's like, "Oh, that's actually—why is OpenAI on that?" Well, because some people use AWS, and they're like, "Oh, you're on AWS. We trust you." Right?

So there is that. When it comes to the enterprise side, I think enterprise is definitely more sticky. Smaller consumers will switch the second there's a reason to switch. Either you're charging them too much or this model is better.

I think the Chinese labs have a real advantage. Give them credit: they are definitely matching the United States in certain categories, right? In coding and these other things. But really, it's cost. The cost is really just however much you want to pay. At the end of the day, if you can make it cheaper, you can just run it on your own infrastructure. But I don't think that's an existential threat, because everyone's going to innovate.

Yeah, okay, I like that take. What about—and I'm just kind of firing at you now—but what about Citrini's 2028 superintelligence deflation crisis thesis? What do you think about that playing out?

I think that's stupid. Yeah, I don't think it matters. Technology is deflationary in a sense, right? But we're just going to get more stuff done. There's going to be more money in the economy, and things are just going to move faster.

You can already see it with coding. When we invented the steam engine, we did more trade. When we invented the kind gym, we picked more cotton. When we invented the internet, we did more communication in business. AI tokens just let you do more work.

What's expected of you compared to 5 years ago? 5 years ago, you would have been a legend if you were using this. Now that's the standard, right? So you're just going to get more work done.

I don't see employment falling off a cliff. I think it's different kinds of employment. It might be more specialized, but there are arguments on both sides. People will tell you, "Oh, well, you don't need the junior software engineer." Well, honestly, that junior software engineer is pretty cheap, and the tokens are about the same thing. As long as he knows more or less what he's doing and I can check his work, he's probably fine.

I feel like it's actually worse for some of these more senior people, unless you're doing something very, very niche, right?

This is a curveball, but just generally speaking, how much money does post-economic mean? What does it mean to be post-economic? Lifestyle, whatever.

What's a good salary in America? It's maybe $200,000 or so.

$200,000 is good. Not if you live in New York City—it's not great—but it's very, very good.

In New York City, maybe $300,000, maybe $400,000.

If you're making $300,000 or $400,000 in New York City, you're lit. Yeah, you're lit.

All right. Now, what if you don't have to work? What if you're making interest?

That's post-economic to me. So, yeah. If your capital base can sustain your lifestyle very comfortably, you're very rich. You are rich.

I don't actually consider people who need to work—I mean, there's no shade to people who need to work. That's great. I work too, right? But I think it's more so that if your capital base can sustain your lifestyle very comfortably, you're very rich.

So why are you going to continue working? Do you plan to continue working for the long haul?

I mean, what is there to do, you know? I actually like technology. I want to build. The thing is, I don't have to work for people I don't like. I don't have to stay anywhere I don't want to be, right?

But to me, it's more like there's this AI, hyper-to-the-moon thing. Maybe it's a great idea to be involved in that in some way or some form, and you've got to find your meaning somewhere. I think, also, if I wasn't working—I said this before—but I don't think I would have actually known enough to make these decisions, right?

So, no, I think working is good. I think learning is good. I think, you know, just do what makes you happy in life. I'm not a Reddit bro who's going to sit around and play video games, right? So it's a different kind of lifestyle, I guess.

What is the purpose of the Bubble Boi account? I'm assuming your boss knows about it—maybe helped you get hired, I don't know. But what is—

Definitely not.

No. Does your boss even know who it is?

No. No, I don't think anyone really knows who it is.

So what is the purpose of the account, and why do you post so much? Post your trades, telegraph what you're doing.

No, I want to stop. Yeah, at this point I'm just like—

No, I actually said I want to delete it, because at this point it's like, what benefit is there? I mean, it's great to talk to you. People like you are really the benefit. Other than that, there aren't really many benefits anymore.

I really just made this account for fun. I've had it for a long time. I like meeting people through it. I run a group chat where it's all electrical engineers of different specialties, so meeting and talking to them is great.

Stuff like that—I think Twitter or X is great for networking. It's the best networking because it's not like LinkedIn. You just meet people who have the same interest and might have a different view from you. That's really the value of X, right?

I mean, me posting about stuff—I really just wanted to do a victory lap while I could, honestly, because I've just been beat.

You took a pretty good lap, by the way. It was a pretty good lap. It wasn't the greatest ever, but it was a respectable lap.

Yeah, I had to take a victory lap. In my eyes, the purpose of the account will probably change. I don't think it's going to be P&L shots or really recommending companies, but maybe more focused on the technology and making serious posts about that. People can interpret from that what company to buy, right? Yeah, of course.

Do you recommend a retail trader focus on and approach trading style-wise to outperform this market? Do you recommend people do what you do—find a little intricacy and never put money behind your second-best idea? How do you, as a retail trader on Twitter who's online all day, outperform these markets?

It's a hard question. I think if you're a retail trader, forget coming up with good ideas. That actually doesn't matter. You don't need a good idea. You just need good risk management.

You could buy 10,000 stocks, sell 10,000 stocks, and if you're really good at cutting losers, letting the winners run, and getting rid of anything that's correlated, you'll be fine. I would say just focus on risk management.

Focus on being like, "Oh, I have all these stocks, and then this news happened and all these things went down. Let me figure out why." Let me be like, "Okay, this went down 20%. Why is that? Is it worth it? Are people in that sector also going down? What's the story? How much of this is a risk?"

Am I long semis, or am I long memory and long Intel and long Photronics, and short PCBs, or whatever, right? Know where the risk has really come from.

Trading is all about risk management. It doesn't actually matter if you have a good idea. You're going to have good ideas, and you're going to have bad ideas. The only thing that matters is that your good ideas run and you make more money on your good ideas than your bad ideas. I am not the best trader either, to be honest.

Let me be real. On the short side, I never made money shorting companies.

It’s the worst. I hate it.

The short side’s terrible. It’s really hard to short companies and make money. I’ve gotten things wrong. I haven’t managed risk. But because I’ve been trading for a long time, since my college days, the only way I’ve been able to reach this scale is by understanding risk management. That’s the only way. I probably would have been here earlier if I knew how to do that.

I would say, literally, just know how to manage your risk. Know where your risk is coming from. It’s not, “I own Apple. I own Google.” It’s, “I own a feature. I own tech. I own these exposures. I own Chinese companies.” If you own Chinese companies, you should know what’s going on in China, right? The currency and all this other stuff. I think risk management is all that matters at the end of the day.

Okay. One of the last things—first of all, that was beautiful. One of the last things I want to ask you is maybe the most important question: What is your next Intel? What is your next big one? Do you have it in your mind yet? Are you looking for it? Are you just focused on Intel? Where are you with that?

Intel’s still not done. Like I said, it should be $600 billion today. Just basic math: $600 billion. It should be $1 trillion probably soon. It should be $2 trillion maybe in 3 years. If everyone’s been asking me what the next Intel is, just buy Intel. The fact that it’s not going down from its 30% jump tells you everything you need to know.

Other things I’m confident on—this is obvious, right? SanDisk is making an innovation called high-bandwidth flash, HBF. It’s like high-bandwidth memory, but it’s for flash. They’re probably getting samples of that by the end of the year, so I’m very interested to see that. To me, that’s like, “Don’t kill yourself. This HBF is coming. Stay in the game. This HBF is coming. Pay attention.” I’m really excited about that.

I mentioned a company, but it’s not a huge company, and it’s not my number one thing, but I like this company. The ticker is FLEX. Flex basically makes liquid-cooling equipment. We’ve seen that they’ve telegraphed that the voltages of their systems are going up. They’re going from what was called 48-volt to 800-volt or something. You’re going to need more cooling.

I think the main reason why I like Flex, though, is because they’re not just making these little cooling blocks. They’re designing them with the people who are designing the chips. Contrary to popular belief, if you design it with the people designing the chips, and you’re pushing this much voltage, the cooling matters a lot. The cooling needs to be designed for your chip. It can’t just be a generic thing. That’s what we’ve been doing for all of history.

Now they’re going to make special cooling blocks for specific chips. I think that’s underpriced. I think the market right now is just seeing this as a commodity cooling company. I see it as a consulting company that’s going to consult every company making these crazy high-voltage chips on how to cool them.

Damn, sick chill. Thanks for the show. Also, Dills, thanks for the gifted subs and Intel men in the chat. That brother, that brother is up. Oh, wait. Okay, last thing I want to know is why did you add to Intel before earnings? You were just fully convicted that it was going to crush and pop. Was that just a gamble, or was that, “I know something”?

I definitely didn’t have any kind of inside information. When you look at the history of earnings, trading earnings is great, by the way. Institutional people trade earnings.

If you look at earnings, earnings surprises and earnings beats will typically come after a crazy miss. But if the miss isn’t structural—Intel missed last time, and I lost quite a bit of money—but the reason why they missed was that they didn’t ramp up capacity enough to meet the demand.

I also knew that their EMIB technology was great and people were trying it, but this isn’t inside information. You’re the CEO of a company, and you miss because you didn’t see this huge demand for your product. Now you’re trying to ramp up supply for your next product. What are you going to do? You’re not going to miss again. You’re going to go back to your team and say, “Get this figured out.”

There are a couple of ways you can get better and get more out of the same process, and Intel figured it out. People were going to say, “We knew that they telegraphed it.” They know the timeline of 14A and when it’s going to come out. 14A is at 0.9 PDK. PDK is just, you know, it’s good to go. 0.9 means it’s almost 1.0. They were at 0.9. They were telegraphing this.

On packaging, they were telegraphing this. You just have to pay attention and know what matters. I’m not always right on these things, but it was a gamble. I just had a feeling that there was no way he was going to mess it up twice.

Yeah. Well, it’s helpful to hear your perspective on how you think about it. I appreciate you coming on. I’m a fan of yours because I think there are a lot of people on Twitter who are anonymous profile pictures and are super deep on a sector, a trade idea, or a concept. You get obsessive about it, hyper-understand it, and publicly have alpha.

You’re a really smart guy, but you’re a guy who’s on Twitter. You respond to people. You can communicate with people. They can get in contact with you. You post your thoughts publicly, and then you’re a monster. I think it’s inspiring for a lot of people on this app who are trying to figure out how to trade and navigate markets. It can be done. There are people who aren’t Citadel doing it publicly with a thesis that’s logical, linear, and makes sense, and then putting the bow on top.

Congratulations on the trade.

Thank you, and thank you so much.

Thank you for coming on, man. Maybe we could do a part 2 at some point.

Definitely, man. Thank you for your time. It was great fun.

Hell yeah. You’re the man, dude. Anything you want to show before you leave? Anything you want to mention?

I gotta mention my boys at Through, since you're big in crypto. I want to mention my friends at Through. You can see it on my profile. My bad.

So, mention my friends over there.

Shout out Robert Chang, baby.

Yeah.

Hell yeah.

All right, brother. Have a good one. Bye.

Peace.

That guy rocks. I really enjoyed that. What’d you think? Bubble Boi interview. These are the kinds of guys who get missed on a lot of these big finance podcasts and shows, and they’re the ones I’m most excited about. I’m most excited about the random anonymous person who’s just a guy.

He’s obviously a really smart guy, but he’s not a top engineer at Jane Street. He’s just a guy who gets obsessive about an idea, understands it super deeply, finds an opportunity, crushes it, puts size behind it, and then is super right in public. To me, that is the storyline of this stream. This is what I’m interested in. This is what I care about.

I think you could be Bubble Boi. I could be Bubble Boi. You could have hit that trade. It’s possible for you to do.