In 1996, roughly 8,000 companies traded on US exchanges. Today, fewer than 4,000 do. Over the same stretch, the top ten names grew from a rounding error of total market cap to roughly half of it. Two data points, one reflex: the street calls it "concentration" and moves on to the next chart.
Wrong frame. Concentration is the symptom. The disease is a liquidity migration — capital abandoning the public tape for private structures — and it casts a crypto-shaped shadow that most of this industry still refuses to look at directly.
The trigger isn't a token. It's a bell. Warburg Pincus, one of the oldest global growth private-equity franchises — 60 years in business, north of 200 portfolio companies — rang the NYSE opening bell to mark the anniversary. Then CEO Jeffrey Perlman sat for an interview and said, without hedging, that the IPO is structurally broken.
No crypto desk blinked. They should have. Because Perlman described, in private-equity vocabulary, the exact market structure this industry has spent three years pretending it invented — and then misread the direction of the flow.
Context: What Perlman Actually Said
Strip the fund-manager polish and the argument is mechanical, not rhetorical.
The public market is shrinking as a venue, not as a store of value. The count of listed companies has halved since 1996. The survivors consolidated. The top ten US names now carry roughly half of aggregate market cap. When the investable universe narrows to a handful of mega-caps, "index exposure" stops being diversification and becomes a single concentrated bet on one AI-narrative cluster. You don't own the market. You own the winners of a narrow race, at the price of assuming the race stays narrow.
The IPO has stopped functioning as price discovery. Perlman's numbers — a 30-to-40% listing discount plus a three-to-four-year lockup — mean going public costs the seller real money and real time. A 35% haircut at listing, then years of illiquidity, isn't a financing event. It's a tax on being public. So issuers route around it: strategic buyers, all-cash, immediate settlement, no discount, no lockup, no roadshow.
Capital is rotating, not evaporating. If public markets offer fewer targets and worse exits, the marginal dollar moves to where the structure is better — private credit, growth equity, alternatives. Warburg is a direct beneficiary. The question Perlman raises himself — can the private market absorb the capital public markets are shedding? — is the entire ballgame, and he knows it.
Rates are structurally higher, and it matters less than you'd think. Higher debt costs should make leveraged buyouts harder. Perlman shrugs. Why? Because the buyers doing the acquiring are cash-rich strategics, not leveraged financial sponsors. When the acquirer pays from its own balance sheet, the cost of debt is irrelevant to the transaction. That one sentence quietly demotes the leveraged-buyout model — the thing PE is famous for — to a second-tier exit channel. The exit that survives higher rates is the one financed by someone else's cash pile.
AI is Warburg's largest single investment — and it isn't a model. Not a frontier-lab stake. Not a foundation-model bet. AI implementation inside traditional portfolio companies. The "sell shovels to the gold rush" trade, dressed in fund language.
Now hold those five points next to crypto's three-year refrain that "real-world assets are coming on-chain." The misalignment isn't a nuance. It's structural, and it's the whole reason the RWA trade has produced more decks than P&L.
Core: The Same Migration, Two Different Maps
Here is where I stop summarizing and start deconstructing, because the surface read — "institutions are tokenizing RWAs, therefore bullish" — is precisely the error that keeps a certain class of newsletter alive and its subscribers poor.
The migration Perlman describes is real. The destination crypto assumes is not.
Mechanics, on-chain, the way I'd run them on a flash-news drop at 3 a.m. Jakarta time.
1. The denominator problem is a market-structure default, not a crypto bug.
Perlman's frame: more capital (numerator) chasing fewer public targets (denominator). Crypto has the identical disease in mirror image. Two assets — BTC and ETH — absorb the overwhelming majority of aggregate market cap and liquidity, while thousands of "listed" tokens scrap over the remainder. We call that concentration too, and we blame narratives, VC unlocks, retail rotation.
Wrong diagnosis. Concentration is what happens to any venue once listing becomes cheap and exit becomes expensive. Public equity halved its listed count because the cost of being public — disclosure, litigation, the 30-40% discount — exceeded the benefit. Crypto multiplied its listed count by orders of magnitude because the cost of being listed — a contract deploy, a pool seed — collapsed toward zero, while the exit (real liquidity, real buyers at a real clearing price) stayed scarce.
Two opposite listing curves. One identical result: a handful of names hold the depth, and everything else is a long tail of thin books. The long tail isn't a market. It's a queue of exits that can't clear — which is exactly Perlman's problem, wearing a ticker instead of a CUSIP.
2. Token launches are crypto's IPO — and they fail the same way, for the same reasons.
Perlman's 30-40% IPO discount has a direct on-chain analogue, and it isn't subtle: the airdrop dump, the TGE-day fade, the post-launch bleed that traders now price in before the token exists. That discount is the market telling you the float is mispriced against the lockup that follows.
Launch day is a promise; the code is the betrayal.
Look at the structure honestly. A token generation event is an IPO with worse governance, no audited financials, and a lockup enforced by a vesting contract instead of an underwriter. The three-to-four-year IPO lockup Perlman describes reappears on-chain as cliff-and-linear vesting — same economic function, less accountability, faster settlement. When the unlock hits, price behaves like a stock coming off lockup. Every cycle. Because it is the same mechanism, running on rails that forgive nothing and reconcile instantly.
One line for why the 2024-2025 launch calendar bled the way it did: the token market imported the IPO's structure, inherited the IPO's disease, and left behind the one thing that made IPOs work — price discovery with genuine buyers on the other side of the trade.
3. The capital isn't coming to your public chain. It's going to custody wrappers.
This is where people get angry, so precision matters.
When the RWA narrative says "institutions are coming on-chain," the implied picture is capital flowing onto open, permissionless rails — your L2, your DEX, your pool. I've spent real hours tracing where these flows actually settle, and the answer is boring and unambiguous.
Tokenized treasuries — the BlackRock BUIDL class, the Franklin BENJI class, the Ondo structures that followed — settle into wrappers where the token represents a custody claim, not a bearer asset sitting in a permissionless pool. The chain is a ledger of record for the issuer. It is not a venue where an anonymous participant provides liquidity and earns the spread. The flow is real. The openness is cosmetic.
Based on my audit experience tracing these flows, the pattern never varies: the institution uses the chain as a settlement layer it controls, and routes around every part of the chain it can't. You get the block-explorer screenshot. You don't get the order flow.
That isn't conspiracy. It's arithmetic. A regulated manager cannot custody client assets in a pool it doesn't control, can't freeze, and can't reconcile against its auditor's ledger. So it builds the wrapper, rents the chain for settlement finality and audit trail, and keeps the custody, the yield, and the spread. Every time.
4. On-chain private credit is the honest version of the thesis — and it's small.
Where the RWA story has genuine teeth is private credit: Maple, Goldfinch, Centrifuge, and the tokenized-fund structures that followed. Here the mechanics are real. A loan is originated, collateralized, serviced; the token is a claim on the cash flow. It's the same trade Warburg runs, in a smaller, more transparent, more fragmented package.
And it's tiny — because on-chain private credit carries the same problems as off-chain private credit: you need borrowers, underwriters, and a workout process for when loans go bad. The chain solves settlement and transparency. It does not solve origination, credit judgment, or recovery. Those are the hard parts, and no token standard touches them.
Perlman's rotation — capital moving from public to private — is exactly the tailwind that should push on-chain private credit. And it does, at the margin. But the institutional capital doing that rotation is flowing to Blackstone, Apollo, and Warburg's own vehicles, not to a permissionless lending pool. The rotation is real. The routing is walled.
5. Rates transmit to crypto through the exit, not the entry.
Structurally higher rates hurt leveraged portfolio companies — refinancing pressure, interest drag, longer holds that compound both. That's the conventional read, and it's half-right.
The half everyone misses: higher rates don't stop the cash-rich strategic buyer, and the strategic buyer is the exit channel that survives. So the durable exit in a high-rate world is the one financed by a corporate balance sheet. Which means the venture-style crypto exit — token listing, retail float, exchange liquidity — is competing for capital against a channel that doesn't feel rates at all. Crypto's exit model is rate-sensitive. Its competition's isn't. That asymmetry decides who clears.
6. AI is the tell, and crypto is reading the wrong line.
Warburg's largest single investment is AI implementation — retrofitting AI into 200+ traditional companies. Not betting on which model wins. Betting on the unglamorous, cash-generating work of upgrading incumbents.
I ran a version of this in 2025, partnering with two AI startups to test autonomous agents executing smart-contract interactions. The finding that mattered wasn't the agents. It was that value accrued to whoever controlled the execution environment, not whoever wrote the agent. Same lesson as the RWA flow: the margin lives at the layer you control, and it lives in service delivery, not in the primitive.
Crypto's AI-agent narrative keeps pricing the primitive — the agent, the token, the model — while the money, by Warburg's own allocation, sits in the implementation layer that touches real P&L. A fund putting its largest check into "AI that fixes a portfolio company's margins" is a signal about where durable returns live. It is not on your token.
Contrarian: The Angle Nobody Will Print
Here's what the interview coverage leaves on the floor.
The institutions doing this rotation do not need a public blockchain. They need a database with an audit trail — and a public chain is a worse database for their purposes than a permissioned one they own.
The RWA thesis assumes institutions want decentralization. They don't. They want settlement finality, auditability, and counterparty control. A public chain delivers the first two and actively denies the third. So they take the first two and engineer around the third. Every time.
That's why "RWA on-chain" has been a three-year storytelling exercise: the story assumes a demand — permissionless access, composability, censorship resistance — that the actual buyers of these products are structurally unable to want. A regulated fund cannot want composability. Composability is a liability to any entity that must reconcile positions and freeze accounts on subpoena. You can't sell a feature that is, from the buyer's seat, a bug.
The deeper asymmetry: the same public-market concentration Perlman laments is what drives capital into private structures — and those private structures are the least likely to adopt open rails. The more capital rotates to private equity, the less of it needs a public chain. Crypto's biggest structural bet is riding a migration whose endpoint is the opposite of what the bet assumes.
Now the counter-argument, because a thesis that can't survive its own stress test isn't a thesis. The obvious pushback: "Permissioned today doesn't mean permissioned forever. Rampers, interop layers, and composability bridges eventually reconnect the walled gardens to open rails."
I've heard it for three years. It fails on sequencing. The walled garden gets built first, captures the assets, and sets the standards — and by the time "reconnection" is technically possible, the economics are already settled in the incumbent's favor. The bridge becomes a toll booth, not a commons. Permissionless access arrives, if it arrives, as a premium feature sold back to you, priced by the custodian. That's not adoption. That's tenancy.
Crypto also runs its own version of Perlman's front-end-strong, exit-end-weak moat asymmetry.
Watch the venues. The exchange that absorbed a multi-billion-dollar fine is now more entrenched, not less — because the fine came bundled with licenses, and licenses are the deepest moat in the business. A new venue cannot buy its way to the same regulatory standing; the entry ticket is measured in years and legal budgets, not in code. That's the crypto edition of Warburg's franchise: the moat isn't the product, it's the permission. Warburg's real edge is LP relationships and a 60-year record, not any single deal. Front-end strong; exit-end — the liquidity, the buyer, the clearing price — weak and cyclical. Same asymmetry, different asset class.
Layer-2 fragmentation is the third mirror. Dozens of L2s, the same scarce user base, liquidity sliced into ever-thinner fragments. That isn't scaling. That's the private-market fragmentation problem in on-chain clothing — too many venues, too little depth, everyone competing for the same exit that won't clear. Fragment the denominator and you don't create markets. You create queues.
Chaos is just data we haven't labeled yet. Here's the label: every structure in this piece — public equity, private equity, tokenized RWAs, L2s, exchanges — is fighting the same war over the same scarce resource. Depth. Real buyers at a real clearing price. The venues that have it win. The venues that don't, tokenize, relist, and rebrand.
Takeaway
Perlman's post-mortem is right about the disease and honest about the symptom. Capital is leaving public markets. IPOs are structurally broken as exits. Private structures are absorbing the flow.
What he didn't say — because he doesn't have to — is that the private structures absorbing that flow are the least likely to settle on open, permissionless rails. They'll rent the chain for its audit trail and keep the custody, the yield, and the spread.
So here's the question the next twelve months must answer, and the one crypto keeps dodging: if the capital migrating out of public markets is landing in walled private structures, and those structures route around your public chain, then what exactly is the tokenization thesis a thesis about — the technology, or the story we tell ourselves while the depth sits somewhere we can't reach?
Arbitrage isn't just liquidity waiting for a mirror. Sometimes it's liquidity walking out a door you never noticed — and locking it behind you.