What Do The Midterms Mean For Crypto And AI?
Jason YanowitzRobMatthew Sigel
- The midterms create a possible fiscal tailwind for Bitcoin while adding obstacles to AI infrastructure growth. Rob frames Democratic control as “a feature for Bitcoin and a bug for the AI trade,” and Matthew Sigel agrees: divided government might mean less austerity, more spending and stronger demand for hard assets. Rob remains skeptical of a lasting decoupling, expecting Bitcoin to stay closely linked to Nasdaq and liquidity over the next 6–9 months.
- AI’s selloff is exposing how much of the growth story depends on continued access to capital. Sigel says Texas represents 30% of the AI data-center pipeline; tighter credit pushes projects toward additional equity, making dilution part of the repricing. Rob’s concern is the interconnected revenue “daisy chain”: companies with negative unit economics need financing to keep growing, and “if one part doesn’t work, another part of the stack doesn’t work.”
- Sigel sees distressed value in data-center landlords, with substantially more risk in GPU-owning neo-clouds. His buying case rests on long leases, mostly investment-grade counterparty backing and valuations assigning no terminal value to expensive facilities. Buying the GPUs makes the investment five times more expensive while shortening contracts from 15 years to three; Galaxy, he stresses, is “a landlord.”
- Open-weight adoption is challenging the assumption that frontier models capture every valuable AI workflow. Rob sees startups moving away from frontier models, while Yanowitz argues that proprietary context, reliability and distribution increasingly determine the edge. Sigel counters that enterprise trust and customer retention still matter: VanEck needs “a legitimate institution,” and aggregate compute demand can grow despite shifting model shares.
- Stablecoin payments can expand rapidly without equivalent growth in outstanding supply. Rob says velocity has risen roughly 3× since GENIUS passed, while payments have grown around 25–35% quarter over quarter, with some quarters near 50%. His mechanism is faster settlement and less capital tied up on balance sheets; Visa’s direct stablecoin settlement has reached $25 billion, versus roughly $7 billion a year earlier.
- Tokenization adoption leaves the question of investor returns unresolved. Yanowitz puts 95% conviction on major capital markets moving onto public blockchains, but Sigel sees competition, pricing pressure and fragmented liquidity rather than obvious “winner-take-most” economics. His preferred exposure is Figure: he cites 50% margins, 100% annual growth and roughly 20× earnings, while warning that Circle’s expectations remain demanding.
- Prediction markets’ strongest argument is rapid information aggregation, with political scrutiny likely to intensify. Rob argues that polls, models and private analysis all feed into prices, allowing markets to recognize likely outcomes before conventional coverage. He says approximately 95–98% of Polymarket visitors do not trade, making information distribution a broader benefit, and expects prediction markets to rank among Washington’s hottest topics next year.
1. The midterms expose AI’s dependence on financing
Yanowitz timestamps the discussion at November 5, 12:06 p.m. Eastern. The previous rebound had taken Bitcoin above $80,000 and lifted the Dow 3%; now Nasdaq is down roughly 4–6% intraday, with Nvidia, Broadcom, Qualcomm, Nebius and CoreWeave under pressure. The initial celebration of gridlock has given way to a reassessment of the AI buildout.
Sigel identifies the Texas governor’s race as the surprise: 30% of the AI data-center pipeline sits in Texas. Building that entire pipeline would imply $5 trillion of investment and nearly a million jobs. He immediately qualifies the scale—“90% of data centers don’t get built”—but still sees roughly $1 trillion and 100,000 jobs at stake after discounting the headline opportunity.
Tighter credit means projects require more equity, so investors discount potential dilution. Sigel says much capital formation is levered off Anthropic’s balance sheet; expectations that it would receive investment-grade treatment now face Financial Times reporting that it “might be junk-rated.” He is also hearing about two hedge-fund liquidations, without knowing whether Situational Awareness is involved, and sees no specific morning news explaining the entire move.
Rob’s concern is structural: much of the AI stack has negative unit economics, and its participants’ revenues are interconnected. “The correlation is somewhat close to one.” Separately, Yanowitz worries that generalist VCs may struggle to raise more LP capital to bridge financing gaps. He estimates around 200 AI companies could need substantial new VC funding if they cannot secure GPU supply or lower compute costs.
2. Long leases support the landlord trade, while GPU ownership adds risk
Sigel’s positioning favors Bitcoin miners and landlords, where mostly investment-grade-backed, 10-year leases provide unusual valuation support in digital assets. Conservative capitalization rates and DCFs suggest the market assigns no terminal value to multibillion-dollar facilities. His challenge to long-term investors: “If you’re not buying from forced sellers, what’s the point of having that longer-term outlook?”
The distinction matters when Yanowitz calls Galaxy a neo-cloud: “They’re a landlord.” Sigel says purchasing GPUs makes the investment five times more expensive and shrinks contracts from 15 years to three, worsening debt and liability mismatches. If credit fails to improve, he floats a reversal of the attempted CoreWeave–Core Scientific combination: “It could be Core that ends up owning CoreWeave.”
Rob’s pushback is that customers saving money with AI does not establish sustainable economics for their suppliers. Sigel points to roughly $1 trillion of annual hyperscaler operating expenses: 10% efficiency would save $100 billion, alongside Microsoft’s buyout offer to around 25% of workers. He also cites NextDC’s announced take-private and 50% stock jump as evidence of underlying value, while conceding that the unwind will take time.
3. Open weights shift the moat debate toward context and trust
Rob sees companies, including later-stage startups, replacing frontier models with open weights for many workflows; he cites Coinbase and Ramp discussing the shift. Yanowitz sees the same at Blockworks: once reasoning becomes sufficiently capable, the bottleneck becomes domain context, reliability and latency. The durable advantage may move toward “proprietary data, proprietary workflows and distribution advantages.”
Sigel tests that argument against Applied Digital’s North Dakota campus, where Meta is an anchor tenant. A remote location could become problematic if training demand gives way to latency-sensitive inference. Meta, he says, values switching between training and inference within an integrated facility; improving technology and advertising also make “making the customer wait a second or two” less consequential.
Enterprise purchasing is another constraint on substitution. VanEck will not simply contract with an open-source wrapper: “We need a legitimate institution.” Sigel sees the moat in retaining customers through a harness that permits cheaper models without surrendering the relationship, and thinks Anthropic is handling that reasonably well. Rob counters that trusted routers, secure execution and US open-weight models could eventually accommodate those requirements.
Rob’s concession is that large enterprises remain early in adoption, sometimes still working out their Copilot deployments. Startup substitution and enterprise expansion can therefore happen together. Sigel agrees with that uncertainty but continues watching overall token demand, GPU hourly prices, available supply and adopters’ costs rather than monthly market-share changes. As Rob puts it, “none of us can actually answer this question today.”
4. Bitcoin’s fiscal argument still competes with its liquidity exposure
Sigel rejects Yanowitz’s suggestion that there are no sellers left: “There are always sellers left.” Bitcoin and gold’s relative strength might instead reflect lower chances of austerity under divided government. With spending authorization approaching in about a month, he expects compromise to mean more spending, potentially weakening the dollar. He agrees with Rob’s proposed framing of Democratic control as “a feature for Bitcoin and a bug for the AI trade.”
Rob expects Bitcoin–Nasdaq correlation to remain high over the next 6–9 months because liquidity and capital-market uncertainty still dominate. Sigel adds that Bitcoin has tracked software particularly closely, invoking AI’s disruption of software and ability to find vulnerabilities against Bitcoin’s slower upgrade cycle. His fiscal bull case retains a clear limit: tighter credit generally hurts everything; even gold usually falls, just less.
Rob finds the Texas conditions relatively reasonable: union jobs, community consultation, transparent agreements, protection against higher electricity and water prices, water neutrality and removal of tax subsidies. Stronger balance sheets could still build. Sigel separately suspects Pennsylvania’s Governor Shapiro might restart permitting around inauguration; both the identity behind the reported discussion and the timing are explicitly his “hunch.”
Asked to allocate a hypothetical $100, Sigel first asks about wealth, horizon and suitability. His own portfolio is approximately 80% SPY, with the remainder divided among Bitcoin, his own fund and a little venture. His conviction is that monetary debasement continues, encouraging Bitcoin adoption and greater appreciation of the power infrastructure supporting both Bitcoin and AI.
5. Stablecoin turnover can grow while trading demand contracts
Yanowitz leaves Galaxy’s institutional gathering with 95% conviction that stocks, bonds, currencies, commodities and capital markets move onto public blockchains: “the boulder is rolling down the hill.” Sigel challenges that certainty with falling stablecoin AUM. Rob’s first correction is that stablecoins plus cash equivalents, including Treasuries, are slightly up over the quarter, although not by much.
Rob says stablecoin velocity has increased roughly 3× since GENIUS passed, with payments growing around 25–35% quarter over quarter and some quarters around 50%. Faster settlement means businesses need fewer idle balances: moving from T+2 toward instantaneous settlement pulls credit risk out of the system and releases capital. His broader analogy is that stablecoins are “digital oil for tokenized assets,” supplying the prerequisite for a larger economy.
Visa’s direct stablecoin settlement illustrates the progression: $25 billion now, roughly $7 billion a year earlier and basically zero six months before that. Rob also points to Rain, where his firm is an investor, and its pipeline of very large companies. These customers want quicker cross-border settlement, less balance-sheet capital and multicurrency access; blockchain enthusiasm is incidental to those business needs.
Weak spot trading explains much of the subdued issuance. Rob is not sure he saw a single company generating $10 million of genuine stablecoin-payments revenue before 2024; earlier demand overwhelmingly served crypto trading. Leveraged perpetuals need less collateral, spot volumes are sharply lower, and the memecoin trade becomes “a circle that keeps getting smaller every time.” Payments growth is emerging against that contracting trading base.
6. Figure offers Sigel clearer economics than the broader tokenization trade
Sigel thinks Circle is a winner in an uncertain market, but next year’s estimates assume substantial USDC growth with little margin contraction. Current margin pressure makes those numbers difficult “unless something changes dramatically.” Rob says Circle is gaining institutional share and highlights USDC’s early payment use cases, including Visa’s support, while arguing that much of the Street does not understand the demand developing inside these businesses.
Figure is Sigel’s largest position in this group. Its chosen market is bespoke and institutional: helping lenders originate home-equity lines and first-lien mortgages, then hosting trading before and after funding. He likes making the blockchain the regulator-accepted “source of truth.” His cited operating evidence is 50% margins and 100% year-over-year growth, rather than a generalized promise that everything will become tokenized.
Sigel uses the rule of 40 to combine growth and profitability. He says two companies globally score above 140, naming Palantir, and contrasts its roughly 100× earnings multiple with Figure’s 20×. His Figure exposure exceeds Circle by more than 5 to 3. Venture shareholders may still be rotating out. Meanwhile, he is avoiding money-losing small caps such as BitGo “for now,” believing they need an altcoin cycle to generate profits.
7. Prediction markets aggregate information faster than the debate credits
Rob recalls Polymarket getting roughly five outcomes wrong across more than 100 election markets, with Mississippi among the surprises; he says Kalshi did similarly. Pollsters’ improved accuracy does not undermine that result. Treating polls and prediction markets as competing systems misses the mechanism: polls, models, bespoke analysis and other information become inputs into trading.
The more valuable function, in Rob’s account, is identifying likely outcomes before news sites or conventional models acknowledge them. Prices aggregate what participants know and disseminate the result rapidly. He cites approximately 95–98% of Polymarket visitors as nontraders, with something similar at Kalshi: the audience consuming the information extends far beyond the people placing bets.
Rob is buying Anthropic “no IPO” markets and would be shocked by a listing before November 30; Sigel notes that year-end IPO odds nevertheless remain 76%, distinguishing the deadlines. Politically, Rob expects prediction markets to become a major Washington subject next year—potentially third after Trump corruption and AI. He argues that their information benefit supports their continued existence.
8. Tokenized equities still lack an obvious winner-take-most business
Yanowitz points to tokenized-equity activity rising 400–500× from a small base, ICE’s investment in Trade.xyz and Citadel’s lawsuit. He thinks issuer opt-outs have been overstated: some companies are effectively asking for 30 or 60 days to decide. For him, the activity supports the adoption story; the unresolved question remains whether value reaches tokens or company equity.
Sigel’s objection is investability. Tokenized listings require balance sheet and can duplicate or fragment liquidity; “ETFs are actually the OG tokens.” More access benefits consumers, particularly overseas, but competing products obscure pricing power. He holds relatively small Circle and Coinbase positions, little Robinhood, and finds Interactive Brokers and Schwab interesting at current prices; VanEck’s venture investments, including Dinari, provide a different route into potential infrastructure winners.
Rob shares the skepticism about today’s implementation despite long-term optimism: “I’m not sure venture is making any money.” The visible benefits may accrue mainly to issuers receiving wider distribution and platforms receiving trading volume. That leaves a meaningful gap between believing markets become less geographically constrained and identifying which current tokenization businesses retain the economics.
Full transcript
We’ve subbed out Santi. Actually, a lot of big changes today. We ditched Hasim. He had a good one-episode run, but episode two wasn’t really cutting it. Thankfully, we got Rob back.
Rob’s back. He’s returned.
Rob, welcome back. We missed you. And we’ve subbed out Santi today. Santi is nursing his leveraged long Micron bags. We’ve got the head of digital assets research and friend of the pod, Matthew Sigel from VanEck. Matthew, welcome to the show.
Gentlemen, good to be with you.
Great to have you here. I have to timestamp this because the market is moving so much today. It is Thursday, November 5th, at 12:06 p.m. Eastern. I say that because by the time we end recording, the market will be in a different place.
We’re going to skip all the niceties and early stuff. I came in ready to talk midterms, but we have to talk markets, which are obviously intertwined with that. It’s the same conversation. We had the midterms, markets took a dip, and then there was a sharp rebound. Bitcoin got pushed above $80,000, stocks ripped higher, and the Dow was up 3%. Everyone said the market actually loves gridlock, gridlock is good for the markets, and the Democrats are going to pump money into the markets. Things are good; the market will be fine.
Then today, November 5th, we woke up to a complete reversal. The Nasdaq is getting absolutely smoked, down 4% to 6% intraday, led by semis and AI chipmakers: Nvidia, Broadcom, Qualcomm. All the neoclouds are getting smoked too: Nebius, CoreWeave. Everything is getting completely smoked right now.
Noted: neocloud Galaxy is also getting smoked?
I was at the Galaxy event in New Orleans last week. I’ll share some takeaways from that later in the pod. Matthew, what are you thinking when you see all this? Give us a brain dump.
The market is repricing AI capex in a big way. The Texas governor was the big surprise for the markets. Thirty percent of the AI data center pipeline is in Texas, and we did some work on it. If that whole pipeline were built, it would mean $5 trillion of investment and nearly 1 million jobs.
At any given point, 90% of data centers don’t get built. But even if you handicap that number, you’re talking about $1 trillion of investment and 100,000 jobs. I think today the market is waking up to that economic impact. There’s probably downside to Nvidia estimates if they literally can’t find data center space to plug their chips into.
At the same time, credit markets are reasonably tight, so a lot of projects are demanding additional equity on the margin instead of debt. That raises financing concerns. It’s an unwind for sure. We’re hearing there are two hedge fund liquidations today, so people are forced sellers in the market, even though there wasn’t any specific news this morning that would have prompted this.
When you wake up to this type of tape, you start to think about whether Anthropic can get out at all, and at what valuation. A lot of the capital formation is levered off its balance sheet, which people first thought would be investment grade. Now the Financial Times is reporting it might be junk-rated. All of that requires the mix of equity to debt on these build-outs to shift more toward equity on the margin. That’s crushing the stocks as they discount potential dilution.
The likelihood of Anthropic going public is getting smoked on Polymarket too. The odds of an Anthropic IPO by each date are down between 10% and 20% each day. Rob, what’s your take when you see all this?
My view isn’t dissimilar to Matthew’s. I’ve been hammering the Polymarket Anthropic no-IPO market because I’ve generally thought—and we’ve been talking about this a bunch internally recently—that the markets were already jittery in October. A number of those big AI names were down 10% to 20% on the month. There was a lot of conversation about whether this was just positioning for the worst case and what would happen once we got some clarity.
The markets were up yesterday, which seemed pretty clearly to me to be short covering. I think we saw quite a bit of that, but it was hard to understand exactly what was happening. Following today, I think we’re probably in for a situation where we’ll have to think through the terminal growth of a lot of these AI-specific companies and how they get there, as well as how the financing markets react.
We’ve talked a lot on this podcast about the daisy chain of revenue within AI right now. They’re all essentially interconnected, and the correlation is somewhat close to one. If one part doesn’t work, another part of the stack doesn’t work, and they’re all likely to go down together.
They require a significant amount of capital because almost every part of the AI stack today has negative unit economics. Very little of the AI stack can survive on its own with the current projected growth rates, debt outstanding, interest payments, and expected demand. If equity and debt markets continue to tighten and parts of the market can’t grow themselves out of negative economics, we’ll see broader risk-off behavior.
That will affect crypto and other parts of the equity market. There’ll be some rotation, obviously. I don’t think people are going to get entirely out of equities or go entirely risk-off, but there won’t be as much capital going into the tech trade as there has been over the last year or year and a half.
1. Buy The Dip Or Bear Market?
Matthew, when you look at a beating like this, I think the main question everybody wants answered is: buy the dip, or the beginning of a correction or bear market? Setting crypto aside and focusing on AI stocks, how would you answer that question? Or, Rob?
I also want to put a point on that, Matthew. You mentioned a couple of hedge funds that look like they’re unraveling. We’ve been through this in July with Situational Awareness. I wonder if they’re one of the funds you were talking about, and how you expect that to play out and when it ends, on top of Yano’s point.
I don’t know about Situational Awareness specifically. It depends on your individual circumstances. I don’t take on leverage; I’m long-only. If you’re not buying from forced sellers, what’s the point of having that longer-term outlook?
I’m pretty concentrated right now in Bitcoin miners and landlords. It’s very rare to have this much valuation support anywhere in digital assets. You can look at these leases—they’re 10-year leases, and for the most part they have investment-grade counterparties backing them. You can use a very conservative capitalization rate and look at a DCF, and the market is pricing in no terminal value for these multibillion-dollar data centers.
To be negative at these prices, I think you have to assume that a bunch of white elephants are being built. We haven’t seen any evidence that’s the case in terms of the demand profile or what’s happening to GPU prices. I’d push back a little on the unit economics point because you can see that the companies most focused on taking costs out of their businesses are adopting AI.
There’s about $1 trillion of hyperscaler opex every year. If you can get 10% efficiencies, you’re talking about $100 billion in savings. Microsoft just offered that big buyout to about 25% of its workers. That’s a separate question, but in a year, I think we’re going to look back and see historical value in some of these names.
2. Can AI Companies Survive?
NextDC, an Australian data center company, announced a deal overnight, and its stock is up 50% as it’s being taken private. It’ll take time for that to shake out, obviously. But if you’re not levered and you can take a longer-term horizon, then in my personal opinion, this is more of a day for buying than for selling. The valuations are very different from bubbles I’ve observed in my career.
Yeah, but Matthew, you made a separate point: Microsoft is talking about buying out some of its employees because it can save money through AI efficiencies. But most of the companies offering this, outside the large hyperscalers, are running at negative unit economics.
There’s been a lot of talk about Anthropic and OpenAI reaching a point where they might become cash flow positive and operate on their own. Even the big neoclouds, like CoreWeave, can’t meet the demand they say they’ll meet unless they keep raising capital to cover their interest expense. If the debt or equity markets seize up, they won’t be able to meet that demand.
At the next level down, this is why you saw some of the conversation yesterday about what’s happening in Texas and other states: maybe we should all pile into the big hyperscalers because the startups and Series C and Series D companies won’t be able to compete anymore. There’s a conversation about the public markets versus the broader AI spectrum, and frontier models versus large VC-backed companies. But almost all of those private companies, outside of maybe Anthropic, are burning significant amounts of cash to grow at their current pace. If the VC dollars keep coming, they’re fine. If they keep getting equity, they can hopefully grow out of it.
But I’d bet there are something like 200 AI companies that have raised significant amounts of money and will be unable to operate without a lot more VC funding if they can’t get GPU supply, GPU prices go up, or compute costs don’t come down. The question downstream is how LPs look at what’s happening in the public markets. Do they become more risk-off? Every big generalist fund is raising a ton of capital by telling LPs, “We’re in a new paradigm. The debt markets are tightening, and there may be less interest in the equity market, so we can see the future better than those risk-off credit providers. Give us more LP capital so we can support this gap in the market.”
If LPs stop believing that story, there’s a huge gap in the market for additional equity capital. That’s the perspective from my seat as a private investor, not a public investor. It’s also connected to what worries me about the daisy chain we’ve talked about before.
3. Content Of The Week
Yeah, I think that’s a fair bear case. When you talk to CoreWeave and the other neos, the majority of their token demand isn’t coming from VC-funded companies. What you’re describing is always part of the VC lifecycle: most companies fail, and the strong get stronger. I’m not denying there’ll be lumpiness here. But with the physical constraints around plugging in intelligence, it’ll go to the highest bidders, which will be the guys with strong balance sheets. I agree there’ll be lumpiness around that.
4. Are Frontier Models Losing Their Edge?
But when we look at S&P 500 company transcripts and surveys of purchasing managers and hiring managers, what they’re saying about where they are in their AI adoption lifecycle is that it’s really early. We were at a conference, and we asked the audience whether they’d feel disadvantaged if the colleague sitting next to them had access to AI and they didn’t. Every hand went up. With valuations like this, at least in some of my names, you have to zoom out and look at the long term. Otherwise, you’re selling when everyone who’s levered and has a short-term horizon is selling, and that’s generally the wrong move.
Can I ask you a question about that? One thing we’re seeing across a lot of our companies and startups, even up to Series D and Series E, is that they’re switching away from frontier models to open-weight models for most workflows. They no longer feel they need frontier models for the workflows they’re doing. You’ve heard Coinbase, Ramp, and a lot of large companies talk about this. How does that affect your view of what’s happening more broadly at Anthropic and OpenAI, their ability to get to the public markets, and the neoclouds?
Nearer term, Anthropic needs to show momentum for its IPO to go well. You noted that the betting odds are still 76% that the IPO gets out by the end of the year. That has come down a bit, but it’s still relatively high.
Yann LeCun often says hyperscalers take turns outperforming. I think you’ll see that in the AI space as well. The pod shops and short-term guys are going to hyperventilate about every month-to-month change in market share. But then you zoom out and look at overall token demand, what’s happening to GPU hourly pricing, what supply can come online, and what early adopters are doing with their costs. We just haven’t lost conviction in that story.
I’m curious how you guys think about this. Rob, one interesting thing you said is that teams internally are no longer using frontier models. We see this at Blockworks. The reason you used to lean on a frontier model was that the scarce resource was raw model capability: the ability to reason across messy, unstructured information. Everyone would race to the latest model, and every couple of months it would shift between Claude, ChatGPT, and sometimes Gemini.
But once you have a sufficiently strong reasoning layer, the real constraint becomes domain-specific context, reliability, and latency. Given the conversation you guys are having, I’m curious how that affects the market and the stocks. Matthew, you think the neoclouds will come back and that this is a buying opportunity. But in the founder—
I just want to distinguish something. When you called out Galaxy at the beginning of the show, you called them a neocloud. They’re a landlord; they’re not buying the GPUs. Once you layer in buying the GPUs, it’s five times more expensive, and your contract length shrinks from 15 years to 3 years. The liability mismatch in some of the neos and their debt levels are definitely problematic.
Remember, a year ago CoreWeave tried to buy Core Scientific, and Core Scientific revolted and said the offer wasn’t high enough. The tables may turn here in a year if the credit markets don’t improve.
It could be Core that ends up owning CoreWeave. I just wanted to distinguish between the neoclouds and landlords, where I see a lot more risk in the neoclouds.
Yeah, that’s a good point. My question to you guys: I’m not in the investor seat; I’m sitting in the founder seat, but I see how we’re spending on AI. The last cycle was dominated by the idea that the best general models plus unlimited compute would win everything. That justified a lot of the neocloud buildout with CoreWeave, Lambda, and the rest, because everyone needed these massive, undifferentiated GPU clusters to train and serve these models.
But the second the market starts questioning the pace of the buildout—as Rob mentioned with Coinbase, we see it internally too, and we’re doing this as well—the durable edge may move to whoever has proprietary data, proprietary workflows, and distribution advantages. Maybe it moves away from unlimited compute. I’m not saying it never comes back, but maybe we’re moving away from the unlimited-compute world.
Maybe. We just went out and visited Applied Digital’s mega campus in North Dakota, where Meta is one of the anchor tenants. One of the big debates on a stock like that is that they’re building in a region that isn’t near any major city. Will they be able to serve inference? If you’re right and the training mix goes down a lot, will they be left with a white elephant with no value?
But if you talk to Meta, they say they like having that integrated facility where they can switch back and forth between training and inference. Technology is improving. Frankly, if you’re going to start inserting ads in your AI service, which the hyperscalers clearly are doing, making the customer wait a second or two isn’t that big a deal. They’re watching an ad in the meantime.
I’m not so sure about that. The moat is going to be holding onto the customer and offering a harness that gives them just enough flexibility to take advantage of open-source, lower-cost models without losing them. I think Anthropic is managing that situation reasonably well from a unit economics perspective. An enterprise like VanEck isn’t going to contract with some open-source wrapper. We need a legitimate institution.
I’m just not so sure that the switching costs are that low, or that these training investments are going to be worth less than zero, which is how the stock market is valuing them right now.
I think what’s interesting is that there are a bunch of variables here, and none of us can actually answer this question today. Clearly, there are a bunch of different perspectives. In what I’m seeing, I think most startups have already started to move towards open-weight models.
You’re completely right that VanEck probably isn’t going to use a Chinese open-weight model, and I’d expect a lot of large enterprises won’t do that. It starts earlier on because startups are a little more cost-conscious, a little less concerned, and more willing to test things out. They’re willing to move more quickly and keep trying things out. They may not be as concerned about some of the data security implications of using an open-weight model from another country as somebody like VanEck is.
There are a lot of companies building harnesses and routers that maybe VanEck would trust. Those could theoretically allow a secure, trusted execution environment alongside an open-weight model. They might not want to use a Chinese model, but we’re almost certainly going to see—and are already seeing—U.S. open-weight models as well.
That’s the barricade for some of the frontier models and some of the downstream effects. At the same time, you made the point earlier that there are a ton of large companies where AI isn’t that penetrated yet. Even where there is a lot of penetration, people are still using Copilot and trying to figure out exactly what their stack looks like.
There are variables on both sides, and I think the unknown is so large. That’s part of the reason there’s as much jitteriness as there is. I don’t know if you agree with that.
Yeah, I agree with that. Not much to add. Fair take.
5. Can Bitcoin Weather The AI Selloff?
Yeah. Guys, let me take us into crypto a little bit. I would say crypto has held up surprisingly well, considering everything that has happened: an absolute market beating in AI, midterms maybe not going the way crypto folks wanted, and the biggest hack of all time at Binance. This blue-wave election tsunami.
My take is that Matt Hougan had a great tweet on this: There are no sellers left in crypto. I’m very curious to hear from both of you. Matthew, maybe you want to start with your general take. We just talked about the impact of the midterms on AI. What’s the longer-term impact of the midterms on crypto markets?
There are always sellers left in crypto, so I’m not sure I agree with that one. Gold and Bitcoin both outperformed yesterday, with a little bit of Bitcoin-Nasdaq decoupling. I think that might have reflected lower chances of austerity in this divided government. It’s one thing to campaign, then you have to govern, and governing is about making compromises.
There’s a compromise coming up in about a month, when Congress has to authorize next year’s spending. The fiscally austere part of the Republican Party is basically decimated after this election. A reasonable base case is that Trump compromises with Democrats, which means more spending—and more spending on nonsense that alienates a good part of the electorate and some of our trading partners. The dollar was down, and BTC and gold both outperformed.
That would be the medium-term bull case. It would honestly be quite good for the fever to break a little, since BTC-Nasdaq correlation has been so high over the last couple of weeks. The bear case is that Bitcoin’s correlation is still pretty linked with liquidity. To the extent that credit markets tighten up, it’s generally good for nobody. Gold usually goes down the least, but it still goes down. That’s what I think the bulls and bears are grappling with here.
6. Fidelity, Blockworks MCP, DAS London
Rob, do you have a thought here?
I don’t know if I have much to add. To Matthew’s point, there are always sellers. There could be more capitulation. I struggle to see a world where the correlation between Bitcoin and the Nasdaq doesn’t stay pretty high for the foreseeable future.
I think it continues to be very linked to liquidity. A lot of the excitement for crypto recently has actually been outside the core crypto communities. You’ve seen more AI founders talking about crypto, and more of the tech community outside the core crypto crowd talking about it recently, especially when it comes to some of the AI-crypto crossover work that’s being done—particularly stablecoins and what we’re seeing happen in fintech with tokenization.
Insofar as all of that fits within a bucket, more tokenization, more stablecoin transfers, and more excitement for things like Pearl and others we’ve talked about would continue to increase liquidity and interest in Bitcoin. It strikes me that we’re probably going to remain highly correlated to the Nasdaq and to continued buildout across these new tech ecosystems and new areas of innovation.
7. Democrats Take Aim At Data Centers
All that said, I think the pace of change is still very high, and the uncertainty is still very high. Capital markets are generally teetering on the edge, and you’re seeing—and will continue to see—a lot more volatility because of that. I think Bitcoin is going to be linked to that, and I don’t think that changes over the next 6–9 months, as far as I can tell.
We need to see what the next Congress is going to look like. One congresswoman came out today and said we need to ban superintelligence and get out in front of AI. I think she might have deleted the tweet, but we need to see exactly what the consolidated view of the Democrats will be when Hakeem Jeffries takes over as speaker of the House and the new Congress starts meeting, and what that means for the future of these buildouts.
We also need to see what happens in some of these states. Matthew talked earlier about the governor’s race in Texas. That had a lot of people concerned. I think it’s Gina Hinojosa, if I’m saying her name right—I’m blanking on it—but she put out a framework for how she would think about data center buildouts in Texas. She said she wants to require union jobs, make sure there’s real engagement with local communities through town halls and things like that, and ensure full transparency so these buildouts aren’t done under NDAs. She wants to guarantee that electricity and water prices won’t go up, and make sure data centers can be water-neutral, which many already are today. She also wants to take away the tax subsidies.
If you put that list together and say, “This is what Texans need to allow more data center buildouts in Texas,” that strikes me as relatively reasonable. A tax subsidy is more about states competing with one another than something you need in order to build there. And, to Matthew’s point, people with stronger balance sheets can still build without the tax subsidies.
If the point is, “We want to continue to allow this buildout, but we want to create a framework that we think is better for local communities,” then I think the market won’t be as concerned, the correlation will probably continue, and Bitcoin will hold up as well. But Matthew, I don’t know if you agree with any of that, as someone who spends more time on these products.
The Wall Street Journal had a really interesting piece today. Apparently, one Democratic governor is already saying behind the scenes that they need to restart data center construction to save the jobs. I have a hunch that was Governor Shapiro in Pennsylvania, so he could start permits right now. I have a hunch that’s going to happen around the inauguration.
As for the Texas governor, everything she put out there was reasonable, except there was this targeting of men who own the data centers. I was like, it’s actually shareholders who own this, and some of these billionaires’ wives may have a claim on these assets too. I wasn’t sure about that part.
As for the Bitcoin–Nasdaq correlation, it’s specifically software that BTC has been highly correlated with. If you look at the semiconductor-to-software ratio and track that against Bitcoin, Bitcoin tracks the software side. That makes sense to me from a fundamental perspective. I know you didn’t want to get into hacks, but I can’t help it: AI is disrupting a lot of software, lowering barriers to entry in software, and making vulnerabilities easier to find. BTC is perhaps more exposed given its slower upgrade cycle.
I wasn’t saying the BTC–Nasdaq correlation will collapse, although it does tend to oscillate quite a bit. It would obviously be bullish if it fell and moved more with gold, I agree.
8. Is Gridlock Bullish For Bitcoin?
Let me throw out an idea, Matthew, and you tell me whether you agree or disagree: Democratic control here would be a feature for Bitcoin and a bug for the AI trade, because more spending and gridlock tend to be a classic hard-asset setup. At the same time, you may have state-level resistance to data centers, which ends up being a direct tax on the hyperscaler growth story that has been carrying the Nasdaq. I know everything is correlated, but perhaps this is how you get some uncorrelated results here. Agree or disagree?
I agree with that take.
A client calls you and says, “Hey, Matthew, I’ve got 100 bucks. How do I allocate that today, considering the market we’re in?” How do you think about that?
How much money do they have? What’s their investment horizon? I’m a suitability guy here. No, but in all seriousness, my own personal allocation is very heavy in SPY, essentially 80%. I split the rest between Bitcoin, my own fund, and a little bit of venture.
I have pretty strong conviction that monetary debasement will continue, and that the secondary effect will be rising Bitcoin adoption and more appreciation for the underlying power infrastructure that’s crucial to both Bitcoin and the AI buildout. That’s how I’m exposed, but older people or people who need income will obviously want to change those allocations.
Yeah, look, you two are the better investors here, but I tend to be, as Rob knows, very left of the curve on the bell curve. I’m a very smooth-brained investor.
Maybe that makes you the better one.
The way my brain thinks after the midterms is basically: blue Texas means the AI trade is less attractive. Debt isn’t going anywhere. Money printing only goes one way. Buy Bitcoin.
9. Wall Street Moves Onchain
Maybe this is a good time to talk about the Galaxy event. I’m not sure how much I’m supposed to share or allowed to share from it, but Galaxy Digital hosted this fantastic event in New Orleans this week. You go to an event like that, and there are C-suite executives from every single bank and financial institution. The ball is rolling down a really steep hill. We spent 15 years getting a boulder up a hill, and now the boulder is rolling down the hill. We’re probably 1 year into the boulder rolling down, and you just can’t stop the crypto trade. That’s my belief.
I’d put it at 95% conviction that stocks, bonds, currencies, commodities, and capital markets move onto public blockchains. I think the question for the investor listening to this is: Does that go into the equity of private companies building in crypto, or does it go into tokens? That was the clearly obvious thing from this Galaxy event: everyone thinks money and markets are moving onto blockchains. We can agree on that. I don’t think it’s easy to take the other side of it, but—
I’ll take the other side. I’m curious about Rob’s view on this. Stablecoin AUM is down pretty considerably this quarter, and maybe some of that is going into the bond market because yields are higher. But I think it’s something that should be asked and answered: Are we in this super-adoption phase of tokenization? I think that’s up in the air.
ETFs are actually the OG tokens. Look at the capital formation there, and at prediction markets and sports betting markets on ETFs, and the duplicative capital involved in putting things on-chain. I think there are still a lot of open questions there. My position sizes in those companies are smaller than in areas where the constraints are more visible to me and there’s pricing power, which is on the infrastructure side.
Obviously, that’s the wrong trade for today, but that’s how I feel about the conviction on the 30,000-foot view. Individually, where’s the pricing power? For tokenization, a lot of it is in the consumer’s hands, and everyone is cutting prices. You can see that across the board. We looked at all the exchanges and brokers that reported this quarter, and it’s all about pricing pressure and competitive issues. Now perps are competing, and CME is suing its own regulator. There are a lot of open questions about that tokenization trend.
10. Why Stablecoin Adoption Looks Different
Going back to the simple stablecoin question, why is market cap falling?
Yeah, I have a perspective on this. If you look at stablecoins plus cash equivalents, like Treasuries and other tokenized cash equivalents, it’s slightly up over the last quarter—not a lot, but slightly. But you have to put more than just stablecoins on one side of that. What you’ve actually seen since the GENIUS Act was passed is that the velocity of stablecoins has gone up something like 3x.
If you look at the data that tries to separate trading in stablecoins from payments in stablecoins, payments have grown something like 25% to 35% quarter over quarter since then. For a couple of quarters, it was about 50%. I think part of what you’re seeing is that, if this is money that moves with instantaneous settlement, you don’t need anything near the M1 or M2 supply to run a business on top of it. You shouldn’t need to hold a ton of it, because you’re pulling a lot of credit risk out of the system. You’re going from T+2 to instantaneous settlement. This is even more pronounced when you think about cross-border payments, where we’re seeing a lot of pickup.
You’re seeing a little of this in the numbers. It’s still small, but Visa came out and said, “We are now doing $25 billion of direct settlement in stablecoins.” That number was, I think, $7 billion about a year ago, and basically zero in the six months before that. That’s when it started growing really quickly.
We’ve talked a lot about what Stripe has been doing and the way they’ve been building their business. I think about it from a Rain perspective, because we’re investors there. I look at the people in their pipeline now, and it’s many of the biggest companies in the world. These companies are saying, “I realized now that I can go cross-border with less credit risk, quicker settlement, and less capital on my balance sheet. I can serve a more global user base in a multicurrency way that I couldn’t before.” Stablecoins are the tool to do that. They don’t really care about blockchains or crypto otherwise.
The crypto token markets have gotten crushed between the very extractive meme coin cycles and some of the hacks that have made people very concerned. Stablecoins were primarily used before to trade those things. As that trading activity has come way down, as you’ve seen in the Robinhood and Coinbase earnings and everyone else’s, the other side of the market has continued to develop around real use cases. That’s what we’re seeing in payments, and that’s where we’ve put a lot of our money.
On the tokenized asset side, I’ve said a lot that stablecoins are like digital oil for tokenized assets. You needed them before you could build the rest of the economy on top of them. Now, I think it’s a little more up in the air who accrues value, because obviously the DTCC is launching its product right now. You’re seeing a lot of these closed-loop systems trying to use tokenized assets. How does that actually look relative to what we’re seeing with tokenized equities on Backpack, Binance, Kamino, and Solana? It’s unclear how that plays a role in the rest of the crypto ecosystem, but it does seem to me that most people feel they can serve a global user base much more easily with tokenized assets.
I have this broader thesis that money and markets are going to be less geographically constrained over time, even as the countries themselves try to be more insular. I think AI is part of that, tokenized assets are part of that, and broader debasement is part of that, along with people wanting access to better-quality assets and money.
Rob, is it too much of an oversimplification to summarize all of that by saying there’s weak spot trading demand, and that’s why crypto prices have been soft for a lot of the year and spot volumes are at lows? When people aren’t actively buying that much crypto, they don’t mint new stables as aggressively.
I think that’s a big part of it. Historically, before the GENIUS Act, and really before 2024, stablecoin demand was different. 2024 was the first year we saw real payments in stablecoins. Obviously, a lot of startups started in 2020 and 2021 saying, “We’re going to…” USDC has been around since 2018, and they talked a lot about payments. There were a lot of proof of concepts that were talked about and hit the headlines. But if I look across the startup landscape and think about the actual data I used to see, I’m not sure I saw a single company do $10 million of revenue from real stablecoin payments before 2024.
All of the stablecoins outstanding and all of the volume before then was basically just trading activity for on-chain crypto assets. I don’t have the data in front of me, but more and more people have moved to perpetuals and leverage, and perpetuals continue to grow. That requires less collateral as people increase their leverage. I also think perpetuals volume is quite down from the highs of roughly a year ago, and spot volume is way down.
The fact of the matter is that a big part of why people wanted to trade crypto assets was volatility. They wanted outsized returns. That’s why they moved to perpetuals. The meme coin trade is a circle that keeps getting smaller every time, until it just washes everybody out, in my mind.
Yeah. All that’s fair, but if you look at street estimates for Circle next year, they assume significant growth in USDC outstanding and not much margin contraction. Then you look at the quarter they reported, and it’s margin pressure. They’re going to miss those numbers for next year unless something changes dramatically.
It could be the normal volatility of supply and demand, similar to what I mentioned on the infrastructure side. It’s going to be lumpy, but the expectations do seem pretty high from where I sat.
I will say that Circle is definitely taking market share in the institutional use case. Although, frankly, I think people are probably underselling how much success USDC is having very early on and how much Visa is pushing USDC for real payment use cases.
A lot of these sell-side analysts call me and ask for my perspective. I think there are some good ones out there, but broadly, the Street has absolutely no idea how to think about this market. They’re just not really in the know about what’s happening. They’re looking at a lot of headlines and not really understanding what drives most of this demand down in the trenches, personally.
11. Which Crypto Stocks Are Worth Buying?
Yeah. Matthew, do you have a view on the publicly traded crypto companies? Let’s put Coinbase and BitGo in there. We could throw Figure into that, and Circle too. Do you really like some of them or really dislike some of them?
Yeah, but I haven’t touched the money-losing small caps and won’t touch them for now. If you look at BitGo, they’re not making money, and they need an altcoin cycle, as far as I can tell, to really start making money.
Circle—I don’t think it’s crazy expensive, but, like I said, expectations are pretty high. There’s tons of competition. I think they’re a winner in a very uncertain space.
Figure is my largest position in that space. It’s more than 5 to 3 Figure versus Circle. Despite the smaller market cap, they’ve chosen to go after a very bespoke institutional market. They make it easier for other institutions to originate home equity lines of credit and first-lien mortgages, then host the marketplace for those institutions to trade those assets back and forth, both before they’re funded and afterward. If you add up their profit margins and their growth, those two metrics are often combined into something called the Rule of 40.
If you’re over 40 on your top-line growth plus profitability, you’re a reasonably healthy software company. There are 2 companies in the world above 140, and that’s Palantir. You pay 100 times earnings, and with Figure, you’re paying 20 times earnings.
It’s a complex story, but I really like starting with institutional bespoke markets, making the blockchain the source of truth, and having the regulator accept the blockchain as the source of truth. The stock is very cheap now. That’s not helping today; there are still a bunch of VCs in that name, and I think the shareholder base is probably turning over quite a bit. But that is my top pick in the tokenization space, where there’s not only product-market fit, but 50% margins and 100% year-over-year growth.
Yeah, interesting. Rob, we’ve made it 47 minutes, and you haven’t talked about Polymarket. I tip my cap to you, sir.
12. Did Prediction Markets Beat Pollsters?
I think I did say I was trading the Anthropic—no, I did. You got it. Yeah, but I teed that one up. I’d be shocked if Anthropic gets out by November 30th at this point, but—
As an unbiased participant here, I’d say: Who’s the winner, the pollsters, who were shockingly accurate, or the prediction markets, which were also shockingly accurate? We’ve got a biased participant here, but I’m curious to hear.
What I will say is that right after the election, on the night of the election, there was universal acceptance that the prediction markets did a great job again. Polymarket and Kalshi were very accurate. Maybe more important than being accurate, Polymarket got only about 5 of its more than 100 markets wrong. A couple of those were really shocking; everyone got Mississippi wrong. Kalshi was the same.
There was universal excitement: These guys did a really good job, and they’re clearly a very important source of information. Maybe the more important thing isn’t just that they’re an important source of information leading up to an event and a predictive measure. It’s also how they aggregate information as people trade, giving us a quicker read on the likely outcome and letting people react more quickly to new information. That information gets disseminated more quickly than it does elsewhere.
That was the bigger part. It was very clear on both Polymarket and Kalshi that some of these races were probably over before anyone on a news site or using a model would have told you they were over.
Then there was a little pushback in the last day or two: The pollsters were more accurate; they’ve updated their models and done a good job. That’s true. Or the models were more accurate, and prediction markets are just reacting to the models. I think that was the wrong conversation. I don’t understand why that argument happened, because at the end of the day, what are markets more broadly—not just prediction markets, but markets? They’re the aggregate input and output of all the information that goes into buying and selling.
13. The AI Trade Gets Crushed
All the different models, all the information, public or non-public, and all the bespoke analysis people have—whether it’s a public model or a poll—get put into that market. The output is everybody’s actions in response. I don’t think the argument that they’re a better predictor of what’s happening is the right way to frame it. They are great predictors when you look at the output, but the biggest thing is that they’re great aggregators of information, showing what the market thinks is likely to happen. They react more quickly than everything else, and there’s clearly a broad common good that comes from them, even if you aren’t a trader. You see this with both platforms. I think Polymarket’s stats show that 95% to 98% of people who visit the site aren’t traders, and it’s something similar for Kalshi.
Rob’s doing his best here to make sure he gets a subpoena from the Democratic Oversight Committee. I’m just joking, but obviously these markets are going to be in the hot seat over the next couple of years, I would think.
Yeah, listen, I think it’s very clear. I’ve spent a lot of time in D.C., and I was on the Digital Asset Market Subcommittee for the GMAC before it was disbanded. I think it’s very clearly true that prediction markets are going to be a hot topic, probably more of a hot topic than crypto or a lot of the other things. Trump corruption, AI, and prediction markets are probably number 3, to be honest, in terms of the things that will be talked about in D.C. next year.
I think it’s very obvious that these things need to continue to exist, and I think most market participants understand that there’s a common good here. I’m always happy to talk to anybody in D.C. about this going forward. I already talk to a lot of people there about it.
14. Who Wins From Tokenized Equities?
One thing to cover here is that, quietly, it was a very interesting week or two for tokenized equities. You had stock perps coming on-chain, Citadel introducing this lawsuit, and, last week or two weeks ago, ICE investing in Trade.xyz, which we already talked about. You had all these companies opting out of tokenizing their equity, which I think was overblown because they’re really just saying, “Don’t do it right now. Give us 30 or 60 days to make this decision.”
Matthew, I feel like you’re a good contrarian for some of my usually overly bullish crypto takes. I’d be very curious to hear your take on tokenized equities. Stablecoin volumes and supply have been flat, but tokenized equities are up 400 or 500 times, granted from a small starting point. I’d be very curious to hear your take.
Maybe some backdrop: The way VanEck is approaching this space from an active perspective is that we have our venture fund, with Wyatt and Juan running it and writing checks to founders up to Series A. Tokenization is a big part of their thesis, and they’ve underwritten some of the critical infrastructure that has potential if this space takes off, companies like Dinari.
Then we have Pranav, who’s running the flexible hedge fund and can go into tokens and equities. He’s had an incredible 4-year stretch. Being a hedge fund, he has a lot of flexibility and discretion. Then we have Node, which is more diversified beta to the entire crypto value chain.
On tokenized equities, who’s making money off this besides venture? There was a brief spurt with HOOD, but on the public-equity side, the way I see it, there are a ton of competing products. It takes balance sheet to list these products in tokenized form, so you get duplicative, fragmented liquidity. That’s good for the consumer to some extent, especially consumers who may not have access to these securities overseas.
But I’m struggling to find the winner-take-most dynamics in the space. I see it as more positive for consumer welfare. In the public equities I’m running, there are pretty small positions in Figure and Circle, and not that much in Coinbase or Robinhood. I think things like Interactive Brokers and Schwab are actually pretty interesting at these prices. But the pure plays are losing money, and I don’t see the winner-take-most dynamics in the space. That’s why we have the venture side, and I think we’ll hopefully have some winners on that side.
I’m not even sure you said who’s making money besides venture. I’m not sure venture is making any money in tokenized equities. I’m actually—I think the only people—
Well, at least it’s not marked to market every day like my book is right now, right?
That’s fair. On the tokenized-equity side, I don’t think it’s like that. We’ve remained pretty dubious about the current iteration of what it looks like. I’m bullish long term on tokenization, but with the current iteration, it’s unclear to me that it’s accruing much value to anyone other than maybe issuers, who are getting more distribution, and some of the distribution companies—the Robinhoods of the world—who are getting some trading volume out of it.
Matthew, we do Content of the Week every week. I’ll let Rob go first while you think about yours. Rob, what do you have?
I’m still working through my backlog of Halloween horror, as everybody knows. There’s a show, I believe on Peacock, called Crystal Lake. It’s a prequel to the Friday the 13th movies, and it’s actually quite good. I’m more of a movie guy than a TV show guy, but if it’s a little rainy in New York this weekend and you want to keep the Halloween horror going, it’s a good one to binge.
All right, I’ll take it. Matthew, I’ll let you round it out.
I’ve got two for you. I’m a biography person, so I’ll tell you what’s new on the docket. There’s a biography of the founder of UPS, the UPS man, from 100 years ago. It’s an excellent story: he left school at 11 years old and founded UPS.
Then there’s a biography by the CEO of Pioneer Natural Resources, a foundational oil exploration company. He lived in Tehran as a child. I think both of these companies and both of these stories are important because these are founders who built the physical backbone of the economy. As the AI buildout runs into constraints around power and capital, it’s helpful to go back and read how some of these foundational entrepreneurs solved those issues.
I also have an oil biography for you. On Santiago’s recommendation, I’m reading a book about Siggy Wilzig called Unstoppable. This guy was—
Amazing. I’ve read it. Great book.
Yeah. A Holocaust survivor, he arrived in America penniless and rose to become a prominent Wall Street CEO in both the oil and banking industries. I’m about a third of the way through. I finished Ari Emanuel’s book, The Call.
So good.
Yeah, I think I read it because maybe you had mentioned it, Rob. I ripped through that.
I called it out a few weeks ago, like three weeks ago.
Yeah. I did a little audiobook while traveling like crazy right now. The guy’s an absolute nut, but—
Did you do the one with the guy from Better Call Saul, Odenkirk, who narrates the book?
Did he narrate that?
Yeah, Bob Odenkirk did, I think. The main one everyone was listening to was Bob Odenkirk narrating the book, which was a great pick.
Oh, I like this narrator. He’s very unique.
Yeah, fantastic.
Matthew, congratulations on everything. I enjoy following you on Twitter and reading all your stuff out of VanEck. Obviously, I have a lot of love for Pranav and Yan and the entire VanEck team. Congratulations on everything you guys are doing. Thank you for coming on. Rob, good to see you. Good to have you back, Rob. We missed you.
Good to see you in London next week. Live pod.