VanEck Research Head: Why This Isn't The AI Bubble Everyone Fears (Here's Why)

Transcript

And I don't think that there's a systemic issue to the Bitcoin price because as you noted, the more the hash rate falls, the more that those who can stay connected will earn, you know, additional profits. >> Just like the streets of Manhattan, I'm going to keep this quick. Over 70% of our viewers are not subscribers. [music] Yes, you. Every week we interview some of the best entrepreneurs and investors in the financial industry. The easiest way to support our channel is to hit the like [music] and smash the subscribe button. Now sit back, relax, and enjoy today's show. And we are back with AI Supercycle, uh the uh head of digital assets research at VanEck, uh Matthew Siegel is joining us today. We're going to be chatting about

when this equities trade is going to rotate back into digital assets. I'm excited to get into it with Matthew Siegel. Matt, thanks so much for joining us today, man. >> Good to be with you. >> Absolutely. Um Matt, maybe you can just give us an update on, you know, what what you're seeing out there. You know, it feels like we're transitioning, you know, we had a market condition where it felt like no matter any kind of good news that we had, the market wouldn't react positively in the digital assets. You know, we had DTCC coming on, clarity looked like it was coming through. These things just weren't generating any kind of upward momentum.

It feels like we've transitioned into a market where now we have Saylor selling, clarity doesn't look like it's going to go through. It's not like we have any negative momentum, either. Market feels totally apathetic. Um and it feels like the equities market and the AI trade is just swallowing up all of the liquidity. Can you just tell us how you're seeing this? Is it Are you seeing this, too? And And what is your reaction to what's happening out there in the market? >> Yeah, I think the first 5 months of the year were a very concentrated market. And [clears throat] what you what we saw was that the companies that were spending the most on capex had the the best performing stocks. Uh and

since June 1st, that that relationship has flipped. Um I'm putting aside the last few days where we bounced off the situational awareness bottom, it was the company spending capex that that were getting punished. Uh so, it you know I I think Bitcoin and crypto kind of fell into the software bucket, right? So, the the semi to software ratio was something that a lot of people were tracking. And in the end, Bitcoin is software, happens to be open-source software. Uh but, the you know, the things that you can do with Claude code and with Codex, they're they're having

meaningful impact on some of these open-source software projects that are, you know, they their upgrade cycles are are not as fast, not as top-down driven as uh like Web 2 companies. You can't just force an upgrade. Uh and we started to see, you know, vulnerabilities. Uh maybe it started with like the Zcash issue, uh but I think that kind of software underperforming is a lot of what's been weighing on Bitcoin and crypto. Uh and and then it's just, you know, the 4-year cycle and people kind of respecting that pattern. So, I'm hopeful that um this kind of changing equity market regime that we've seen since June

will create some more balance, uh some more dispersion in the returns, and will bring back, you know, the ability to harvest alpha from kind of single stock positions, single asset positions, rather than just the factor bet. I don't think it has to be, you know, A works and B doesn't, B works and A doesn't. Uh I think the market will be more nuanced than that. Uh And in terms of my own positioning, still pretty exposed to the to the AI infra theme via the Bitcoin miners.

Optimistic that Q4 will bring a bottom to Bitcoin and and some again some more balance to the market, but I'm also like kind of happy to buy on good news when there's some volume rather than play this chop which we seem to be in with the Bitcoin price. >> Okay. Let's talk about where you're positioned now and then we'll get into maybe the core Q4 bottom and and and some more of that thinking. But where you're positioned now and there's some tickers here but effectively it's Bitcoin miners converted into AI data centers because they've already built out the infrastructure. They've got power and energy capacity in rural areas and this

is all of the same raw components that we need to do a data center and AI. They can take the same hardware that they already have switch it to a different product or different end end product with the same raw materials end up becoming way more profitable. Mara, Riot APLD WULF. These are some of the tickers here. And so um you know, this seems like it's a trend that's going to continue to play out. Data centers are just now getting built out and and there's still plans for a whole lot more.

Um one question is how long do you think this trend to continue for? And then the other is like what kind of adverse effects here happen on Bitcoin if so much of the computing power and the hash power ends up leaving and going into the AI world? You know, obviously we have this mechanism where things balance out. Hash power leaves to AI, the difficulty adjusts and it becomes easier to mine Bitcoin, but at what point do we reach some systemic um effects of a lot of hashing hash power leaving the Bitcoin network? Could that pose for some systemic issues um to the Bitcoin price?

>> Yeah, so our largest source of returns in the strategy that I manage, which is the node ETF, and we launched uh now 15 months ago, uh and we've managed out perform Bitcoin by nearly 100 percentage points, uh and that was driven by, you know, identifying uh early that the Bitcoin miners were being uh you know, valued at uh a very low multiple of the megawatts of power that they control relative to the few data center REITs that existed at the time. Um so, you know, the early Bitcoin mining model was uh dilute shareholders with equity

issuance to go out and acquire Bitcoin mining ASICs faster than your competition so that you can at least keep up with the growing hash rate. But, because the you know, your revenues fall by 50% every 4 years, it it was kind of a melting ice cube problem and a very capital intensive uh not particularly profitable business model. Uh and what you know, what we can see now from the debt issuance, not just from the hyperscalers, but on the back of these data center uh construction leases, is the cost of capital for these names has collapsed. They're able to fund uh their capital needs in the debt markets instead of the equity markets, and each successive lease that's been signed uh

has pretty much been at at better economics. Uh and then you add on the lower interest rates to that, and there's been a lot of um value creation. So, uh you know, that that's where we've been when we you know, that that the the change in the market regime uh that started in June obviously hit the leveraged folks way worse than us. We don't run any leverage. We're very conscious of trying to underweight the companies that have the most leverage. And when we saw that it was taking down you know, a fund like situational awareness where we had a lot of overlap with the positions, but but here was a leverage seller being taken out by the prime brokers. We took that opportunity

to lean in. We made the largest single day trade since we launched last Thursday morning at the open exiting almost 10% of our kind of lower volatility, lower beta exposures and doubling down on some of the minors that that we like the most. So, you know, those are the actions that we've taken. We haven't seen any fundamental deterioration in the return on capital that the hyperscalers are expecting from this investment. In fact, if you read the transcripts from Amazon and others, those returns are tracking better than their earlier expectations and companies that had been renting older GPUs for $2

an hour are now trying to refresh those contracts quite a bit higher. So, you know, first and foremost we're looking at the fundamentals. They're still improving and it got to a point where some of our favorite names were pricing in just the value of their existing leases without anything ascribed to the terminal value of these data centers or the pipeline platform and the prospect for for new leases. And that work was done even assuming that the 10-year rise is another 100 basis points. So, we saw a pretty pretty good margin of safety at the lows where Leo was blown out.

And so, we know we've increased those positions. Uh you know, we can get into some of the historical analogies. People are comparing this to you know, the the railroad capex boom of the 19th century. I wrote a long piece of that that we can get into it. Um you know, that lasted 20 years. We're in year three or four here. So, I I think there's a long way to go. Uh and and I don't think that there's a systemic issue to the Bitcoin price because as you noted, uh the more the hash rate falls, the more that those who can stay connected will earn, uh you know, additional profits. So, some of the names that we've that we've added to like Bit Deer and Mara, which still have

quite a bit of optionality around whether to, uh you know, repurpose their facilities or or keep mining Bitcoin. There could be a Bitcoin price here where people are talking about when pivot back. Uh so, that type of optionality uh we really like. And and in terms of increasing exposure to, say, uh some of the altcoins or some of the kind of tokenization infrastructure companies, uh um waiting for more uh clarity to emerge on clarity or rule making by the SEC or just better prices accompanied with some volume. Right now, the volumes have been pretty dismal across uh across most of the crypto tokens.

>> Yeah. Absolutely. I mean, that that is a you know, it's a great great explanation. Um and it it's, you know, something that I think we've been watching for uh for quite a while. Um and now we're we're starting to see how this plays out. And so, when it comes to these facilities, that you mentioned this optionality where they, you know, now are seeing the profitability of going and converting their data uh their their their hash power into AI data centers for compute. Uh but you mentioned how this might be a pendulum that ultimately swings back, presumably if the Bitcoin price appreciates such to such an extent where now, despite, you know, maybe another having, block reward is lower, it still becomes more profitable for them to come

back into Bitcoin and contribute their hash power and their compute into the Bitcoin network rather than some of these AI networks. Does this create some sort of like a reciprocal flywheel where they're almost chasing one another in a spiraling upwards? Can you just talk us through what some of the like the financial thinking is as you sort of see this pendulum shift back and forth? >> Sure. Uh uh Fred Thiel, who's the CEO of Mara, was on a podcast a week or so ago, and he he said that if Mara wanted to, they could fit all of their Bitcoin mining machines into the one new facility that they just purchased in South Texas.

Uh so, for those miners who are running containerized facilities, uh there's I think I'd just call it optionality to use these facilities for the purpose that suits them best at at any given point. So, it I'm not calling for a pivot back for those facilities that have already been repurposed. We estimate, for example, that CleanSpark would need 360K per Bitcoin in order to, say, tear up the lease that just that they just signed. So, you know, that that's not going to happen.

Uh but in terms of moving these ASICs around into underutilized facilities while they prepare to lease them and make additional CapEx, you know, I think that's going to be a viable strategy. So, you call it a reciprocal flywheel. I would call it optionality. I would say that companies leveraged companies that are serving a new market are going to have a lot of volatility and that's what we saw with Bitcoin in the early days. That's true in AI. If you can package these two volatile businesses together, you know, maybe they can even themselves out and I think that is is still under priced by the market.

>> Yeah. Um it certainly feels that way. It you also pointed to how sustainable this is. Maybe you could just give us a quick run down of your article and how this comparison shakes out to the railroad. You mentioned that was a 20-year build out of capital capital expenditure. We're in year three of the AI build out. Maybe you could just help us understand why this comparison feels accurate. And I think that's what a lot of people are starting to question is how sustainable is this. We've seen, you know, these large capital debt raises by the hyperscalers. Then it felt like a lot of that capital started to move down the supply chain as the hyperscalers started to move. You know, then we we got DRAM, we got chips, we got memory.

And now it feels like things have reset. The hyperscalers are kind of coming back to life. Have we just gone down the supply chain, reset, now we're going to run back to the same playbook or you know, do you think you also mentioned dispersion. Do you think there's maybe some some new market conditions at play here that we should keep an eye on and why, you know, this go around is maybe not the same playbook as what we had previously seen in the AI equity markets? Yeah, I mean my my bet is that the fundamental scarcity here is now atoms, no longer bits. You know, we spent three decades de-industrializing the United States because commodities

and manufacturing capacity opened up around the world. We learned from COVID that these integrated supply chains are vulnerable and so there's a reshoring requirement here and then you add on to that the demand shock from AI where these this intelligence cannot be served globally, right? France does not want their AI served from a US data center. So you essentially have to double build everything and you lose those economies of scale. Plus the you know, the NIMBYs have made it so hard to build anything in this country that puts a fundamental cap on how quickly this capacity can come back on.

So the the you know, maybe the Bitcoin miners were lucky, but maybe Bitcoin kind of always anticipated that the world would shift back from bits to atoms. You know, we can debate that, but but that's my thesis and those who push back on it point to the railroad boom of the 19th century as a great example of a capex cycle that transformed the economy but destroyed a lot of the initial capital. So we can compare it on the scale and then we can compare it on how it was financed. So on the scale the US essentially put 3% of GDP into rails for almost two decades. We are now I'd say

five years this is year five of a post GPT and we're only hitting that 3% of GDP number once this year. At the current equity valuations, there's no way the market is discounting that this continues for two decades. In fact, most analysts see it peaking by by 2030. So I I see a disconnect in in the scale first and foremost. The second is that the railroad boom was actually government engineered. Uh, US Congress passed a law in 1862, the Railway Act, uh, and it granted, uh, millions, hundreds of millions of acres

of federal land to these railways, but the titles would not transfer until the whole network was built. Uh, and not only that, but the US Treasury floated construction bonds, and those construction bonds were subordinated to private capital. So, the government engineered a financing boom. The largest railways were selling bonds overseas that they were marketing as safe, uh, which were dependent on these land transfers, these land sales, and the companies didn't even have the title to the land at the time that the bill was passed or that the bonds were sold. Uh, so that

that was a bubble, right? And the other thing is that these these rail networks, they didn't provide the true utility until the East Coast and the West Coast were were connected. You can contrast [snorts] that to an AI factory, which as soon as it's trained the chips uh, or has connected to to the grid and has the proper fiber, it can start uh, training models and serving inference immediately. There's no need for, you know, some type of uh, global network to be to be synchronized. Uh, and then the last major difference is the the contracted backlog that is supporting uh, the financing around this. So, uh, you know, there was no one buying rail tickets in 1870 for a uh, a route that wasn't completed until

1885. There was no forward freight market. Uh, it it was speculative based on land sales. Contrast that to these data centers where just the four largest cloud providers alone, Amazon, Google, uh, they they have more than $2 in contracted backlog. Uh, Microsoft and Oracle are half of that. Many of those deals include customer prepayments, customer supplied GPUs. These contracts are more than 5 years in in duration. So, there is you know, legitimate purchase orders associated with the compute that will come out of these factories and the financing

is based on that, not based on a government handout to the to the private sector as it was in the railway capex cycle. >> Yeah. So, now the argument the conclusion of this argument is that it's much more sustainable. >> Yes, more sustainable backed by a private sector rather than government handouts and secured by multi-year backlogs by the hyperscalers. >> Fantastic. Um Matt, only got a few minutes left with you. Really really appreciate that and I've got this bookmarked so I can read through the entire thing a little bit later. We've got newtype trader in our chat here that's saying

he can't remember if it was 2021 or 2023, but you were calling that Solana needed to correct the pause button on the L1. Uh newtype trader here is curious, what would Solana need to do for you to remain bullish? Um I think you know, that that's one question and then I'll I've also I want to follow up with we see a lot of L1s starting to reduce their inflation and commitments to economic security. Probably a little bit of a different question, but what would Solana have to do in order for you to remain bullish on Solana? >> Yeah, um you know, we took down our kind of L1 exposures across the firm after the election when many of the tokens doubled and there was no acceleration in

in real world adoption. No app that that kind of went viral and took in uh you know uh global um capital. And what we've seen since is it's been these corporate chains. Um Circle has one, Stripe has one, Robinhood has one. These kind of quasi-permissioned blockchains that allow uh public companies to customize the user experience and extract some of the economics. And I I understand that's not in the spirit of open-source development and um you know new cryptographic money, but um the folks who want to use these networks

at scale want predictability uh of the fee stream. So, I know the question was about Solana, but I'll like pivot to ETH for a minute cuz it was an example of a chain that had very volatile transaction costs. Uh and many large institutional um uh participants, you know, want more predictability. Uh so, generally these corporate chains have taken a lot of market share uh and that's been the focus of our investments um since the election. Uh you know, Solana doesn't have that that fee variability, but it you know, Solana still um until recently in New York, you couldn't really buy USDC on Solana. And I think

we saw another headline this morning, Wells Fargo working on tokenized uh deposits. It's going to be their own custom-built blockchain. Uh the banks and the regulated entities don't want to put real capital on top of these open-source chains. And then to the extent they do, they end up having to uh dilute their interest by supporting, you know, three, five, or 10 L1s rather than just one. So, there hasn't you know, you have a market share loss, and then you have uh a lack of kind of winner-take-all characteristics within those that remain and that's left us, you know, very underweight these tokens. I think there could be a give you know, there's many people who

think like me. I'm not the only one who's thinking this. So there's it's one way positioning. I'm sure and if there's a Clarity Act that somehow passes although the odds of that are at a year-to-date low. I think there'd be an enormous relief rally in some of these coins. It would provide a disclosure regime where we actually would know who are the beneficial owners. You you know, you wouldn't have people tweeting and shilling this stuff without having to disclose their ownership. You would see how many tokens the labs owns versus the foundation etc. And it's been that lack of disclosure. I think that has made many institutional investors ignore this space. So I want to see that change before we can get properly bullish on many of the L1s.

>> And last thing I have for you on this front also on the L1 front is ETH, Solana, and Near all came out with proposals to reduce the amount of inflation that they have going to pay validators. Ethereum as early as this morning. Solana I think put out this proposal earlier this week and then Near has this fund that would generate some yield and that yield would go to paying validators. All of this would take the inflationary supply of ETH, Solana, and Near respectively to down to almost zero if these proposals go through. Still very early in the proposal process. How relevant do you think these inflationary measures are to the market conditions for each of these L1s? Do you think they have some

weight or is this thing really getting held up by the rise of these consortium private permission chains with the banks and the lack of clarity? >> Yeah, I mean we did a an exercise couple quarters ago where we looked at the average inflation rate of L1s compared to their user growth and fee generation and you know, there's definitely an inflation problem across the space and we started this conversation by comparing the software companies to the crypto tokens and you know, the the year-to-date outperformance of semis versus software forced a lot of these software companies

to have a reckoning with how many new shares they were printing and some of the dilution in that space has come down a lot as well. So, I think it's a an appropriate thing to explore. I think the details details will matter. Let's see also what the impact is on you know, companies that are relying on these staking revenues to fund initiatives like Bitmain would be a good example. They're often touting their staking income. So, that you know, this could collapse. There could be second order effects, but I think it's wise now, you know, six years after the L1 kind of uh birth to you know, rethink some of the inflation.

>> Yeah. Certainly. Matthew Siegel, director uh uh head of digital asset research at VanEck and PM of the actively managed VanEck on-chain economy ETF as well as the node the node fund which is doing incredibly. Matt, thanks so much for joining us, man. Really really appreciate it. Great insights just across the spectrum from AI equities all the way to digital assets, L1s and beyond. Really appreciate it. I'm going to check out your your article here on the comparison of the AI buildout and the railroad buildout in the US. Uh love to have you on again real soon and thanks so much for joining.

>> To be with you.