Eleven bullet points arrived in my feed this week, forwarded three times before they reached me. Ten were opinions. One was a fact โ that someone had said something. The date stamp read October 25 and then stopped, offering no year, which means the note cannot be anchored to any point in the liquidity cycle: not to the ETF inflows of early 2024, not to the summer consolidation, not to whatever quarter we are actually standing in.
I have read enough crypto research to recognise the genre. A well-known investor's views, compressed into a list, circulated as though compression were the same thing as verification. Anthony Pompliano is bullish on Bitcoin, gold and land over the next twenty years. He considers AI badly undervalued, expects venture capital and public markets to converge, and believes capital will concentrate into a handful of quality companies. He hands young investors four sectors to watch and predicts a TeslaโSpaceX merger. The ledger remembers what the market forgets: not one of those claims carries a verifiable number. They are positioning, dressed as forecast.
Pompliano's product has never been the protocol. He came up through traditional finance and asset management, built a podcast network and a newsletter, and now runs a fund. His comparative advantage is narrative organisation โ taking trends that already exist in the data and packaging them into a direction retail capital can act on. That is a legitimate business. It is also a business whose output should be read as sentiment rather than research. When someone with a public book tells you an asset will perform well for twenty years, you are receiving a sell-side view with excellent distribution.
The macro map matters more than the man. Since the spot ETF approvals, the marginal buyer of Bitcoin has changed identity. In 2024 my firm spent most of the year translating on-chain metrics into compliance-safe language for more than fifty institutional clients, and the whitepaper we published on post-ETF liquidity flows ended up bringing in ten million euros of new mandates. The lesson from those meetings was not that institutions loved Bitcoin. It was that their questions changed. Nobody asked what Bitcoin is any more. They asked what it correlates to, who else holds it, and what happens to the order book when the ETF flow reverses.
That is the frame I brought to the note. Pompliano groups Bitcoin with gold and land โ the classic debasement trade, scarce assets against currency dilution. Coherent, old, and defensible. But the grouping does quiet work: it reclassifies Bitcoin as a hard asset and quietly retires the earlier story in which Bitcoin was a technology network. Those two framings have different holders, different correlations, and different failure modes.
A twenty-year horizon is not a forecast; it is a structure. It is long enough that no single observation can falsify it, and long enough to match the vocabulary of pensions and endowments, where allocation decisions are made on decade-scale mandates. The twenty-year horizon is an alibi, not a thesis. A claim becomes useful to an investor only when it specifies its own death: what would have to happen for it to be wrong? On Bitcoin the answer is knowable โ a sustained break in the halving-era security budget, a correlation regime that never decouples from risk assets, a regulatory reclassification. None of those appear in the list. What appears instead is duration.
The debasement frame also hides a measurement problem. Since 2020, Bitcoin's correlation to the Nasdaq has risen precisely when it matters โ in drawdowns, when diversification is supposed to pay. Its correlation to gold has been unstable in both directions. Grouping an asset with gold does not give it gold's behaviour; it gives it gold's marketing. I have watched institutional allocators build digital-gold sleeves and then discover, in the first stress week, that they had bought high-beta technology exposure with a commodity narrative attached. Volatility is not risk; impermanence is. The asset that survives a cycle is not the one with the best story but the one whose holders do not need the story to stay.
Then there is the part of the Bitcoin thesis that never makes the bullet points: the security budget. After the fourth halving in April 2024, the block subsidy fell to 3.125 BTC, and the next reduction lands around 2028 at 1.5625. Annual issuance is now under one percent of supply โ genuinely scarcer than gold's annual production growth. But miner revenue is subsidy plus fees, and fees have spent most of the last decade below five percent of the total. The network's security is still financed by inflation, and that inflation halves on a schedule. Meanwhile hash power keeps concentrating; the practical outcome is a small number of pools large enough to matter for block template construction. Decentralisation that survives in the whitepaper but not in the pool distribution is a consensus of paperwork. A twenty-year bullish case that does not address the fee market is a case that has not read its own cash flow statement.
The recommendation I found most interesting was compute. Pompliano lists compute markets among the sectors Wall Street has not yet dug into, and here I have direct experience, because in 2025 I helped build a decentralised compute market connecting AI researchers with GPU providers and ran a pilot with three labs. The technical debt in that sector is not marketing โ it is verification. You cannot easily prove that a rented GPU is the model and bandwidth advertised, you cannot easily prove the training job ran as specified rather than being quietly downclocked, and you pay a latency penalty for scheduling across geographies. Every serious compute network I have audited solves one of those three and hand-waves the other two. The binding constraint on decentralised compute is not demand; it is proof. Until attestation is cheap enough to be default, token subsidies will keep doing the work that product quality should be doing โ and when the incentives stop, the utilisation graph tells you who the real customers were.
We have a habit of this. The data availability layer was sold as the foundation of a rollup future, and most rollups today do not produce enough data to justify a dedicated availability layer. The demand was architected before it existed. We built the cathedral before the saints arrived.
The AI agent claim has the same ambiguity, in a more consequential place. Saying agents will reshape finance is not a forecast until you say which layer. There are three, and their difficulty differs by an order of magnitude. The execution layer โ agents holding keys, paying, settling โ is already partly buildable with account abstraction, scoped session keys, and intent-based routing; the hard problems there are revocation and auditability, not cryptography. The intermediary layer โ agents replacing brokers and advisers โ runs straight into fiduciary duty and KYC attribution, where no regulator has decided who is liable when an agent mis-sells. The clearing layer โ machine-to-machine settlement at volume โ needs standards nobody has ratified. Almost all of the near-term value sits in the execution layer, and almost all of the near-term headlines sit in the other two.
Which brings me to the only claim in the list with real information density: that venture capital and public markets are converging, and capital is concentrating into fewer, better companies. This one is observable rather than rhetorical. Companies stay private longer, IPO exits congest, secondary markets expand, and public investors reach primary assets through SPVs and tender offers. The direction is right. The conclusion people draw from it is not.
Concentration and returns are not the same variable, and the note collapses them. If capital crowds into the same head assets, those assets' valuations rise and their forward returns compress. That is not a theory; it is the 2024โ2025 AI primary market, where the strongest companies attracted enormous rounds at multiples that require flawless execution for a decade. The end state of concentration is winners whose valuations are too high, not winners whose returns are higher. Anyone reading the convergence thesis as a buy signal for late-stage private exposure is buying the valuation, not the trend.
Look at the four sectors as a set rather than a list: private equity, prediction markets, compute markets, data brokerage. What they share is not technology. It is a stage of maturity โ traditional finance has already proven the demand exists, while Web3 has not yet found a token model that captures it without subsidy. Prediction markets are the outlier in a good way: they generate real fee revenue from real events, one of the few non-incentivised income streams in the sector. Compute markets have paying customers whose behaviour is distorted by token emissions. Data brokerage has obvious demand and an unresolved question about who owns the data being sold. In three of the four, the direction is probably right and the current instruments probably do not capture the value. That distinction is the difference between a sector call and a trade.
One more structural omission deserves attention: the note has no year, and therefore no cycle position. That sounds like pedantry until you try to use it. A cluster of long-term-bullish-plus-AI-is-undervalued commentary is not randomly distributed across time โ it tends to appear when a narrative is accelerating, not when it is bottoming. Bottoms produce silence, capitulation, and people leaving the industry. Euphoric lists with four sector recommendations for young investors are a mid-cycle artefact. Read as a thermometer rather than a map, the note does say something useful: attention is rotating from the frontier to the foundation, and that rotation is usually early in a narrative and late in a price.
The contrarian read is that the recommendation list is not an opportunity list at all. It is a map of regulatory arbitrage.
"Wall Street has not yet dug in" is a filter with two faces. It means not yet priced, and it also means not yet validated. What the four sectors share at the structural level is that their excess return comes from a compliance cost differential โ the gap between what a lightly regulated venue can offer and what an incumbent can offer under its licence. That gap is real, and it is also temporary by construction. When the classification question is settled, the incumbent enters with an advantage the startup cannot match. CME does not need to invent event contracts; it needs a rulebook. A hyperscaler does not need to invent GPU markets; it needs a contract. Projects that treat regulatory ambiguity as a moat are renting their business model from a legislature. The ones worth owning are those with network effects that survive clarification โ liquidity that stays because it is deep, not because it is unobserved.
Prediction markets are the clearest test case. The live dispute is not whether the technology works. It is whether an event contract is a swap, a commodity, or a wager. That single determination rebuilds the cost structure of the entire vertical overnight, in either direction. Anyone who bought the sector because a fund manager listed it has bought a legal outcome, not a product.
So where does that leave a portfolio? Watching three things, none of them price. The event-contract classification, because it decides whether prediction markets are a derivatives business or a gambling business. The geography of GPU export control, because it decides which compute networks can serve which customers. And the first serious retail wrapper around private-market exposure, because that is the moment the convergence thesis stops being a structural observation and becomes a risk transfer to people who cannot price it. Stability is a myth; liquidity is the only truth. The question worth carrying into next quarter is not whether Bitcoin performs over twenty years โ it is who is holding the other side of that horizon, and whether they can stay.