Solana 对决 Hyperliquid、Ethereum 为何被高估，以及 2026 年的加密市场 | Logan Jastremski

Logan from Frictionless Capital is back on the show. Logan's a special person. Every time I talk to him, I learn something new. I struggle to keep up with most of what he's saying. Hopefully, you guys are smarter than me and can keep up better than I do, but Logan, welcome back, man.

Thanks for having me back. I appreciate you guys.

1. Logan’s Crypto Thesis in 2026

Yeah, of course, man. Good to see you. I have a place I want to start, which is when I think of the Logan thesis and the Frictionless thesis, it is that high-throughput blockchains will win, right? You guys were early to Solana and pretty early, I think, to Aptos and Sui. I'd love to revisit that thesis and have you tell me what you got wrong there and what you hopefully got right.

Yeah, it's a good question, and I would broadly say yes. We were one of the few people beating on the Solana drum in the wake of FTX's collapse. In part, it was because we felt like the industry was not going to scale, and if you enabled application engineers to build more expressive applications, they could, at the end of the day, make more money and serve more customers. I think that part was directionally correct, looking back on it.

If I were to separate the applications built on top of the blockchain from the underlying infrastructure itself, the main thing that we've gained a greater appreciation for regarding the underlying blockchain—the L1—is that the way to actually make revenue is through execution. I'm fully bought into the Fleep [?] and Santi worldview, which is that everything is discounted cash flows. I believe that is true, but I think the big thing is: How do specific applications make money, and how do the underlying blockchains make money themselves? And I think the thing that we have really come to appreciate is that you can almost value the underlying blockchain like an exchange, where it's effectively how much volume you do, and then kind of your take rate, and that can be measured either in bips, as in a traditional exchange, or with priority fees. And so, I think we've just gotten a greater appreciation for throughput does matter for scale, but throughput without execution, you don't have any revenue. And so, you really need to focus on the execution quality to actually have a good sustainable business model long-term.

Can you double-click on what that last part means?

The thing that we did get right is that once you enable more throughput, it allows applications to build more expressive apps. When we were looking at Ethereum back in the early days, developers were removing lines of code because they couldn't build the applications they wanted to build because gas fees were too expensive. When you enabled more throughput, it allowed application engineers building on top of the underlying infrastructure to do what they wanted. The crude and often directionally correct analogy is going from dial-up to broadband. You could build things like YouTube.

But for the underlying blockchain, how does that actually make money? In the past, it was, “Bitcoin went up X amount, so my chain will go up by some derivative of that.” I think that is no longer, or should no longer be, true. I've been on this podcast and many other podcasts saying I think Ethereum is the most overvalued asset in the world, and I still believe that.

In large part, I think it's made people a lot of money and has this monetary premium, but I don't think that will sustain forever. I don't know if that's 5 years down the line or 10 years down the line. I think you do need to have a model that points to these things making money, and you should be able to run a discounted cash flow on them. With the blockchain itself, what really matters is execution.

As you scale up trading volumes, the interesting thing for the underlying blockchain is specifically priority fees, if it's a permissionless blockchain. Those priority fees ultimately get redistributed back to the stakeholders across different blockchains, and those stakeholders earn a form of real yield. We've gained a greater appreciation for the actual core drivers of generating these priority fees, and typically, it's trading volumes and access to contentious pieces of state.

2. Are L1s Overvalued?

Logan, I have 2 questions here. Maybe I'll start with the one around ETH being overvalued, because I know that's the one people are going to get upset about. Is ETH overvalued, and are all the other L1s overvalued? Or is ETH overvalued while the other L1s are undervalued, would you say?

Pretty much everything, I would say, is overvalued.

Everything—all of the chains, or just every token in crypto?

Specifically, the underlying infrastructure with L1s. We actually built this interactive dashboard internally that we're going to share soon. We wanted to share it with our LPs first, but it essentially breaks down almost like a balance sheet statement of Solana specifically.

What it's showcasing is how these things can be valued on both the positive and negative sides. The main driver of the model is how much trading volume you're generally doing and then taking a certain amount—either in bips or through your take rate—from that trading volume to generate priority fees. Those priority fees are ultimately redistributed back to stakers.

When you run these models, I think you'll find that a lot of things are overvalued. But if you find some diamonds in the rough, I think this exchange model showcases that these things can and will scale if you continue to attract different trading volumes and have the assets that people want to trade.

I think Hyperliquid has done a relatively good job recently of branching out beyond crypto assets and expanding into commodities. Part of it is that you need assets people want to trade, tight execution, and liquidity. I think they've done a very good job of expanding beyond crypto assets.

Maybe say one more line about the valuations here. Because you compare these blockchains to exchanges, the revenue multiple for an exchange is what? 2× revenue, 3×, or 4×?

I don't actually know.

That low? I think it's like 10. Santi—

6 or—

Okay, somewhere between 3× and 10× revenue. I would actually guess it's closer to 2× than to 10×.

If you look at the CME and some of these non-crypto exchanges, those are valued generally between 20× and 30× price-to-sales.

Price-to-sales? Okay, I didn't know that. Well, then that kind of negates my question. My question was going to be: If something's trading at 5× revenue, if an exchange is trading at 5× revenue and a blockchain is trading at 100× revenue, but given where we're at in the cycle and where we're at in prices, it does feel like a lot of these things are down 80% or 90%. Are you saying they have another 80% or 90% to go?

I hope not. I hope not. In our dashboard, I have these numbers. The CME is approximately 27×. Robinhood is 36×. It was higher, but it's come down. That doesn't mean those are cheap, either.

Santi's more of a P/E expert than me. I would say generally that because these crypto networks and the L1s themselves don't really have a bunch of cost overhead—we've talked about this a lot, and you guys have talked about this on the podcast a lot as well—Hyperliquid, and even Uniswap back in the day, was making X amount of money per employee versus some of these much larger venues, where they were also making money but just spending a lot more money.

Effectively, the networks are almost at zero. You have the validators running this, and they pay for the hardware costs and bandwidth costs, but they're also generally still profitable entities. I've been really trying to hype up that if you break down what blockchains are, they're just databases. They're databases that connect these previously isolated ledgers.

You had the JPMorgan ledger, Wells Fargo's ledger, the New York Stock Exchange ledger, and the CME ledger. All they're really doing is sending data packets back and forth to synchronize information. That was the whole idea behind going from low throughput to high throughput: You needed to pass more data packets between all these parties to synchronize information.

And I think if you press that even further, what is this technology good for? It really is just a global ledger, if you think about it. As they get more performant, they get better at things like trading. So, if I were to think about—and I do think about—the largest TAM within the blockchain market, it really is this global exchange, at least from the underlying infrastructure side.

That's different from the applications that are built on top of it. What you would really want to do is connect all of these heterogeneous exchanges around the world and create this global exchange. That has really been the thing that we've been thinking about, because when you start to think about trading volumes and where blockchains are today versus the global trading markets, it is massive.

Today, just take Solana, for example: they're doing about 2.5 billion in daily trading volume. It's a big number, but when you look at the New York Stock Exchange, they're doing $100 billion to $200 billion a day. If you look at all equities markets around the world, that's $800 billion to $1 trillion in daily trading volume. Then you look at FX and derivatives and that broader market—these markets are doing $20 trillion-plus a day.

I think these markets are going to continue to grow. In my mind, the peak use case of these blockchain technologies is getting everybody onto one of these single systems and isolating the liquidity and different trading pairs.

So, Logan, I think the demand side is less contested these days. Nasdaq on the blockchain is closer to that reality. We're still at $2.5 billion of volume a day. If you go to the NYSE's volume, that's roughly $100 billion a day, give or take.

Yep.

So, I think we can all collectively—and even the market probably understands this—say there's a pretty high possibility of that happening. Whether it happens on Solana or Hyperliquid, TBD. But let's just assume Solana is at $84 today. Its FDV is $52 billion, and its P/E ratio is 80.

I think where one can make a lot of mistakes is when you look at P/E ratios and ask, “What's the growth?” Again, we go back to: What volume growth are you going to see, and where is it going to come from? Commodities, other applications that are being built on top of Solana—I want to get to that discussion later on in the conversation.

But just anchoring on these model inputs, because I'm looking at something you shared with me, and I have the benefit of being an LP in you guys, I said, “Okay, I do believe that Solana gets to a nice volume—$100 billion.” But where I want to have a discussion is: Where's the take rate? The take rate today is 2.5 bips. Before I answer where I think that goes, what is your view on where the take rate goes, and how can you support 2.5 bips when the NYSE and other markets are much lower than that?

Yeah, so in this model, all we're really doing is taking the annualized trading volume, which ultimately today amounts to about $1 trillion in annualized trading volume, and then taking the run rate, which is the revenue captured through priority fees, and assigning a take rate.

The take rate's a little—I wouldn't say it's one-to-one. What we are seeing on the prop AMMs within Solana is that the take rate is generally pretty tight. Generally, it's about 1 bip, maybe less—1 to 2 bips, generally. That is pretty comparable to Binance VIP 9. I think Binance VIP 9 is doing about 2 bips.

So, to answer your question, if you look at the New York Stock Exchange and these more institutional trading venues, they're going from 0.5 bips to 2 bips. On FX, it's even less. There are also a lot of second- and third-order systems that are taking their cut of the fee, like the clearinghouse, et cetera. So, I wouldn't say it's exactly one-to-one when you're looking at the exchange volume.

I do broadly think that as volume increases, the take rate will come down. The main driver in why we really think blockchains can be valued like exchanges is that you want to make sure you're doing as much trading volume as possible, because that is the key driver of the network making revenue.

If you're not focused on trading, I think you're really screwed, because trading has shown—and execution quality has shown—that people will pay for better execution quality. If you're not focused on that, it's hard in my mind to actually be able to run a discounted cash flow model on that, because you're not actually making the revenue.

3. Solana vs Hyperliquid

Logan, what do you think Solana got right and got wrong over the last couple of years?

If you look back to 2024 and 2025, I think the interesting thing is that with exchanges, obviously you want liquidity and tight execution. But the other thing is that you want assets people want to trade.

If you look at Binance and how they got their start back in the 2016–2017 time frame, they were listing all the ICO coins. If you wanted to trade some of these emerging coins, you had to go to Binance, because it was one of the few places where you could actually buy and sell these coins.

When I look at Solana, and if you think about these blockchains as exchanges, what they had in 2024 and 2025 was meme coins. It's kind of a listing scheme. You want the assets that traders want to trade. Extend that to Hyperliquid: they have commodities. People don't currently want to trade crypto assets, so you have to have the assets that people want to trade.

Even just looking forward, that listing process is massive. If you have the assets that people want to trade, generally you'll have more trading volume and more revenue.

Because of the permissionless nature of assets being natively on Solana, particularly memes, and even with pump.fun or TRUMP coin, people had essentially spot exposure with leverage because of how much volatility the meme coins experienced, which people loved. Now you have leverage with things like commodities on Hyperliquid.

When I'm thinking about the future, the things that all blockchains really need to improve upon are having the assets that people want to trade.

This naturally leads us into a discussion around Hyperliquid. The volume that they've seen is impressive. The revenue in Solana, to put it in perspective, at the peak of the meme coin mania—I think January 2025—you saw around $500 million in revenue. That has come down dramatically. Now it's $22 million, so it's super concentrated.

How do you think about underwriting Solana versus Hyperliquid here? And what is the internal allocation at Frictionless between Hyperliquid and Solana?

I'll answer the first part. I don't know if I'll comment on the second part.

The way we view it is: What is the largest market in the world? I think that is global trading.

What I criticized Hyperliquid for early on was not being decentralized. I think they have 16 to 20 validators now, and generally they're all in the same geographical region, within Tokyo. I've become less of a decentralization maxi, if I'm being honest, and more of a product maxi. I'm like, “Hey, show me the revenue, show me people love your product, and let's go.”

I appreciate that they've taken a more opinionated point of view and focused specifically on trading. But when I look at what the technology of blockchains was actually good for, it was really having all these different databases around the world, interconnecting them, and making them real time.

Zelle was a unique feature from all the big banks. They just got together and did a kind of “trust me” that they'll do reconciliation afterward. Because they all trust each other, they'll just make the payments in real time, but it still took them a while to actually settle the payments on the backend.

With blockchain, because settlement is in real time and data propagation is high, you don't need the “trust me, bro.” You can just make trades or transfers on the blockchain.

When I think about where we are today from the blockchain standpoint and where Hyperliquid is, I think they've built a very good centralized exchange. That is a great product, and they've made a lot of money. They've built a great perp product, and I think their unique innovation outside of the vault product, which was great, was ultimately these 24/7 markets.

If you look and listen to what they're trying to do with the New York Stock Exchange, they will also have 24/7 trading. So, I feel like the value proposition may be slightly diminished over time.

I love it. Is it fair to say that you have a spectrum? On one hand, you have fully decentralized, completely open, and on the other hand, you have a more centralized, opinionated product. Maybe those people can win because they're fully open and decentralized.

On the other side of the spectrum, you’ve got the New York Stock Exchange, the CME, and people like that. They’re closed, but extremely efficient. You’re saying Hyperliquid kind of sits in the middle. So, how do they win?

More concretely, if you want to tackle global trading across the world, I think you need to make a product that allows neutrality for all those trading parties. Generally, what we’ve seen if you start to study high-frequency trading—like in Flash Boys—is that you get closer to the matching engine. The New York Stock Exchange has its data center, or matching engine, in New Jersey, and they give everybody the exact same cable length in New Jersey to make sure that you’re trading on equal and fair information.

The racers build private fiber lines and microwave towers so they can get the information to that matching engine first. The only potential hiccup—or maybe not a hiccup, but it’s kind of the status quo with Hyperliquid—is that it’s located in Tokyo. This may matter or it may not matter. I honestly don’t know.

What I think blockchains that are going after the global market will allow you to do is uniquely capture information flow, or order flow. What I’ve learned in trading is that order flow is everything, so you want to trade against non-toxic flow. Robinhood also understood this pretty early on. They were able to monetize the retail user base, sell it to market makers like Citadel, and take a couple of pips off that trade.

If you can make a trade in a certain region around the world and have it executed better than it would be on a centralized exchange, I think that’s where it gets pretty interesting. I don’t know how nerdy we want to get into the engineering design of it. We call it colocation, which is the traditional model for matching exchanges today, where you play latency games, versus this global network, which we’re calling localized congestion.

So, the long-winded answer to both of your questions is that the largest market in the world is global trading, and trading is only going to continue. I think the exchange that optimizes for global trading will be the winner over the long term.

In the short term, it’s different, because—I can’t remember which trader. Santi, you may know—you can have a good business but a bad price, and it just doesn’t make it a good investment. I think that’s the question people really need to ask themselves now. You can have a theoretically amazing business over the long term, but if the price is too high as an investment, it still may be too high.

Yeah, I mean, I go back to your model, and when you open-source it, I hope people will reference this episode. I think where I’m unsure is Solana at $50 billion. You’re sort of predicating on where the volume comes from, which is what I’m struggling to understand.

I see a very clean trajectory with Hyperliquid. Hyperliquid is sitting at under 60% or 70% of Solana’s valuation—or less. What is it now? Hyperliquid at $30 billion?

If you look at FDV, it depends on which one you look at—whether it’s circulating market cap or FDV.

Yeah, let’s get that right, because Solana, for context, is sitting at $80, and it’s, call it, $52 billion fully diluted. It obviously has a market cap of $48 billion. So, Hyperliquid’s at $42 billion, or call it $43 billion, FDV.

Got it. So, $43 billion, $10 billion circulating. You could argue, as Jeff Dorman does, that you should be looking at something other than the full $42 billion. I think CoinMarketCap reports $25 billion as its outstanding token value metric, which adjusts for things like tokens that have been permanently locked or burned.

And so I’m curious. I think most of this argument is about where the volume comes from. You’re sitting on the venture side, and you’re seeing projects and developers. I can wrap my head around Hyperliquid. I don’t see as clean a trajectory for Solana, but maybe this is the time that you should be investing, because it wasn’t clear when Solana hit $8. Yano was on the timeline talking about it.

Is it memecoins? Is it NFTs? Is it Solana doing what Hyperliquid is doing with HIP-3? Walk me through your different scenarios.

If you look at Solana specifically—and really all of blockchain, if we’re being honest—you could almost view NFTs, whether on Ethereum or Solana back in the day, as a live test to continue building out the infrastructure, get more performance, lower latency, and figure out the market structure. I would say that’s step 1.

The next step is the tokenization that everybody’s talking about. You want to have the assets that people want to trade. Right now, for better or worse, the assets people want to trade generally aren’t crypto assets. They’re commodities, AI stocks, robot stocks—all of the above. They’re tokenized products and real-world assets.

You need to bring all of those on-chain. The listing process—and what I think Hyperliquid has done fairly well—is democratizing access to listings. Combined with the vault product, they have instant liquidity for these new assets on day 1.

If I go back to Jason’s earlier question about what Solana got right, I think it was this permissionless asset listing that just happened to be memecoins, with liquidity there for people to trade. You really need to expand that process of permissionless asset listing not only to crypto assets, but also to commodities, equities, and foreign exchange.

I think pre-IPO stocks are also interesting. I’ve heard people make money there, but it’s all tradable assets. If you think about it, the way I view all blockchains is that they’re asset ledgers. If we were to look back and be intellectually honest with ourselves, blockchain was really the next evolution of finance. It wasn’t Web3, as most people called it.

I think we have the data to show that the fastest applications to reach $100 million in revenue are really finance- or trading-related. For us, it’s about doubling down on the things that work: focusing on the trading side, making sure you have an asset-listing process, and continuing to tighten execution and spreads. That’s happening with these proprietary AMMs, which are very interesting.

4. Why Was Hyperliquid So Successful?

Santi, can I turn that into a question for you? If you had to choose a long right now, what would you long?

HYPE.

Are you long HYPE?

I have way more exposure to Solana because it’s locked and stuff like that, but I think HYPE offers a more compelling risk-adjusted return. This is not financial advice for anyone, but I just see a much cleaner trajectory.

Maybe we should anchor the conversation around this again: Where does the volume come from, and how sticky is that user? How sticky is it?

I would argue that HYPE has created a very loyal customer base. My view is that pretty much every other blockchain is going to want to do what Hyperliquid is doing. Let’s assume it’s competitive to launch it. They’ll be successful, and they’ll throw massive incentives at teams to build.

The question is whether the users on Hyperliquid are stickier than users on any other blockchain. I think the stickiness is marginally higher because of the lore. It’s sort of pristine, and it has no venture backing.

I don’t think they have first-mover advantage, actually. You have dYdX, and on commodities—

You’re right.

On commodities, you’re right. I’m just saying that as a perpetuals platform, they’re late to the perpetuals game, but they were the first—

So, people should read the Colossus article about Jeff. One of the things you want to see in a founder—you, Logan, are getting very technical, and Santi, I know you focus more on the quantitative side and the financials—is what founders can break out and create as a lore about themselves, and who can become a mythical figure for some cult.

Need grit, for sure.

Yeah, and he’s built this cult following around himself and around the platform. It’s the first perpetuals platform that has probably done that.

I view it—I obviously like to nerd out on the technical details—but I largely think all the tech details are mostly irrelevant. They really only serve to make the product better.

I think if you're objective about Hyperliquid, they've made a great product. It's hard to argue with that. They've made leverage easy, asset listing easy, and they have the vault product, which makes liquidity easy. So they've kind of just nailed all the basics.

I think even with Santi, it's hard to argue the value accrual is there, and they are doing a very good job. In my mind, people care. They talk about the lore currently, but people also trade on Robinhood. I think they'll generally trade on really any platform if you have the assets that people want to trade.

Now, commodities are traded more on Hyperliquid than on Ethereum. The most traded asset on Hyperliquid is still Bitcoin. But then it goes commodities, then Ethereum, then HYPE, then, I think, the indices, then the L1s, and then there's this long tail. It's a big long tail, and I think equities are number 7.

They're doing a nice job of building this everything exchange. It's the cleanest narrative. You always want to look at—if you want to take the lens of Stan Druckenmiller—the chart is very clean. Anytime there's a reset and a rebound, this is the thing that catches the bid fastest. To me, that's something that I've just learned time and time again.

You want to look at when there's a flush, like a drawdown in the market across the board: What rebounds the fastest? Hyperliquid just does that time and time again. I'm not saying that Solana doesn't have potential. You could argue all of them are overvalued, too. Perfectly normal.

But my point around Hyperliquid is that, unlike any other L1, it has an increasingly diverse volume composition. I think you can look at aggregate volume, but I also think you want to look at the composition of the volume, which is probably a more interesting and nuanced conversation.

For Hyperliquid, it's very interesting that they're moving away from crypto correlation and entering commodities. You could argue, “Hey, look, it's obviously got a lot of interest because of what's happening geopolitically.” But you add the AI trade, and energy is just something that, in my portfolio, I want to have exposure to.

It's positioned really nicely to capture not only way more volume that has never existed in crypto on-chain, but also something that breaks the correlation with just meme coins. If you look at the composition of volume in not just Solana but pretty much any other blockchain, most of the revenue comes from retail flow and stuff like meme coins. Pump.fun, at one point, was over 50% of the revenue of Solana, if not more. Pump.fun is still doing quite well, but it's just come down a ton.

So, what are you seeing from your standpoint? I don't know what you would respond to that, but I'm more curious: Are you seeing anything interesting on the front lines—teams, prediction markets, or anything else that might really enhance the volume in other blockchains?

First, on the broad comments, I agree with you. I think they have built a very good exchange product, and people like to use it. Focusing on the product a little bit more, I think that is the core reason why people like it.

Back in the day, I liked FTX before it blew up because they made fiat onboarding and offboarding super easy. As you use a product, the easier it is to use, the easier it is to recommend to a friend. So, I think they've done a good job.

For Solana specifically, what's been more interesting recently is this idea of proprietary AMMs, which is essentially moving market-making logic onto the blockchain itself and almost embedding it in a smart contract.

The way that I describe this is that, with standard exchanges around the world, whether that's the New York Stock Exchange or the CME, they give you that exact same cable length so that you don't have an unfair advantage. But they don't actually let you take, in a crude analogy, a thumb drive, put your market-making logic on it, and stick it into the matching engine itself.

That's kind of what these proprietary AMMs allow you to do: run the matching engine with the market-making logic essentially side by side. I think that has been resulting in very, very tight spreads. If you look at the data, Solana has actually been doing much better than Hyperliquid on the spot side.

What they have not figured out is perps. So, I think the interesting next question is whether you can take this proprietary AMM logic for market making directly on the blockchain itself and apply it to perps. That's where it gets a lot more interesting.

Obviously, you just need the assets that people want to trade. Commodities are hot, and crypto assets are less hot, but can you expand that to equities? If you get things like Backpack or Superstate to allow KYC, then you have KYC'd equities on blockchains as well. I think those are ones to watch.

Increasingly, what we're seeing now is that, with one of our portfolio companies, Titan, they've moved the routing logic that used to be off-chain actually on-chain, directly within the smart-contract logic. When you trade, that latency to execution time is on the blockchain itself, whereas historically that was off-chain.

Now they actually have the best execution, which we relate more to the National Best Bid and Offer within TradFi. It's guaranteeing that you get the best execution across all these aggregators across the space because they've moved this aggregator logic onto the blockchain itself.

In Chris Dixon's words, you originally start with what you had in the past and then copy that to the new technology stack. After a while, you figure out what actually works with the new technology stack.

If blockchains are these distributed ledgers, we're now starting to get to the point where people are actually reinventing and building the crypto-native logic for trading, which is slightly different from order books. This logic is living on-chain and allowing some unique things to happen.

That's been an interesting area to watch. While it's not one-to-one, I think it has the potential to actually be even better over the long term.

I think you're as equipped as anyone out there to answer a question that I think is looming for a lot of folks: What did Solana miss? Or is that a disadvantage that allowed Hyperliquid to emerge? What did Hyperliquid do extremely well, and why was it hard, or has it been hard, for someone like Solana to do what Hyperliquid did?

I think you sort of hinted at it earlier, but it's a question that I've heard a lot of people ask: What made Hyperliquid Hyperliquid?

I think their vault product was exceptional. If you looked at what happened with 1010, a lot of the market makers just pulled their liquidity and stopped market making. If you have a vault product, you can internalize that flow as a backstop.

What they've done with their listing process has also been fairly unique. If I view it purely from a technical standpoint, it goes back to what I was talking about, where you have co-location versus this new paradigm that's less proven but, I think, potentially has the better architecture for a global exchange.

Because all the nodes, if you will, were co-located in a single jurisdiction rather than distributed around the world, it makes things a lot easier. That latency, when you're not propagating packets around the world and everything is clustered into a 10-millisecond range, makes it a lot easier to design than having a global node set with 1,000 nodes around the world that have to propagate information.

It's mostly, purely, physics from a technical standpoint. It's harder to propagate things around the world if you're further away. Combined with some of the unique things they did on the vault front to provide liquidity, the asset listing, and obviously just making a very good perps product, I think they've done exceptionally well.

Solana wasn't truly competitive on the execution side until recently with these proprietary AMMs. Jason Yanowitz has a bunch of great data on this, and I've been appreciating all the research and charts that you guys have been putting out. Those proprietary AMMs are really eating into all trading across Solana, specifically because they've moved this market-making logic on-chain.

For all intents and purposes, from a physics standpoint, that's almost equivalent to what Hyperliquid is doing.

There's a reason Kyle left Multicoin to go build his own proprietary AMM, right?

Yeah. Kyle's still doing Kyle things. The proprietary AMMs are very interesting, and they're the thing that has made Solana, at least in the short to medium term, a lot more competitive.

The execution is not 50 basis points, or 10 or 20 basis points. It's 1 to 2 basis points, and they're actually more competitive than a lot of centralized venues like Binance.

5. Opportunities In 2026

Logan, we've focused most of this on Solana and Hyperliquid. What else in crypto is getting your attention right now?

Santi actually had a good post the other day about DePIN. I think we were early DePIN supporters, and I still think, conceptually, it works.

On a high level, what I think more so now is that what people originally thought was the Helium model—where it's almost inflationary tokenomics, you give people quasi-equity or tokens in this model, and people go out and do things in the real world for you—I think in large part has failed. Because Helium was the first, they got to run that experiment, and it allowed them to bootstrap their network. If you look at everybody else who tried to do it, it really just didn't work.

And so I broadly think of DePIN now as almost analogous to ordering pizza from downstairs in my building on Uber Eats. But that doesn't make that pizza shop a technology company. Similarly, with DePIN, if you're using the blockchain for coordination or payments, and even just rewarding people in tokens, that doesn't make them a crypto company. They're still a company that has to go out and build a real business by themselves. They have to get revenue, and the financial engineering, I just think, does not work.

And so I think in large part it's becoming more true that if you want to use the blockchain for coordination or payment rails, that will work. But don't do clever financial engineering where you're early-incentivizing people, because I just don't think that's actually worked.

So I'll tell you the biggest upgrade I've done in my investment framework: focusing more on supply than demand. Supply is actually harder to understand. You come from Tesla. I think the value that Elon has built is really decomposing supply, and he talks about it in a number of podcasts, like SpaceX and Tesla. He just focused on the supply chain, building a company from the ground up and challenging every single part of the supply.

Similar to blockchains, I go back to your model that you will open-source. The supply is where I get stuck. When you think about DePIN, my point on DePIN is that I think we can all wrap our heads around how freaking fascinating it is that Hivemapper, with like $50 million in CapEx funded by the community, not the company, has been able to map like 36% of the world's roads—orders of magnitude less than Google and other players.

You see this phenomenon again and again and again. We've been able to create these really dense and accurate networks, and you see Hivemapper and Helium do really well. The point that I'm trying to make is that the supply side is sort of where we're getting stuck. I think what we're probably getting wrong as an industry is the take rate and the ability to monetize these networks.

If we just assume that trading in the real world comes on-chain—I know you and I talked about stablecoin transfers, like Tempo—maybe answer this question: Do you think that Tempo, with stablecoin transfers, is an interesting business model? If you had the opportunity to invest in Tempo, would you do it?

No. I would ask them what their business model is, because at least what we have seen in quote-unquote permissionless blockchains is that stablecoin transfers are a loss leader. From an engineering standpoint, it's really a race to the bottom, because as soon as you enable parallel processing—which really all the modern blockchains do today—you can almost infinitely parallelize, and stablecoin transfers are one of the easiest things to do.

I would ask Tempo, “Are you going to do a similar model to Circle, where you have, like, Tempo USD and you're monetizing the float on the dollars in that system, or are you doing a subscription model?” Because the actual act of just doing a stablecoin transfer is not a moneymaker. Unless they can own the end-customer relationships with the end terminals, maybe there's interchange. There are other things, but just the actual money movement is not very bullish.

And so, for that reason, we haven't been very excited about stablecoin chains because I think they just don't make money.

Yeah. What's the—if we go to the tail ends—where are we totally missing the huge opportunity here, and what is the absolute bear case? Is it that these are common goods, and so the infrastructure doesn't accrue the value; it's the apps? For L1s in general?

Yeah, for L1s in general. Hyperliquid is application-specific, so it's a cleaner value capture in my mind, but generalized L1s are different. Kyle and other folks have responded to my DePIN post like, “Dude, you're overlooking Jevons paradox. There's just way more demand that's going to come online. You're an idiot.”

And maybe that's true. I'm willing to be wrong. I just am not seeing it as clearly right now. So, for the L1s themselves, I think you will value them like exchanges. The reason why we've also generally been interested in Aptos is that they put out this piece like a year and a half ago that they're only going to be focused on trading and kind of this world of trading.

Now, just starting and focusing on that doesn't mean that everything's going to pan out. You still have to get the liquidity, you still have to bootstrap traders and all that, but we appreciate the opinionated direction. I think what we really will continue to see, even across L2s, is more trading-focused blockchains.

Base even said that. They were focused on creators, and then they were like, “Hey, this doesn't make any money. Blockchains are ledgers. Let's focus on trading.” But even with MegaETH or any L2, like Arbitrum, because they're a single sequencer, they're selling co-location.

If you can co-locate next to the matching engine, you have the opportunity to potentially make more money because you're more informed than your counterpart. This is the same thing with Hyperliquid, and it's kind of why we bucket them in that same vein. If you can co-locate next to the matching engine, there is alpha there because of physics and trading.

And so the reason why we have built this model is because we're trying to be intellectually honest with ourselves: How do you actually run discounted cash flows on these things without saying everything is a special snowflake like Bitcoin?

6. The Tesla Thesis

Logan, I got dinner with your co-founder maybe a year and a half ago. You guys were long some tokens that did well and stuff like that, but also very long Tesla. I'd love to hear if that thesis is still true. Again, I don't know if this is actually in the fund or just personally.

But I mean, I'm super bullish on Tesla. I'm biased just because I work there. I think, even going back to Santi and earlier, crypto back in 2017 and 2021 was kind of the only kid in town where you could do interesting investments. Now there's AI, and now there's robotics.

I think Tesla is very interesting, mostly from 2 parts. One is: Can you build an essentially autonomous car? And I think, peeling under the hood, they've done a lot of interesting things just building the neural network to make generalized autonomy. I think people underappreciate how hard it is to do generalized autonomy versus the lighter model.

I've had lots of debates with other investors about autonomous cars and essentially only driving with cameras and a neural network versus lasers, which is what Waymo uses. I'm pretty confident—and still extremely confident, actually—that you're going to be able to do it with neural networks only.

I think the other thing that people are broadly underappreciating about Tesla is just these robots. I view robots in a similar vein to where AI was maybe 5 or 6 years ago. I have lots of thoughts on robots, but people knew that AI was going to be a big thing. Until ChatGPT actually came out, people were like, “Oh, shit.”

I kind of feel like we're in a similar vein with robots. There are going to be billions of robots. Really, what you need is the neural network for the brain, and then you need to actually build a competent robot and build it at scale. All those are relatively difficult things, but I think you're almost going to get a renaissance similar to the early Detroit days, where they built these large manufacturing lines for cars, but for robots.

You're going to be pumping out hundreds of millions, if not billions, of robots a year. Some will be in the humanoid form factor, some in other form factors, but robots have been a thing that we've been increasingly focused on, at least for me, because I think people are underappreciating how hard this will ramp when people start clicking. Those have been very interesting.

What percentage of Tesla's terminal value would you ascribe to robotaxis and robots?

Most of it, if we're being honest. Like, 90-plus percent is robots.

The P/E has always been out of whack if you look at it forever. Objectively, if you're just being honest with yourself, there are a lot of Chinese electric car manufacturers that have been able to beat them on the pure electric-car front. I have a Tesla, and I think they're well-built cars, but if you can buy a similar car that's cheaper, I don't see why people wouldn't do that.

The bet specifically is: Can you get generalized autonomy in this form factor that Tesla has done? I think that's very, very hard to do, and they have really built up over years, from the data collection to the neural-network training and that entire flywheel.

Then there's the robotic side, which to me is now more—it's less of a reality today—but it's hard to bet against Elon. He's probably the best person in the world at moving atoms around and doing that at scale. I think him building humanoid robots will happen, and it will happen at a very large scale. You're paying a premium for that today, but he will get it done.

I call Tesla the Great Wall of China. The wall will go up, and many of you will die to build the wall, but it's going to happen.

[Laughter.]

Are you a buyer of SpaceX at $1 trillion, $1.5 trillion?

I actually got asked this question the other day by a different LP. I think crypto has trained you to, when things initially come out, just watch. Don't FOMO-buy, and then see if they trade down. I'm massively bullish on xAI, Tesla, or SpaceX over the long term, especially with the data center side and space.

I think that is very interesting if you think about the power capacities, and really even just the red tape to build data centers. Elon was the first person to do 100,000 GPUs that were interconnected, where any GPU could talk to any other GPU, and I'm pretty sure they'll be the first to get to 1 million. In his podcast, he talks about all the hoops they had to jump through just to get the permits to build that.

From the SpaceX standpoint, you're really underwriting today, at least, the ability to continue to scale Starlink, which is a great business, and everybody loves the product. What they're going to do—they've been buying a bunch of spectrum—is allow you to do direct-to-cell. Then they're competitive with Verizon, T-Mobile, and really all global carriers. That will be a massive market, too.

Over the long term, what you're really buying in my mind is this increased access to intelligence and them putting more data centers in space. I don't know if it's going to be a direct buy on the IPO day. TBD on what it comes out with.

Over time, though, I do think people are generally underappreciating xAI and Grok. Obviously, they're behind today, and Grok has killed it on the coding side. But what we've seen with the scaling laws is that if you get access to more compute and generally have better data, you can continue to scale. I wouldn't count Grok and xAI out yet, even though they are behind today.

7. Can Twitter Scale To A Finance Platform

I can't let you go without talking about X and its financialization of the platform. Is that in some way, shape, or form related to crypto and what they do with it?

I'm an Elon stan, so I listen to a lot of his interviews. I appreciate people who can look at who's more directionally correct over time, because that's a proxy in my mind for predicting the future, and it also generally helps with capital allocation. Elon has obviously done that fairly well. He's generally a little optimistic on his timelines, but it does happen.

With X specifically, going back to the early PayPal days, what he wanted to build was a global database where you had very low latency, low jitter, and low fraud. I can send you the clip, but he was talking about building the global real-time database. That is what blockchains are. They are very high-throughput, low-latency, low-jitter databases.

This is why, even though I think Hyperliquid is a great product, this kind of global TAM is this global database. I think Elon is going to be on a mission to recreate that, where X becomes the kind of money or WeChat for all blockchains, or for all finance.

I think it's interesting because the main product guy—his name is escaping me—used to be an advisor to Solana and to Solana projects. So they do have an in there. To me, it's just more order flow. They capture the order flow, similar to Robinhood, and then they can sell that order flow to market makers and have another revenue stream.

If anything, what I try to think about is whether trading will continue to increase over time, whether it's going long, going short, just buying spot, whatever. I think, just as Robinhood tapped into the retail market and people thought that was a small market at the time, people today still underestimate how much trading volume in general is going to happen globally over the long term as more and more people have access to capital markets on their cell phones. It's going to be a lot of trades.

We've been talking a lot about things that you're excited about. Is there anything in particular that you are categorically not interested in investing in, or very bearish on?

That's a good question. Within crypto or just broadly?

Let's start with crypto, and then broadly.

I would say that, as of today, it's very hard to focus on nonfinancial use cases within crypto because that has shown where the product-market fit is. Maybe there's an opportunity where DePIN networks can bootstrap their networks over time and then start making sufficient revenue. We just haven't seen that in the data, and I would still say that is a possibility. But I don't think you can do what Helium did.

We've just been hyper-laser-focused on the market microstructure. Obviously, we've studied the different architectures for some time, which are the best for trading and which we think are not as well-positioned over the long term for trading. I think it's just being honest with ourselves about what actually works and what doesn't. If you're building anything on the trading- or finance-related side, definitely reach out to Frictionless.

From the things that we're excited about but wouldn't necessarily invest in, I think it's hard to invest in the wrappers, or anything outside of the core model, at least today, or in compute. Compute has obviously done fairly well with CoreWeave and NVIDIA, but it's capital-intensive.

There's this funny meme on X where NVIDIA or Anthropic puts out a new feature, and it's the Gladiator meme: “Do you live or do you die?” I think that's true. As these models get better, they continue to propagate outwards and eat into more things. You kind of have to go long the models, but it's a very capital-intensive business. They are making great revenues, but I'd be wary about investing in the wrappers or applications outside of those.

The robots are interesting because they're underappreciated in terms of where they will be in the next, call it, 10 years.

8. Which Would You Own: BTC vs NVDA

In the next 12 months, you're going to come on the pod again. What outperforms, Bitcoin or NVIDIA?

Probably NVIDIA.

Why not just pivot to investing in AI?

Santi, what do you think? NVIDIA or Bitcoin?

NVIDIA.

Over 12 months?

Yeah.

Is there a time frame for you where that changes?

No. Well, there will be a time frame where Bitcoin probably outperforms NVIDIA over a 6-month period or so. NVIDIA has been pretty flat, so I actually think it might outperform in the next 6 months or so. Over a longer-term period, I think NVIDIA continues to outperform.

The only scenario where I think Bitcoin outperforms is if everything goes to zero and Bitcoin relatively outperforms NVIDIA. You have this AI apocalypse where capex rises, there's a lot of hype, and NVIDIA falls more than Bitcoin. But Bitcoin still goes down, just to a lesser degree.

Yeah, the problem I've always had with Bitcoin is that it's at least kind of implicit that money printing will continue. That's true, but it's also just not super fun. I've personally always liked the businesses a little bit better, which is why we focused on how you actually scale blockchains.

I don't want to be super doomerish on blockchains. I do think that if we tokenize finance, the market cap of all assets today is 2.5 trillion. I think that's going to go to hundreds of trillions, and the trading of that is going to be massive. If you can own different parts of the trading stack, whether that's execution or order flow, that is massive.

I think there's also a matter of being honest with ourselves. There are other interesting things happening in the world with artificial intelligence and on the robotic side. You're seeing more and more managers expanding their mandates to do that because, at the end of the day, what you're paid to do, in my eyes, is make your LPs money. If the opportunities are maybe not as unlimited as they once were, you really have to go where you think there is alpha and you can be early.

I think there are a lot of interesting things in trading, and at Frictionless we're very early in understanding some of the nuances of the different trading stacks, including order flow and the different value capture across each of these. But there are other cool things, and I think you just have to be honest with yourself about the investable universe. It's large.

9. Final Thoughts

It's getting pretty silly out there. You just saw Allbirds pivot from being a shoe company to an AI company, and the stock is up 10X today, or 8X. It feels frothy. They used to do that with crypto. You'd launch an NFT and be up 10×.

Before we drop, what are you reading, or what is your consumption diet to stay on top of everything going on in the world and in crypto, to inform your investment framework and level up and build an edge?

I think our core alpha was always trying to aspire to be first-principles thinkers. It allowed us to understand that blockchains were bound by bandwidth. It allowed us to understand parallel execution before others.

When I try to expand into these other markets or adjacent markets, it's really just going very deep into the core building blocks and understanding the trunk of the tree, so to speak, before getting out into the leaves. Even on the compute side, it's about understanding different compute architectures, from training neural networks to inference.

I've been nerding out on understanding how most chip design uses electrical pulses versus photons, which are light. There's a bunch of nerdy stuff that I've been going deeper on to build that base knowledge, so that when you move up the stack to different applications or models, or even on the robotics side, you're not swimming naked, so to speak. It's a mixture of smart people and podcasts.

I think the thing that all people really should do is try to identify who's the foremost expert in any given field and why they're winning. Study Kobe Bryant. Study Tiger Woods—maybe he's not the best example—but study people who are dominating in their field. In athletics, this is very common. Study what the greats are doing.

You talk about Druckenmiller and all these other people: Why are they winning so much more than other people? You can do that for any given field, on the robotics side or on the artificial intelligence side. I think it takes some time to identify those people, but then you can really study them—how they thought and their unique insights. I think that's a great foundation model to then jump off of and build your own unique insights, because otherwise it's just like you're grabbing at leaves and generally have a poor understanding of the world.

Logan, always good chat, man. Appreciate it.

Appreciate you guys having me on.

Yeah, good to see you, Santi. Cheers.