At 3:47 a.m. Auckland time, unable to sleep, I pulled a single USDC-to-SOL quote off Jupiter and stared at it the way you stare at a stain on the ceiling. The number that came back showed 0.72 basis points of deviation from the consolidated mid-price. Nothing about it looked strange. What unsettled me was the absence: no pool address, no invariant formula, no x*y=k curve I could sketch on the back of an envelope. Just a price, a signature, and a settlement. Tracing the ghost in the machine, I found it did not live in the code at all. It lived in a trading desk in another time zone, running software nobody outside the firm is permitted to read.
That is the part I want to trouble you with. In the most liquid corner of the most performant chain โ the SOL/stablecoin pair, which clears more professional volume than nearly anything else in crypto โ more than 90% of the order flow routed through aggregators now executes against proprietary market makers. Not against public pools. Not against other users. Against firms.
Most DeFi users still believe they are swapping into a communal well. They are not. They are trading into a dealer's inventory, settled on a public ledger, and the dealer has never once been obliged to show them how the price was derived.
To understand how quietly this happened, you have to remember what the automated market maker was supposed to be. When Uniswap's constant-product curve went live, it was not simply a product. It was an argument. The claim was philosophical as much as technical: that public, open-source software could replace the professional intermediary entirely. Liquidity would come from anyone. Pricing would be an emergent property of a formula that anyone could verify. The market maker โ the person, the desk, the balance sheet โ would be abstracted away into math.
For a while, that argument held. Then the arithmetic caught up with the ideology. The problem had a name by 2022: LVR, or Loss-Versus-Rebalancing โ the systematic, quantifiable bleed that passive liquidity pools suffer because their prices lag the rest of the world. A pool quoting a stale curve is a pool that arbitrageurs can hit, again and again, at the expense of the people who deposited the capital. LVR is not a bug in the code. It is a structural tax on passivity, and it compounds.
That tax is the reason a new class of market maker appeared on Solana first, where low latency and sub-cent fees made on-chain professional dealing viable. Firms with names most DeFi users have never typed into a search bar โ HumidiFi, GoonFi, and a rotating cast of competitors โ began running their own inventory, their own quoting engines, and settling on-chain without ever inviting the public to deposit.
The architecture they sit inside is a three-layer stack, and it is worth naming precisely: a front end (the aggregator's interface) that shows one price; a routing layer that scans public AMMs, proprietary AMMs, and request-for-quote networks simultaneously; and an execution layer where the winning quote settles on the ledger. To the user, it is one click. To the market structure, it is a courtroom with the door closed.
Now the mechanism. A proprietary AMM โ propAMM โ is not a cleverer curve. It is a trading firm's balance sheet wearing a smart contract as a coat. The firm holds inventory, runs private software that continuously re-prices against external reference markets, and posts a signed quote. If you take it, the swap settles transparently on-chain. What does not settle on-chain, what is never published, is the reasoning. Why that price, at that moment, at that size?
That opacity is precisely what solves LVR. A passive pool cannot help quoting a stale number; a propAMM never quotes one, because it is watching the same venues the arbitrageurs watch. It re-prices before it can be picked off. In theory the tax disappears โ not by redistributing it, but by refusing to pay it. And the quotes are genuinely tight. Research from Jump Crypto put median deviation for these venues at 0.72 basis points, materially inside what passive pools deliver.
I want to be careful here, because I have been burned by tidy numbers before. My first newsletter, the Beacon Chain Tracker, grew to five thousand subscribers in 2017 partly because I preferred an exciting chart to a boring footnote. I have since learned the footnote is where the story lives. So: the two most-cited data sources on propAMM growth and quote quality are DWF Ventures and Jump Crypto โ one a market-making firm and investor, the other a market-making firm. Neither is a disinterested observer. The mechanism is convincing; the absolute percentages deserve a haircut.
Even discounted, the shape of the shift is unmistakable, and it contains a detail I have not seen anyone properly interrogate. PropAMMs account for somewhere between 15% and 27% of total DEX volume โ respectable, not dominant. But in the SOL/stablecoin market specifically, that figure exceeds 90%. The gap between those two numbers is the real news: professional dealers are not conquering DeFi broadly, they are conquering the segment where institutional-grade flow actually lives. They took the penthouse and left the public pools the lobby.
Follow that thread from code to culture and you arrive at a consequence with no comfortable framing. Public pools earn fees. Fees come from volume. If the volume that flows through a pool is the residual left after professional desks have skimmed the professional pairs, then LP fee revenue is compressed precisely where it used to be fattest. This does not merely reduce yield. It initiates a loop. Lower fees attract less capital; thinner pools quote worse; worse quotes lose more routing share; the propAMM's relative advantage widens. Passive liquidity is not being killed. It is being slowly repriced toward irrelevance in the only market that mattered.
There is a second cost, less visible but structurally heavier. DeFi's most valuable property was never decentralization as a slogan. It was composability โ the ability of one protocol to build on another's liquidity because the terms were public and the interfaces were open. A closed-source, permissioned market maker cannot be composed with. You cannot build a lending market, a structured product, or a hedging strategy on top of a price you are not allowed to inspect. The most liquid venues in the ecosystem are now opaque edges, and opaque edges do not compose. They terminate.
And then there is the dependency nobody advertises. Ask where a propAMM gets its sense of the true price. It does not emerge from the chain. It is pulled from centralized exchanges โ Binance, Coinbase, OKX, Bybit โ whose order books remain the final arbiter of value. Which means the more professional on-chain market making becomes, the more authority accrues to the centralized venues that set the reference. Solana's on-chain market is, in a real mechanical sense, a derivative of someone else's order book. Unearthing the human story behind the hash rate means acknowledging that the humans in question are sitting at centralized exchanges, and the chain is following them.
The routing layer adds a final complication that has received almost no scrutiny: the aggregator is now a gatekeeper. If proprietary market makers must be admitted in order to be seen, then whoever controls admission controls the flow of an entire market's order volume. That is not a protocol function. That is an infrastructure power, and it is exercised without published criteria, without public accountability, and without any obvious mechanism for appeal.
I have spent enough time in market microstructure to be wary of what comes next. In foreign exchange, dealers historically won the right to a last look โ a final instant to reject a trade if the market moved against them. It is one of the most contested practices in electronic trading. Nothing in a closed-source quoting engine prevents the same behavior, and nothing in the disclosure regime would reveal it. I am not asserting that it happens. I am noting that the structure permits it and the architecture cannot disprove it.
Concentration is the risk that keeps me up more than transparency. When more than 90% of a market's professional flow passes through a handful of firms whose balance sheets, hedges, and failure modes are invisible from outside, you have rebuilt a systemic risk profile on top of a system whose selling point was the absence of exactly that. One blown inventory position or one mispriced quoting engine does not produce a bad print. It produces a liquidity vacuum at the moment liquidity is most needed. And the public pools that might once have absorbed the shock have been hollowed out by the very dynamic that made the dealers dominant.
Regulators, meanwhile, are walking toward a door that is already open. Consider what a propAMM actually does: it commits its own capital, continuously quotes two-sided prices, and acts as a dealer in the technical sense of the word. The SEC's dealer rules and the interpretive guidance around them fit that description uncomfortably well. When Nasdaq, the LSE, and Robinhood eventually migrate real assets on-chain, the compliance bar will not stay where it is today. Best-execution obligations, quote transparency, auditability โ all of it will follow the assets upward, and a closed-source quoting engine will have a difficult time arguing that its pricing was fair when nobody could see it.
Which brings me to the observation I cannot shake. Artifacts of a new digital renaissance are supposed to be open, verifiable, and legible to anyone with a block explorer. What Solana has built in its most important market is the opposite: a reproduction of electronic dealer markets, wearing a blockchain as a settlement layer. It works. It is efficient. It solves a real problem. And it is, in almost every meaningful respect, the thing DeFi told everyone it had already replaced.
Here is where I will argue against myself, because the obvious conclusion โ that DeFi sold out โ is the least interesting one and possibly the wrong one.
The first counterargument is economic, not moral. Passive liquidity was never a viable business at the yields it advertised. Those yields were subsidized by token emissions, and when emissions dried up, the pools were left quoting a curve that bled to arbitrageurs for the privilege of existing. PropAMMs did not steal that business. They revealed that it was not a business. Repricing an unprofitable activity is not a betrayal; it is the market discovering what market making actually costs.
The second counterargument cuts deeper. The reflexive debate in crypto is about whether this is decentralization. I think that is a distraction, and I think the industry is looking at the wrong end of the machine. The genuinely contrarian reading is that the propAMM era does not threaten DeFi's decentralization โ it threatens DeFi's relevance to the institutions everyone is waiting for.
Consider the standard bull case: real-world assets will migrate on-chain, and public blockchains will capture the settlement layer. But an institution does not need a public pool to trade. It needs a counterparty with a balance sheet, continuous pricing, and confidential execution. That is what a propAMM is. That is also what every prime broker already provides, with better legal finality and a compliance department attached. Which means DeFi has spent five years building the market-making template for Wall Street โ and in doing so, it has demonstrated that Wall Street does not need DeFi's public infrastructure at all. It only needs the settlement rail. The three-year RWA storytelling exercise keeps assuming that institutions want the pool. They want the plumbing, and they are perfectly happy to bring their own liquidity.
Mapping the chaotic beauty of market sentiment is usually my favorite exercise. Here, sentiment is not chaotic at all โ it is simply wrong. The consensus mental model of a DeFi swap is a user trading against a pool. The reality is a user trading against an institution. That gap between perception and mechanism is the most valuable thing in this story, and the market has not yet priced it in either direction.
So, the road ahead. Watch for propAMM share crossing 30% of total DEX volume, not just the SOL/stablecoin sliver โ that is where the structural shift stops being a Solana peculiarity and becomes the default. Watch whether aggregators publish their admission criteria, because the moment those criteria become public, the gatekeeper becomes accountable. Watch the dealer rules, and watch whether passive pools respond with hybrid models that borrow professional quoting without borrowing the opacity.
And watch, most of all, the one question nobody in the room is asking. If the most liquid market on the fastest chain is now run by a handful of closed-source desks pricing off centralized exchanges, then what exactly did we decentralize?
We did not decentralize the market. We put a public receipt on a private decision.