AI Agents, Zcash Mania & Crypto’s Ownership Crisis
- Markets were described as a melt-up: they have shrugged off the Middle East growth scare, the Qs are up 25% in a month, and one panelist estimated that earnings growth was something like the best since late 2020. Bond yields rallying on Iran-related inflation risk remain the residual worry, but the practical read was blunt: “If you're not long risk, you're getting left behind at this point,” a classic melt-up characteristic.
- Saylor publicly entertaining Bitcoin sales—previously “blasphemy”—was read as a confidence device for the STRC re-peg, not distress. STRC was cited at 99.98, with only until the 15th ex-date to try to issue at par; if it re-pegs this week, the panelist covering the trade estimated “low billions” of firepower. The BTC-sale signaling was viewed as “not very punitive for BTC, but pretty accretive to both MSTR and STRC.”
- The Zcash discussion framed it as “true PVE”: nine years of proof-of-work distribution, no obvious private-buyer dump, limited DeFi-exploit exposure, and now Multicoin building a position with Ansem drawing attention to it. The 10%-of-Bitcoin target attributed to Arthur Hayes was dismissed as having “no way” of happening soon, though 2% was considered possible from a level near 1%. Zcash was presented as a candidate for the cycle’s big alt winner in the ETH/DeFi, Solana and Hyperliquid lineage.
- One panelist pitched Venice (VVV) as “probably the best-growing fundamental business right now”—adding around $2M of ARR a week, doing roughly $50M, and potentially adding another $100M of ARR—at a ~$500M market cap and high-$800M FDV. Its product and hybrid crypto/AI positioning were challenged. The power-user objection was that privacy proxies do not address leaked context: “It’s not your prompts that leak the information. It’s your context.” Porn and edgy use on open-source models remained a disputed bull case.
- The panel framed the discussed $60B Cursor acquisition by SpaceX as xAI buying its way toward a competitive model and coding position. A model company has two jobs: build a good model and buy enough compute to serve it; xAI shows that compute alone is insufficient, with Colossus at around 100,000 GPUs while its models remain below the frontier. The deal could combine Cursor’s Composer model, talent and data flywheel, and a 5–10% boost to a hypothetical ~$2T SpaceX IPO valuation could pay for it multiple times over.
- The AI thread’s through-line was that software must be built for agents, not just humans. Yan said Hermes, in which the panel disclosed an investment in Nous, was nearing 25,000 GitHub stars and taking share from OpenClaw; he and Piers use agents and a private GitHub repository rather than a conventional CRM. Parallel was described as building a search engine for agents, and Yan’s categorical view was: “If you’re building for humans, you’re cooked.” Low switching costs mean users and agents can move to the best, most cost-efficient model, keeping the labs in a major compute and capex race.
- Ceteris presented Houdini Swap as an exhibit A for crypto’s ownership crisis: a token with roughly $8–9M market cap and something like $2–3M of annual buyback run-rate was effectively zeroed after the team halted buybacks and Sol Strategies acquired the equity without indicating what would happen to the token. His systemic claim was that crypto lacks liquid funds because tokens are broadly unreliable; he argued for enforceable structures such as MetaLeX/ACE, MetaDAO and SOAR-style token-holder rights.
- Pump’s buyback cut to 50% and token burn were judged largely priced in, while retaining cash was considered rational given roughly $700M probably earned before the ICO and a $1.5B raise. The concern is that token holders lack equity’s legally enforceable claim and could be “Houdini’d.” One panelist said an ACE-style equity link could send the token over 2×, while another holder said Pump’s roadmap now looks less clear than Hyperliquid’s: “Now I’m just a community holder.”
1. Melt-up framing: earnings blowouts erased the geopolitical scare
- The opening market read was that equities had “looked past the growth shock” from the Iran conflict, with the Qs up 25% in a month and AI names printing new all-time highs daily, while crypto appetite returned through memes, AI names and “the good businesses that have been doing well.” The open question was the inflation persistence of the conflict; yields had been rallying for weeks.
- One panelist estimated that the quarter was something like the best earnings-growth period since late 2020, off a weak-growth base, while geopolitics had been “a nothing burger, at least so far.”
- The uncomfortable positioning conclusion was blunt: “If you're not long risk, you're getting left behind at this point,” a classic melt-up characteristic.
2. Saylor’s BTC-sale trial balloon is STRC re-peg engineering
- The Saylor discussion framed his willingness to entertain BTC sales—previously considered “blasphemy”—as deliberate signaling across three buckets: BTC, STRC and Strategy shares. The speaker hoped it came “from a position of strength, not fear.” Opening a new cash source for the dividend could build confidence that STRC re-pegs to 100; the amount of BTC required was described as immaterial, so the market was viewed as likely to shrug it off positively.
- The STRC microstructure has two trades: dividend-date buyers, and a crowd recreating a roughly 70-basis-point pseudo-dividend by buying in the low 99s and selling in the high 99s. That activity adds roughly 1% to the capital needed to reach par, but the panelist argued it increases confidence in a re-peg by more than 1%, bringing in outside money.
- STRC was cited at 99.98 on the call. Saylor has only until the 15th ex-date to try to issue at 100; each day that passes reduces what he can extract from the window. If it re-pegs this week, the estimate was “low billions,” plus indirect capacity to issue more MSTR.
- The BTC accumulated would sit on the balance sheet and, at least in the market’s view, be claimable by the equity component rather than STRC. The sale signal was also viewed as favorable for MSTR because it could indicate a more strategic approach to BTC per share.
3. Zcash: the well-distributed anti-rug trade
- The core case presented was that “Zcash is true PVE”: nine years of proof-of-work distribution, years of poor trading, then last year’s breakout followed by a profit-taking phase and a new narrative decision point. The appeal is that holders are not obviously facing private-buyer or team-token dumping, and the asset lacks the same DeFi exploit surface. The caveat was memorable: “Hopefully North Korea doesn't hack the shielded pool.”
- Arthur Hayes’ 10%-of-Bitcoin target was dismissed as having “no way” of happening soon, though perhaps possible in 5–10 years. Zcash was described as nearing 1% of Bitcoin, with 2% considered possible. Privacy has returned to favor, with Monero and other alternative privacy coins also performing well.
- The thesis also cited institutional interest, legitimately good technology, a strong team that has been hiring well, and attention from Mert and Ansem. Multicoin’s move was notable because it had historically been viewed as a Solana-focused fund, yet its largest recent public positions were described as Hyperliquid and Zcash.
- The pattern argument was that crypto repeatedly produces a dominant alt winner: ETH and DeFi summer, then Solana, then Hyperliquid. “Is it going to be Zcash this time? I don't know, maybe. I'm kind of betting on it.”
4. Venice: undeniable chart, contested product
- One panelist pitched VVV as “probably the best-growing fundamental business right now,” adding around $2M of ARR per week, doing roughly $50M, and potentially adding another $100M of ARR. The business was described as having a ~$500M market cap and high-$800M FDV, growing for AI rather than crypto reasons and remaining relatively underowned.
- Revenue may be understated because usage-based API spending from users who have maxed out their plans is not fully tracked. The revenue economics depend on whether models are open source or lab-based: open-source models can run on owned or outsourced decentralized compute, potentially producing better margins.
- The product objection was that most users still trust Venice with their queries, while power users need to share context—fund, work or personal information—for the AI to be useful. As one panelist put it: “It’s not your prompts that leak the information. It’s your context.” That makes the anonymization layer difficult to justify for many real use cases.
- The debate turned to whether open-source video models support porn, with participants explicitly unsure and proposing to test it. Adult or edgy usage could still support a narrower bull case, including bifurcating ordinary high-volume work from more sensitive queries.
- The hybrid crypto/AI positioning was also questioned: one panelist asked why investors would own something that is not a pure play on either crypto or AI, especially after crypto-adjacent assets had been punished over the prior couple of months. The initial airdrop was described as having been widely sold, supporting better distribution.
5. The $60B Cursor deal: xAI buying the coding killer app
- The frame, attributed to a point heard from Rory on 20VC, was that “a model company has two jobs”: build a good model and buy enough compute to serve it. Before 5.5, OpenAI was described as having executed well on compute but lagging on the model, while Anthropic had the stronger model but insufficient compute, producing throttling and complaints that its model was getting dumber.
- xAI was presented as proof that compute alone is insufficient: its Colossus cluster has around 100,000 GPUs, yet its models remain below the frontier and Grok “kind of sucks.” The acquisition thesis is to combine teams, acquire Cursor’s Composer model and training data, and try to win coding, described as AI’s first killer application.
- At $60B, the deal was called the largest private venture M&A deal in history and more than 2× WhatsApp and Wiz. The comparison with OpenAI’s ~$50B valuation and an uncertain, rumored ~$900B Anthropic valuation was used to argue that the price was not obviously irrational. Cursor was discussed as perhaps a roughly 10× revenue business, though the revenue and valuation figures were not fully settled on the call.
- The panel also discussed Cursor’s valuation history uncertainly: it had been around $2B previously, while a recent or prospective $30B valuation was mentioned and then questioned. Regardless, a three- or four-year-old business receiving an exit of this size was called an absurd outcome.
- Paying partly in SpaceX stock could add rapidly growing recurring revenue to a revenue-poor story and potentially lift the equity value. If SpaceX goes public near a hypothetical $2T valuation and the acquisition raises its stock price by 5–10%, “it pays for itself multiple times over”—a “one plus one equals three” scenario.
6. Build for agents or die: the harness is the new UX layer
- Yan described Hermes, in which the panel disclosed an investment in Nous, as nearing 25,000 GitHub stars and taking market share from OpenClaw. Agent harnesses may look like “Markdown files, a bit of code and cron jobs,” but they decide which models and services are used and can become the UX layer for software.
- Yan said his agents now interact with services such as Confluence and Figma rather than him doing so directly. He and Piers are using a private GitHub repository instead of a conventional CRM: each agent uploads the information its owner permits, and the agents communicate updates to one another. “I don't really speak to Piers even though we're building this together. It's literally just our agents speaking to each other.”
- Yan extended the thesis to licensed businesses. He described an asset-management firm building an ETF intended to select the 20 best companies using agents, with the founder planning to hire engineers rather than more analysts. Similar agent-first approaches could appear in biotech and other fields where a regulated or licensed business remains defensible while agents perform much of the analysis.
- The categorical version was: “If you don't have a reason that the AGI needs you as a software business, then you just have no reason to exist. If you're building for humans, you're cooked.”
- The panel connected this to the labs’ capex race. Switching costs are low, so users—and eventually agents—can move to whichever model is best and most cost-efficient. On the services side, Parallel was cited as a particularly effective search engine for agents, alongside Perplexity and other emerging providers.
7. The Houdini Swap rug: crypto’s ownership crisis, litigated live
- Ceteris’ account was that Houdini had been using roughly 50% of revenue for buybacks, with something like $2–3M a year of buyback run-rate against an $8–9M market cap. It had operated for more than two years, was gaining integrations, and Jupiter had discussed integrating it.
- Two weeks before the discussion, the team announced that buybacks were stopping without explaining what would happen to the token. Then, yesterday or two days earlier, Sol Strategies announced that it had acquired Houdini Swap’s equity, again without indicating what would happen to the token. Ceteris said the token was trading at zero with no meaningful liquidity. Its original ticker was POOF—“foreshadowing.”
- Ceteris’ broader claim, beyond his own loss, was that crypto lacks liquid funds because “tokens are just dogshit all around.” In normal capital markets, identifying a protocol with revenue, buybacks and integrations could be rewarded; here, the low multiple existed partly because investors feared the token would be rugged.
- He argued for enforceable token-holder rights through structures such as MetaLeX, MetaDAO and SOAR, saying a pure buyback is no longer sufficient. He also said he owns HNT but cannot buy more because Nova Labs could potentially sell the equity and stop supporting HNT.
- The panel debated whether ecosystem participants would actually ostracize Houdini and Sol Strategies. One view was that the market should reprice non-ACE structures toward zero and force adoption of enforceable models; another was that the incentive to keep using a useful business makes ecosystem-level punishment unlikely. The discussion also noted that token holders and equity holders can fare differently: FTX equity holders had to absorb the failure while many FTT holders exited, and Axie token holders did better than equity holders.
- Another panelist argued that early DeFi’s on-chain cash-flow rules and governance requirements had provided stronger protection; after the industry dismissed governance tokens as broken, value became discretionary: “Your token is worth something until the team tells you it is.” Delphi’s incubation of MetaLeX was disclosed.
8. Pump: the cut was rational, but the discount is about getting Houdini’d
- Pump’s announcement cut the buyback from 100% to 50% of revenue, burned repurchased tokens and coded the arrangement for one year. The burn was viewed as already priced in, while the one-year term provides clarity for twelve months but is insufficient for a multiyear DCF. Significant unlocks were also described as an overhang.
- The defense of retaining cash was that Pump probably made around $700M before the ICO and raised $1.5B afterward. Keeping cash to build the business and maintain years of runway was viewed as preferable to distributing everything immediately and later facing a shortage from a position of weakness.
- The counterargument was that equity has “a contractual, legally enforceable claim on the underlying business,” while the token’s discount reflects the fear of being “Houdini’d.” Greater transparency about spending and a legal tie between the business and token were suggested.
- One panelist said Pump could rise more than 2× if it implemented an ACE structure. The broader comparison was with Hyperliquid, which returned 100% of revenue to token holders without having raised money. Pump has a greater need to build products, but its next product or acquisition is unclear; streaming “didn't really work.”
- A holder said the ICO thesis had been another Hyperliquid, but Hyperliquid’s product roadmap and new-market opportunities now look clearer. Pump remains tied largely to meme-coin trading, even though its revenue is higher than it was a year earlier in worse market conditions. “It seemed a lot more clear when I bought my Pump tokens. Now I'm just a community holder.”
Full transcript
All right, everybody, welcome back to the Delphi Podcast, the show about markets, AI, crypto, a lot of unfiltered opinions. We got tons to talk about today. Joined as always by Yan, co-founder, managing partner at Delphi Ventures, Jose, co-founder and head of Delphi Labs, Siddarth, our head of research, and Jason, our head of markets. Feels like it's a good It's another good day to be doing this. You know, stocks are up early this morning, a lot of geopolitical news. Had some nice crypto moves over the last 24-48 hours. How's How's everyone feeling?
It's hard to feel bad, honestly, at this point, right? Markets—
Zcash.
Yeah. Yeah, I guess Zcash. Everything is looking pretty decent. Markets have, at this point, shrugged off at least the growth shock—or scare—from what's going on in the Middle East. They've clearly looked past that.
The Qs are up 25% in the last month. A bunch of tech and AI names are hitting new all-time highs every day, and appetite is returning to crypto markets through memes, AI names, and the good businesses that have been doing well. The real question now is whether markets have looked past the growth shock but are still held up on what the inflation implications of the Iran conflict are going to be, and how persistent they will be.
Clearly, bond markets—yields have been rallying over the last couple of weeks as well—are under pressure. There's a lot of talk about what earnings would look like for all these AI names. Would they be bad? Then everything just got blown out of the water: new all-time highs across the board.
1. Strategy, STRC dynamics & Michael Saylor’s BTC comments
If you're not long risk, you're getting left behind at this point, which is a classic characteristic of a melt-up. I'm curious what the other guys think.
Yeah, you had one of the best earnings quarters in terms of growth in a very long time, right? I think it was something like the best since late 2020, right? And that's just on the back of weak growth. It's been insane blowouts across the board.
Risk-on, I think. Geopolitics again ends up being a nothing burger, at least so far, so everyone's looking past that. Crypto was flung by the wayside for a bit but has been picking up recently. You had Saylor's earnings call yesterday, which is obviously very big for crypto, and I think he had the recent mention that shook everyone: the idea of him entertaining selling BTC.
Prior to yesterday, that was probably thought of as blasphemy in terms of his appetite to do so. I think it makes a lot of sense to signal it, and I hope it's from a position of strength, not fear. The whole idea is that there are 3 buckets that need to be satisfied: BTC, STRC, and Strategy shares.
It is a bit zero-sum. Everyone has been watching this thing like a hawk because the second that STRC re-pegs to 100, he can start selling those at the market. Historically, this has taken longer this time than ever before, and I think, to some extent, that makes sense. There's much more of it now, so it takes a lot more money to bring it back to par.
As you keep issuing more, financially, it becomes harder to sustain. The idea is that it basically keeps re-pegging until it doesn't, and then he would raise the dividend again. I think there are a couple of games played on the STRC side. You have those who buy it for the ex-date and collect the dividend, and those who recreate a dividend-like return by purchasing it in the low 99s and selling it in the high 99s.
In between ex-dates, you get a pseudo-dividend, which is admittedly smaller. If you're assuming 1% a month, roughly, this is closer to 70 bps. I think there's a slight negative element, but overall, there's a positive element to the involvement of those people. They create a bid in the low 99s, but then create a bunch of supply in the high 99s.
That makes it easier to get to the high 99s but harder to get to 100 because they're increasing the overall amount of buying needed to get to 100 by roughly 1%. The positive side is that, yes, they probably increase the amount of capital needed to buy 1%, but at the same time, I think they increase confidence in it re-pegging to 100 by more than 1%.
You have outside money building confidence in its ability to get to 100, and that money will come in and start buying now. I think the reason he mentioned the idea of selling Bitcoin is that it also drives confidence in getting to 100. You've opened up a whole new faucet for cash that can sustain the dividend.
It's not an insane amount if you think about the amount of BTC he would have to sell to sustain the dividend. It's not a meaningful portion of Bitcoin. I think the market is generally shrugging it off and ideally viewing it more as a positive because it's enough to instill confidence in the STRC bucket that it helps re-peg and keep the flywheel going, so to speak.
Him mentioning that is meant to instill more confidence in the STRC side and ideally help those buyers gain confidence in his ability to continue paying the dividend. The downside is that he basically only has until the 15th to really issue STRC at 100 if it gets there, because at that point it goes ex and the price drops again.
He is limited in his window now, and every day that it takes, the harder it is—or the less he can get out of it. At a certain point, I think the market will probably reverse a bit because enough of this trade is set up that people are basically just buying when they think he can issue, or front-running when they think he can issue, and then getting out as they approach the ex-date.
You get to this crunch-time element. It has been gradually grinding higher. I think it's 99.98 right now, right? Yeah, there you go. Ultimately, that's what people are watching, and then you have broader alt speculation happening alongside it, as well as a handful of names.
How much firepower do you think this Saylor Strategy play has in the near term? Do you have any ballpark numbers you're looking at or expecting?
Low billions, I'd say. Obviously, the longer you have and the more days you get, the higher the chance of more volume. Recently, the volume has been picking up day by day, roughly. Even today, we're almost at 1 million shares of volume, and it's 10:51 Eastern.
You're starting to see more and more volume, and you expect that as people are flipping out of it. If it gets there this week, I think you can get low billions out of it because you also get the indirect ability to issue more MSTR alongside it.
In theory, the BTC that's being accumulated goes onto the balance sheet. The balance sheet is only claimable by the equity component, not this STRC side—or at least, I think that's how the market views it. It tends to be positive for both.
2. Zcash breakout, privacy narratives & alt rotation
I also think the idea that he would sell a slight amount of Bitcoin is favorable for MSTR because it means you think he'll be more strategic about this BTC-per-share approach. That's accretive for those holders. The signaling was basically not very punitive for BTC but pretty accretive to both MSTR and STRC.
Yeah, and with Zcash, as you mentioned, we've seen some speculation further out on the good old risk curve, especially over the last week or so. The big one—I have to give a shout-out to it—the last time we recorded this, we all pulled up this chart, and we drew in a little higher. We had a nice, clean break from this, right?
Nobody expected it to go this high this fast, but you had the Multicoin tweet, then building a position, and then Ansem retweets it, and he's moving markets again, drawing attention to it. I think everybody on here—most people—has a private Bitcoin bag. Where do you guys think this goes, though?
Do you think we get back to— is the expectation that we're getting back to 700, 750? Is it going higher than that?
First of all, if you're trimming, where are you trimming?
So, first of all, on Strategy, we have a crazy detailed report coming out next week on all this stuff. Definitely read that. If all this stuff is confusing for you—and it is really complicated structurally—read this report.
It might take you a while to read because it's super detailed and there are lots of numbers, but you'll understand it by the end of it. On Zcash, listen, I don't know. I'm still bullish on Zcash.
I think Zcash is true PVE. I think this is what's going for Zcash right now. First of all, it had 9 years of distribution, so it has this really long proof-of-work distribution that went through. It traded really badly for years.
Then it had this breakout last year—a nice pop—and then the profit-taking you would expect going into the next year. This was kind of a decision point: was this going to be a real narrative, or was that the only pump? When the Valkyrie fell and all this stuff happened, I think Multicoin coming out is a pretty high signal.
You can meme it pretty well. People are going to have some ridiculous targets. I saw Arthur Hayes tweeting 10% of Bitcoin. In my opinion, there's no way it gets that high. Maybe in 5–10 years, potentially. But it's getting close to 1%, but could it get to 2% or so? I think it could.
Privacy has really come back into vogue over the past year or so. Even Monero has done pretty well, too. If you look at coins that are up over the past year, Zcash is up a ton, but Monero is up, too, and some other alternative privacy coins are also up.
DeFi has a million exploits a day at this pace. Tokens are getting rugged by their protocols, and VC unlock tokens have just been down only. With Zcash, you have this coin that's really well distributed. You're not getting dumped on by private buyers, you don't have to worry about the team deciding, “We don't care about the token anymore,” and you don't have to worry about DeFi exploits.
Hopefully North Korea doesn't hack the shielded pool. I don't know if they're able to do that, but you kind of just have a lot going for you. Also, in crypto, let's face it: a lot of times assets just trend, and it seems like there's always a big winner—a big alt winner.
Obviously, you had ETH and DeFi summer years ago. You had Solana after that. If you had Solana—or if you didn't have Solana—that was a big determinant of how good you felt in 2023 and 2024. Then Hyperliquid took the mantle after that. Is it going to be Zcash this time? I don't know, maybe. I'm kind of betting on it.
It just seems like there's a lot going for it, and it has a lot of institutional interest, too. The technology is actually legitimately good, they have a really strong team, and I think they've been hiring really well recently. Mert has always been behind it, and now they have Ansem behind it, too. Ansem seems to be moving markets again.
One interesting thing with Multicoin is that I feel like it's pretty clear Kyle is a massive Solana maxi. He hated Hyperliquid, he hated Zcash, and the 2 biggest positions Multicoin has built publicly since he's gone have been Hyperliquid and Zcash. So I think it's interesting. It's definitely a change from how you've historically thought of Multicoin—as mostly a Solana fund—and now they're really branching out into other stuff.
3. Venice AI (VVV) and the AI x crypto trade
This latest pump was kind of crazy overnight. Any other alts you guys are looking at?
Yeah, pull up that VVV chart. Show me something prettier than that. What's going on?
Probably the best-growing fundamental business right now. They're adding around $2 million in ARR a week. They're probably doing about $50 million right now and are about to add another $100 million in ARR—not, you know, subscription revenue. The business is good because markets are noisy. No, this is actually sustainable, and the idea is they're flowing value through a token.
I'm sure we can go into the tribulations of that. But, yeah, it's a $500 million market cap and high-$800 millions FDV. So, not a low cap, but also something that I think is pretty underowned. It's growing for AI reasons, not crypto reasons, and you don't really see much about it on Twitter. I think it's just going to keep going here. They're gradually returning value, but mostly just reinvesting in growth right now.
No VC unlocks ready to dump.
Yeah, I mean, there was a big airdrop when it launched, right? I sold mine at that point, and I think a lot of people sold, too. So, definitely better distribution on it.
I wonder why it's not talked about as much.
If you're looking for an AI-overlap narrative in crypto, there are few cleaner charts, especially year-to-date, than this one. I guess it's kind of why everybody argues in the group chats. It has crypto and AI overlap, which is the reason why I never bought Galaxy. Why do I want to own something that's not really a pure play on either? Why not just own something that's strictly AI or equities?
Obviously, VVV has done really well, but I think that's probably the reason. Over the last couple of months, you've been punished really badly for sticking with crypto things that are adjacent to what's going on in the real world. You've underperformed massively. You might as well have just gone to equities and traded it directly there.
I just think Venice fundamentally doesn't make any sense. I don't get the product. Why does this thing make sense?
For most users, you're still trusting them, right? You're still sending the queries to them. They anonymize them with a proxy or whatever. But for anyone who's power-using AI, it's not your prompts that leak the information. It's your context, which is what you want the AI to have in order to answer questions about you.
Every query I'm sending to AI has built into it either our fund context, my context, or some personal context—whatever it is I'm querying. So I just don't get how having this anonymization layer on top of AI really makes sense. Even if it worked fundamentally, if people were using the TEE or end-to-end encryption stuff, which I don't think they are—
They are.
I still don't think it really makes sense.
What do you mean they're not using the TEE? I think most people are just using the Venice-controlled one, right? The Venice-controlled GPUs—the anonymous mode, where it's basically—
I mean, you're wrong.
You're wrong? No. You're feeding into every other underlying model. I think a decent amount of it—I don't know, actually—because they don't break out token usage. They gradually break out token usage.
But I think the obvious application is porn, especially with all the video models, and that's a massive market.
But the video models don't do porn, right? Which video models do porn?
The open-source ones do.
Okay. I wasn't aware that they did. I thought you had to fine-tune them to do porn. Is that where most of the volume is coming from right now?
No idea.
Are you sure the open-source models do porn? Then we need to test it out right now.
I think that's a legitimate bull case if they do. But the last time I checked, they did.
They have the same safeguards as open-source models.
Okay, interesting. I'm confirming that now. What do you get with this? You said subscription revenue is one of the big drivers. What do you get with that?
It's just usage-based.
Just usage-based?
Yeah. I think the revenue is also understated because they don't really track the people who are paying for API usage after they've maxed out their max plan. Revenue in this case is total spend—API spend—which I guess is mostly being passed through to the labs, no?
It depends on whether it's open-source or lab-based.
For the open-source ones, they're running them on their own hardware.
I think it's probably a combination of some of their own hardware and outsourced GPUs from decentralized compute providers.
And there you get pretty good margins.
Interesting. Yeah, I think the porn and illegal use cases could be cool. I just think that for the majority of my use cases—where I would want privacy, if it were available—it doesn't work because of the context. It feels like a very niche market, I guess.
I mean, I guess porn isn't that niche.
You could bifurcate your use. You'd keep your usual, high-volume things in your existing setup, and then use this for more edgy stuff where you're less worried about keeping your entire context window and all that information up to date.
4. Cursor’s rumored $60B acquisition & AI business models
Yeah. It's just hard to necessarily see where all this goes because, as we were talking about before we jumped on, AI in general is extremely hard to predict. The pass-through model is a similar type of qualm that some people have had with other bigger companies, like Cursor, which recently had the $60 billion potential buyout announcement.
But that's always been one of the questions around their model: a lot of the revenue that comes in goes right out the back door because they essentially serve as an aggregator for a bunch of different top frontier models. They've started to develop and build their own models, too, to diversify that revenue stream or have some lock-in, but that pass-through model has definitely had some question marks around it.
The pass-through model?
Yeah, where it's like Cursor. You're paying Cursor, but, to your earlier point with Venice, a lot of the revenue that comes in winds up going out the back door because it goes to paying—
It's like double-counted. Yeah, revenue for the AI. But you assume they have enterprise plans, so they do get better economics than retail. But I agree; it's the lower-margin of the two.
Yeah. I think for the open-source models, the margins are much bigger, for sure. When we spoke to Kimi and stuff, it seemed like that's the model they have: more efficient ways to serve them, where you can have higher gross margins on them. But I don't think it's a super— I mean, Cursor's acquisition kind of validates the model, though, right?
I think they're the— Is it the biggest private acquisition ever? In startup land, it feels like definitely the biggest one I can remember. Wiz seemed big, and WhatsApp seemed huge, right? But this is more than 2× WhatsApp and Wiz.
Yeah, I mean, you had quite a few, actually, in the early 2000s. But at the time, they weren't necessarily private ones. I mean, it's definitely up there in terms of AI. It's one of the largest, for sure.
But it's interesting, SpaceX being involved in that.
Yeah, it's the largest private venture M&A deal in history, it seems like. Yeah, which is wild.
But how much of that is— It's an interesting dynamic because you've got all this interplay between SpaceX and Elon. If you go and acquire Cursor and you've got xAI, part of Cursor's business model is selling access to frontier models—one from OpenAI, one from Anthropic. You're almost acquiring a business that potentially benefits from your competitors doing well, too, in a weird way. I'm just trying to think: what is the reason for them going after Cursor?
Yeah, I think it's just that they need to—the Grok model is clearly falling behind. It kind of sucks, right? It's well below the frontier. But, I guess, in general, one way I think Rory from the 20VC podcast first said this to me—or at least I heard him say it—is that a model company has 2 jobs, right? You have to build a good model, and then you have to buy enough compute to serve it. You have to do both well to win.
Until earlier this year, before 5.5, OpenAI had done a really good job buying compute, but they'd fallen behind on the good model, right? Anthropic was the inverse of that. They built a really good model, but they fell behind on compute, which is why they were having to throttle users and people were saying the model was getting dumber and stuff like that.
Obviously, having just compute isn't enough. I think xAI is the best proof of that, right? They have this Colossus cluster with around 100,000 GPUs, and yet their models still suck. Cursor has its own model—this Composer model—and obviously, they have a lot of data to train it on. So, for me, the bet is that they're going to try to win coding: combine the teams, get this talent, get the data flywheel going, and try to make a model that's competitive in coding, because clearly that's the killer app for AI, at least the first killer app for AI.
Yeah. And then, I mean, even the valuation side—$60 billion doesn't sound insane, right? When you look at OpenAI at $50 billion and Anthropic rumored to be raising at—what is it, $900 billion?—because they just crossed OpenAI's ARR rates at much healthier margins, too, might I add. It doesn't seem like that crazy of a bet.
Yeah, it seems like a good deal for both sides, honestly. I feel like SpaceX kind of needs this. They need a leading model. Cursor is growing pretty nicely, but for them, it's a pretty absurd exit. It's a 3- or 4-year-old business, you know, and it's the biggest private venture acquisition of all time.
Wait, wasn't it over $2 billion last year? What wasn't its valuation—$2 billion?
Two, and then it just—I think it just raised recently at $30 billion.
$10 billion, right?
Or $30 billion, was it? Okay. I thought it was $30 billion. Or they were getting ready to raise at $30 billion. For some reason, I—
Either way, that chart is exponential. Obviously, the Cursor side is clear, right? For the SpaceX side, you're probably paying with a combination of cash and stock.
Yeah.
SpaceX has a big revenue problem, and you're slapping on a bunch of recurring revenue that's ramping very quickly. So you assume the market probably gives that some premium as well, and you get a bump in your equity price that probably more than offsets the purchase. And you have the compute capacity through xAI, which I guess you could say is the most expensive acquisition, but, yeah, it's one of those 1-plus-1-equals-3 scenarios.
Yeah, because it's kind of crazy: if SpaceX really does IPO anywhere near $2 trillion, to your point, Yan, if this acquisition has even a high likelihood of increasing the chance of your stock price rising 5% to 10%, it pays for itself multiple times over.
Yeah, it's also—I don't remember what SpaceX is trading at, but assuming it goes public near $2 trillion, it's like 100× revenue, right? I think Cursor is a much smaller multiple. It's like—
I don't remember what the last revenue was, but I think it's like—
10×, or—yeah, I think in February it was $2 billion. Okay.
Revenue on the $30 billion valuation, so now it's probably inflated a lot.
Yeah, exactly.
5. AI agents replacing traditional software workflows
It's probably a 10× multiple, so it just makes a lot of sense for SpaceX to be doing these acquisitions, I think, especially paying in stock, like you said. I think the really interesting thing—actually, I haven't even thought about this too much—is also Hermes. You've seen this, and we're investors, by the way, in Nous, but it's really been inflecting. I think it was nearly 25,000 stars on GitHub. Everything on my feed is about Hermes. They've been stealing market share from OpenClaw.
These agent harnesses—I kind of go through phases where, to some extent, it's just Markdown files, a bit of code, and cron jobs. But in another way, it's actually the most valuable thing because it's the thing that decides what model you use and what services you use. At this point, what I'm noticing in my workflows is that I don't interact with any internet services. I don't interact with Confluence or these project management things. I'm not interacting with Figma. It's just my agents; they're interacting with everything, actually.
The software products that are most successful are the ones that are built for agents. You kind of saw this with Parallel AI raising $2 billion, right? It's a service that helps you—it's like a search engine for agents. I think you're going to see more and more of that. For the many people using these agent harnesses, the agent harness is what defines what model it wants to use and what services it wants to use. That is pretty valuable. It's almost like the UX layer for a lot of this stuff. But I don't really have a strong take; I'm curious if anyone has any thoughts or if you've been using this stuff.
Yeah, no, that was the point Rory made on the 20VC podcast, right? Basically, you're no longer really optimizing for what your employees want to use, but rather what the agents choose. The metrics they choose are very different—or the metrics they choose based on are very different—from what employees use, and so it makes it hard to predict who ends up winning. But just based on the volume of usage from agents versus employees, they're kind of always on. So if you can nail that, you nail where software wins, because right now everyone thinks all the software is going to lose.
I think collaboration is also a thing that's not really solved yet. I know Piers and I at Tellfire are building a— We both have personal CRMs, and we're trying to find a way to integrate our data to make a public CRM, right? So all our transcripts and messages and stuff like that can go into an overall database. Permissioning is still super hard.
At the same time, I don't really want to use an existing CRM because I have to learn to use it. Claude can just generate one for me with exactly the filters and edges I want. I don't have to reformat everything. The same thing goes for Piers, I guess—he has his own workflow.
What we've landed on is just having a private GitHub repo that you upload stuff to. Our agents are basically—you know, I tell my agent what I want to share and what I don't want to share, and at the end of every day it just uploads that stuff to the repo.
I think in the future, people aren't going to use this boring stuff where, after every call, you have to upload the CRM and say what happened. Agents can just do all that stuff. If they can do it, why would they use a CRM? Why wouldn't they just interact in Markdown files or whatever?
So much stuff changes with this paradigm. Once you start, it's really interesting. When we're collaborating on this, I don't really speak to Piers, even though we're building this together. It's literally just our agents speaking to each other.
I'm like, “I made this update; send it to Piers.” Then Piers responds to me, and I'm like, “Read the update from Piers and tell me what we should do.” You know what I mean? It's so different from what it was before.
I think a lot of software that isn't built for this environment is just cooked. I feel like anything that's just for humans is hard to be super bullish on.
Yeah, it almost goes back to bare-bones or more raw forms of software. The back ends and the actual infrastructure that drives it are where the value accrues. On the UI side, there will be certain cases where that can make sense, but for the most part, your point is that the traditional SaaS model—and a lot of SaaS companies are different, right? So it's not like throwing the baby out with the bathwater—but that traditional SaaS model is entirely upended by this because it's all back-end stuff.
Yeah. I think some companies are adapting to it. I'm not sure—I don't know enough about SaaS. I've never really used a lot of this stuff. I've always been pretty terrible at doing this kind of thing and hated this kind of work, but now with agents, it feels like it obviates a lot of that.
Especially ones that are built around workflow and productivity workflows, automating things that historically involved marketing and emails. All these models can do that even better than some of the software that's out there. So you really have to pick and choose what the use case is, but it's a massive disruptor.
It's interesting. For software, if you're building software, it has to be for agents, in a sense. I almost think of it as serving AGI. You assume there's going to be an AGI, and if you really believe in AGI, you'd think it's going to be a better doctor, better lawyer, and better financial advisor than any human.
But it's not actually able to do any of those things, right? A lot of the interesting software businesses we're seeing are people wrapping a licensed business—basically, a business that's defensible with a license—and building it agent-first from the ground up, competing with incumbents.
I spoke to a financial asset-management firm last week that is building an ETF, just leveraging agents to find the 20 best companies in the world. It's a very experienced asset-management guy who is basically trying to compete with the BlackRocks and Vanguards of this world without having to hire any more analysts.
His goal is explicitly not to hire any more analysts. He literally just wants to hire engineers, and then it's going to be an expression of his taste, right? All the analysts will be agents. I'm seeing the same thing with biotech people doing this for peptides and things like that.
Obviously, agents can't just buy that stuff for you, so you can have something with a license that's licensed in every state or in every country that does this. But I really feel like if you don't have a reason that AGI needs you as a software business, then you have no reason to exist. If you're building for humans, you're cooked.
That's also why a lot of these big labs will stay in this massive capex race, because when switching costs are that low, to your point, 5.5 comes out and you can switch really easily. Now there's an argument that OpenAI is back in vogue, right? But give it a few months, and the next thing rolls out.
We, as the users—or the agents as the users—will just use what's best out there, what's most cost-efficient and gets the job done. On the back end, the model companies are in this massive race. They can't let up.
6. Why SaaS companies may be vulnerable in an agent-first world
Yeah, the models are tough. I do think that, on the services side, some people are building great businesses. Parallel is an example. They've built a really amazing search engine. Perplexity has one, and there are a few others, but Parallel just works really well. It's the one that gives me the best results, at least.
Yeah, you can definitely build some massive businesses serving agents. For sure.
Yeah, which is interesting. I think that part is solved.
Yeah. It's going to be really interesting to see how it plays out.
One thing we definitely need to switch back to crypto, but we need to hit on this because it's a huge story this week. I'm going to pass the mic over to Ceteris in a second here to give his unfiltered thoughts on the Houdini Swap acquisition by Sol Strategies. I know you have strong takes on it.
7. The Houdini controversy & crypto token holder rights
What I want to know, really quickly, is what happened, in case anybody didn't see it. I also think this is a very interesting scenario that we're probably going to see come up more and more around token holders not having true ownership rights.
If you were going to do this backward, what would have been the optimal outcome here? How do you solve this?
I don't know if it was a huge story. I probably helped make it a bigger story, but it's not something new. We've seen it before with Axelar and with other dual-equity-token structures.
I think the biggest problem in the industry is that we don't have liquid funds, right? We have no liquid buyers of these tokens. We have a little bit—a good amount—of venture, but we have no liquid funds. The reason why we don't have liquid funds is because tokens are just dogshit all around.
The Houdini thing was pretty egregious to me. This is a protocol that's been around for over 2 years. They were doing this buyback and burn. They had some staking thing before, but they had over 50 million in circulating supply.
Recently, they've been getting good integrations and stuff, and their market cap was down to 8 or 9 million. So you're looking at this as an—and their revenue, the buyback run rate, was 2 or 3 million a year or something. It's trading at a low multiple to its market cap, relative to the buybacks.
And you're kind of like, okay, what's going on here? I'm someone who's a fundamental investor. I want to look for protocols that are genuinely undervalued. You have this protocol here using 50% of the revenues to do buybacks. You're trading at like a 3–4 multiple. They're getting integrations with other protocols, like Jupiter was talking about integrating them and everything.
And so in a normal capital market, right? Solana wants to be internet capital markets, right? In normal capital markets, you see all this, you do all this work, you buy the token, you're probably going to be rewarded for it, right? Because that's good research you did to identify this. Good fundamental buy.
However, the reason why it was trading like that is because the token was going to be rugged, you know? And 2 weeks ago, the team just did an announcement and said, “We're stopping all buybacks,” and they didn't say anything else. They didn't say anything about the token. They just said, “We're stopping buybacks,” and the token basically went to 0 then.
And then yesterday, or 2 days ago, this announcement comes out that says, “Sol Strategies acquires Houdini Swap's equity.” No indication of the token. The token's trading at 0 now, but even that market cap—there's no liquidity anywhere. It's literally 0.
And, I mean, yeah, listen, I had some of this token. I lost money on it, but I've lost way more money on way more tokens. It's not about the money here. It's really about, if you want to fix the market structure, we have to stop with these tokens just rugging out of nowhere. Nobody can ever have confidence in them.
I really think you have to have a MetaLeX, a MetaDAO, a SOAR, any of these other protocols that are enforcing some sort of token-holder rights. If your token doesn't have any of this, I don't even think a pure buyback is enough anymore either. Maybe if you're Hyperliquid, because they have really good trust, but you can clearly just turn off the buyback and rug the token out of nowhere.
And so, think about it, right? This team had their equity, they had their token, they launched the token years ago, and I'm sure the team sold team tokens over the years, right? I can't say for sure, but I find it unlikely that they never sold any tokens over the years. And then one day they just decide, “Yeah, you know what? I don't want to worry about that token anymore, and I just want it to not exist, you know?” It's funny because the original token ticker was POOF, and that would have been a lot better for this situation, but foreshadowing.
Yeah, it's just annoying. This is the deal with tokens, though. This is the deal with tokens: they've evolved, at least.
Not all tokens, though. There are tokens that are moving in—
Right, but it's like—
—the era, right? Like, the early-vintage tokens, they—
This was all early, for sure. Equity, dual-equity tokens—it's like, I don't know, man. If there's a dual-equity token these days, I just don't think you can buy that token.
The diligence you basically have to do—which is unfortunate, because all that means is it's going to compress multiples as people—
Yeah, it's like—look at Helium, okay? I own some HNT, but I can't buy more because I don't know if Nova Labs is just going to sell the equity one day and rug HNT. They don't have to keep doing HNT buybacks, right?
And so these things just need to be solved. I also think Sol Strategies in this case deserves blame here. If you're participating in essentially rugging tokens, you have this entity that you promote as being good for the Solana ecosystem and Solana-aligned. Tokens need to do well for your business model. That's what you're doing here, right? You're supposed to be supporting an ecosystem with tokens on a blockchain and stuff, right?
I definitely agree with that. Relative relative to like what Pump does, right? They're they're not interested as much in an ecosystem. It's more about their own ecosystem financially, but for them to rug the token, which they actually ended up um redeeming. Uh and and it's less of a thing, but yeah, for Sol Strategies, I I totally agree. They thrive with activity. Activity is is kind of supported by growing token ecosystem.
Yeah, and like the the the it's just annoying because like Houdini actually has As I said, they've been getting more integrations recently and it's like people buy this token because they see them doing buybacks and everything and they see them starting to like get more adoption. And then they just decide that the tokens were zero. And it's like you just We just have to stop this stuff basically. And I don't know. Like I've been I've been rugged a lot and like everything and I usually just like chalk it up to the game. But I feel like some of these things people need to like make a bigger deal out of because I think some of it is a bit of like Peter Griffin when he hurts his knee uh meme, you know? Like the the these things have done like There's there's loads of examples where the opposite happened. Like right, FTX went under. FTX equity holders just had to sit there and take it. FTT holders mostly got out like right, Multicoin sold most of their FTT at like 20 bucks um and and got out. And like many [clears throat] other examples of things I mean Axie is another example where you know, token holders made out much better than than than equity holders. I I just think you like all this stuff's transparent. Like you can see when a token like
This isn't transparent, though. It's transparently—I mean, you knew that you were depending on them to do buybacks, and you know there's a possibility that they stop doing buybacks at any moment.
All these tokens would just be worth 0 then, basically? I think they should trade at a lower multiple. I think they trade at a multiple based on people's confidence that the team will return—
—return value to the token, right? Hence the multiple versus hype. For sure. And I think teams—the—I mean, I think what MetaLeX is doing with MetaDAO and with Pump, these structures where you're linking equity and token, makes sense.
And I think those things should, over time, trade higher as we see more of these rugs. But I do think a lot of this is like, man, you can see there's a reason why early DeFi—everything was on-chain. Cash flows to the token weren't coded on-chain; if you wanted to change them, it was a governance proposal.
All the stuff that people are like, “Oh, governance is broken. We don't need governance tokens.” And now this is what you get. It's discretionary. Your token is worth something until the team tells you it is. And that's the reality of a lot of crypto tokens.
So you have to trust that the team cares, or make them put it into either on-chain—or, in the cases where it's not possible, which is most of the stuff, do something like MetaLeX. Work with a company like MetaLeX—
Yeah, I don't disagree with you. I just think crypto is permanently—if we don't switch to those types of models and we just leave the status quo on these kinds of “your token can be rugged at any moment” scenarios that we have—
You kind of need—yeah, it's just part of what you diligence, right? You have this setup where the team has likely sold tokens, but obviously can't sell equity. And so now all the value is sitting in equity.
And so you have a combination of that, plus the business isn't doing particularly well. And so you have the ideal setup for a team that feels very little loyalty to the token community. Let me get my final exit. Yeah, it's just scummy, man.
Yeah. I don't even care if it's legal. Rugging the token may be legal, honestly. Doing this might be legal. It's definitely legal. It's definitely legal. Yeah.
But it's just like—well, I haven't read the docs. It depends what they promise. I assume they didn't promise anything to anyone, though. I mean, this is the postgame, so—
Huh? Considering how long they've been around, I think back then—
Yeah. I mean, the thing is, a team at this point—I get you. At the same time, a team can't be forever wedded to this token, to this community, right?
8. Pump’s buyback changes, burns & token economics debate
If that's the case, don't launch a token. Yeah, isn't he kind of arguing for equity or, like—? That's horseshit. Don't launch a token if you're just going to sell your allocation and then—I mean, I don't know if they did that here, but they probably did. It's been around for 2-plus years, and then just leave it to die at some point. I just think that's—
I don't think it's binary. You're not forever wedded to the token. You just have to not rug token holders in the process of the exit, right? You don't have to be forever wedded to it, but you just need to not do this kind of sinister scenario where people are just left hanging, with no information.
I think from a legal standpoint, though, I don't know how this would work. In this one, it says here, right? Part cash, part shares of actual Sol Strategies itself. What would that look like—not rugging token holders or LOCK holders in this case? What does that look like, and do you run aground legally if, now, as a token holder, you're converted into an equity holder of this company? The only way to do this is—
Question.
It's the ACE thing. Gabe's ACE thing is the only way for this one specifically.
That would have stopped this. If they'd done that, that would have stopped this, right? Token holders could convert to equity and claim the treasury back via the BORG.
I know. I think that legally, there's no real incentive for either the acquirer or the acquiree to do this. And so I think that you have to punish the business. The market has to punish Houdini Swap and Sol Strategies going forward.
Otherwise, people will keep doing it, right? This means teams within the Solana ecosystem stop working with Houdini Swap, right? If the Solana ecosystem wants to build a vibrant, healthy ecosystem of tokens and a lot of liquid funds that want to invest in it.
Man, I mean, that's one way. That's one way for sure. I also think the market can just reprice everything that's not an ACE, that's not using the ACE model, and over time force people into using something like this instead. I think the market is the most effective way to do it.
That's the only way you have everyone operating in a rational, self-interested way that scales, right? There's just no way that everyone will say, “I feel bad for the $8 million in market cap that was just devoured. I will now shift my business focus elsewhere and avoid them, when potentially they are the best thing for me to use.” It's got to be the other approach, where there are different stakeholder groups making different decisions on the market pricing versus the other stuff. But, yeah, the market is going to punish non-dual-equity-token structures.
I think the market is really just going to—yeah. The reason I don't know if what you're suggesting will work is because, like you said, Sol Strategies, if there was ever an acquirer that would try to do something not to rug the token community, it's them, right? Because of where their long-term interests would lie. And if they're going down this path, I think the propensity for everyone else to shy away from Houdini Swap or something like that is pretty low.
It's definitely low. Don't get me wrong. The market just sending these tokens to zero is the real way you fix it. I agree with you that it's much harder to get people within the ecosystem to stop working with Houdini. I don't think they should work with Houdini because of how their model works. That's a different story. But, yeah.
Yeah, and this brings us to Pump quite nicely, actually. What did you guys think of their announcement? Obviously, I think they announced they're cutting down the buyback to 50%. I thought it was good—all the tokens that they bought back.
Yeah, I think the burning of the tokens was kind of irrelevant. When you've bought back the tokens, I think the market expects you not to just dump all those tokens on the market. They're kind of already out of supply.
A lot of people are upset that they changed the buyback and burn to 50% from 100%. A lot of people said, “They raised so much money; why do they need to continue keeping some revenue?” I think that's kind of a bad reason. Lots of companies raise a lot of money and still keep a lot of retained earnings to continue growing the cash.
I think putting the buyback and burn in code for a year is a good thing. At the same time, if you're doing this on a multiyear DCF or something, you really can't use that because you have a 1-year guarantee; you don't have a guarantee in perpetuity.
So, I don't know. There are good and bad things. I think overall, there's clarity for the market, at least for a year. They have a lot of unlocks coming, so that's going to be tough with cutting emissions 50% and then all the unlocks. I don't know if people have other expectations. I mean, your DCF isn't conditional on money flowing back in through buybacks or dividends, right? DCFs are just what the business expects.
And so—
But it's conditional on you having a link between the business and the token that you're buying. Yeah, but if the idea is that they're saying this value accrues to the token, the only reason they need to return it now is because there's lower confidence. But if they can address that component and suggest that all value at some future point will only go through the token, then you can still DCF off earnings.
It doesn't have to be off a return of capital. And so I think it's just a broader crypto problem of this forced returning of capital, which kind of ties back to the other stuff. But, yeah, assuming that they weren't going to do this would have been kind of silly.
Yeah, I think the burn is irrelevant. It was already priced in. If they did something else, that would have been the shock to the market. And the buybacks had to slow down, right? What you don't want to do is—there are 2 scenarios. Either you slow them down now and they have cash to grow, and ultimately that's what you're buying the token for, or they probably have over $1 billion.
They probably made something like $700 million before the ICO, and then they raised $1.5 billion. So whatever happens, they probably have over $1 billion in the bank. I have no idea what they're spending it on. It's not like it's—I don't know.
I'm still a PUMP holder, by the way, but I think it makes sense in terms of stability. You'd rather them continue to bolster their balance sheet and grow that way versus this immediate return of cash, which you never see from any startup at this stage. You'd rather them be in a scenario where they have years of runway rather than one where they're running short on it and then doing this out of a position of desperation rather than one of strength.
I think that would be fine. The DCF point and stuff is all fine. It doesn't rely on distributions to shareholders, but the difference is that with equity, you have a contractual, legally enforceable claim on the underlying business, its cash, etc., right? I think the worry that people have with Pump is that they get Houdini'd, right?
For sure. I mean, not Houdini'd, but, yeah, that's been part of the discount, right? The discount is a combination of a lot of things—
The fear that you get Houdini'd? They're reducing the buyback in a position where they have over $1 billion in cash. Like you said, we don't know where their financials are.
Yeah, I think what would help them is some clarity—some transparency: disclosing what they've done, how much they've spent, and then potentially trying to create some legal tie-in. Because I don't know if they're going to do that.
Do the ACE themselves. They're running the ACE. I think the token would pump, I don't know, over 2x—for sure over 2x—if they do an ACE.
I agree with that. I mean, because you think about it, they pulled off a pretty amazing trade. They sold a boatload of tokens at $4 billion and then bought them back at like $1.8 billion to $3 billion—probably somewhere in that range.
I mean, it's not a trade because they burned them, but, yeah. The burn value was like half of what they actually spent on them at the time of the burn.
Yeah, well, I'm also wondering what their average purchase price was. To be honest, it's probably around $4.
Because they bought a lot above 2. They only spent a few months there, if that.
Yeah. By the way, we incubated MetaLeX, so full disclosure, but I do think MetaLeX solves literally all of this. It's the right structure for most teams. It could be dope for Pump, too.
And I guess there's probably legal reasons and all sorts of complexity with a business that size, with that much stuff going on, that we're not aware of. That's the tough thing with this: you want transparency, but there's probably a bunch of lawyers giving them very sound advice for why that doesn't make sense.
But, yeah, it also depends on what you're optimizing for. Hyperliquid was obviously an extreme example of this. They never raised money and gave back 100% of revenue to token holders, and that's been an insane success for them, right? But I guess Pump has more of a need to build product. I don't know.
I think the hope for everyone when Pump went live was that it would be another Hyperliquid, right? It's one of the only businesses in crypto that was generating that level of cash flow. I bought into that thesis and bought into the ICO.
And I think now it's harder without some link to the equity. It's just hard.
I mean, even without the link to the equity aspect, Hyperliquid has so much else going on and has had so much other stuff come to market since all of this—since Pump's ICO. And Pump hasn't really done anything, right? It's pretty much—
Yeah, they still have the daily revenue, but streaming didn't really work.
It’s kind of clear what the next step is.
9. What’s next for Pump, meme coin markets & crypto products
Definitely a combination of the two. It’s really not clear to me what Pump would buy—streaming, maybe, whatever. It didn’t really work. Maybe they try again. But it’s not super clear what they would branch into or what business they might buy, whereas for HYPE, it’s pretty clear what the roadmap looks like in terms of products they could ship, markets they can go after, and everything like that.
It just seemed a lot more clear when I bought my Pump tokens. Now I’m just a community holder. Yeah, I think going after new markets is a huge differentiator, obviously, because Pump is also still largely tied to the meme coin trading narrative.
That’s also the difference with Titan. Its revenue is well down from the 10/10 peak, or pre-10/10. It’s actually higher than it was a year ago, right, when market conditions were much better. So, yeah, it’s the dichotomy that I keep coming back to and wanting to kill myself over when I look at my Pump position.
[snorts]
The important mistake was that they had a massive user base, and there were a lot of products that were pretty popular but relatively commoditized. So it was all about being connected to the user directly and then offering something similar, but that hasn’t really materialized. Yeah, it’s crazy that they’re still making this level of revenue. It’s kind of wild.
People are skeptical of that, too, right?
Yeah. I think that’s reflected in the price, for sure.
100%.
They have a lot of future revenue.
Yeah, the future revenue, where the revenue comes from, the token unlocks—I’m pretty sure all this stuff is priced in pretty heavily. So if you get clarity in one of these directions, maybe you could see it turn around well. Surely someone would have done some analysis on the patterns. Has anyone?
Yeah, they’re basically power users of meme coins and token developers. Of course—the devs. There’s still a ton of volume that happens in that world. My question is basically: who’s getting fleeced still?
Dude, just go on YouTube and type in “live meme coin trading,” and you’ll see exactly who’s getting fleeced.
Yeah, but you have to go a centimeter deep and many miles wide because of how low a ceiling these things have, right?
Yeah. The whole idea is you basically launch a lot of them, and a handful of winners offset the majority of losers. Basically, these are like venture meme coin devs.
[laughter]
Hey, man, Pump.fun has charity coins now, okay? I saw they crossed 1 million in donations. There you go. They’re really turning a bright light on the industry. Pump.fun is really a bright guy.
Whatever happened to that Glass Half Full Foundation thing, where they were going to buy me? Did they just stop doing that? The glass became full and then overflowed. I think it was called the Glass Half Full Foundation.
I do think Pump.fun—if they can get another run of making people some money with something, like if we get another meme cycle—I mean, I doubt it ever happens the same way, right? So they’d have to be early to something else. But with the distribution that they have, you would think there could be some projects that would be interested in doing A's.
Like—well, they did. Pump can do A's, right? Yeah, but I mean, some AI project, some stuff relatively outside of it. It’s just tough because traditional fundraising, I think, is so hot right now.
Yeah. Why would you go into something different? But yeah, it’ll be interesting to see how it all plays out. I think that’s plenty. We still have some stuff on the docket. Maybe we’ll save it for next time, but I think we covered a lot of ground here. Appreciate you guys joining as always. Those who tuned in, appreciate you sticking with us. We would love to hear comments, feedback. Drop them in the chat below. We're going to We got some plans for Spain. They're doing some new things. And subscribe if you haven't already. This is the new home for the Delphi High Bind podcast. And we'll be back soon enough. Hopefully markets are just as exciting. I'm sure they will be. Appreciate you guys joining. 95K next cycle. Thanks everyone. Lock it in.