Uniswap founder Hayden Adams published his first blog post since 2019 on Monday night, arguing that automated market makers will take over the world’s largest markets once tokenized assets begin trading against each other rather than against dollars. By Tuesday afternoon, former XTX Markets trader Brian Huang had fired back with a blunt line: 「AMMs are going to zero.」
The dispute is not entirely about mechanism. Most critics who pushed back still accepted Adams’ basic point that pairing assets that move together reduces risk for liquidity providers. The split came elsewhere: whether traders actually want those pairs, and whether the capital Adams expects to sit in those pools is even allowed to do so.
Adams says onchain liquidity already clusters around correlated assets
Adams’ post, titled Correlated Pairs: How AMMs Win the Biggest Markets, drew 229,000 views, 885 likes and 142 replies in its first 19 hours. His central claim is that onchain liquidity has already organized itself into clusters: Ethereum ecosystem assets against ETH, Solana ecosystem assets against SOL, and stablecoins against one another. The reason, he said, is simple. Liquidity providers lose less when the two assets they hold move together.
「No one designed that,」 Adams wrote. 「It emerged organically.」 He argued that tokenization could push the same structure into equities, with pairs such as NVDA/USD shifting toward NVDA/SPY, while SPY/USD would serve as the bridge back to dollars.
He built the case on the 50th anniversary of the index fund
Adams anchored the argument in the 50th anniversary of the index fund. By Jack Bogle’s own account, the First Index Investment Trust closed its public offering on Aug. 31, 1976 after targeting $50 million to $150 million and raising only about $11.3 million.
Today, according to the Investment Company Institute, index mutual funds and ETFs hold $21.88 trillion, ahead of the $18.83 trillion held in active funds, for a 53.7% share.
Adams drew a direct parallel: passive liquidity provision can displace professional market making in the same way passive investing displaced active management. He cited Citadel Securities trading close to 25% of U.S. equity volume, a figure president Jim Esposito also used in November, as well as record net trading revenue of $12.2 billion on roughly $21 billion of trading capital. Citadel Securities does not publish financial statements; Bloomberg reported both numbers, citing people familiar with them.
「Most people read those numbers as proof the system works,」 Adams wrote. 「I read them as entrenchment.」
His mechanism is capital cost. A market maker hedges price exposure and pays for that hedge. An investor who already wants to hold both NVIDIA and SPY takes that exposure at no additional cost. If an issuer must pay a market maker to quote its asset, Adams argued, that amounts to a negative cost of capital. The tighter the correlation, the smaller the gap between a passive AMM curve and an active strategy, and the easier it becomes to undercut the firm.
Brian Huang says the model does not fit real market making
Brian Huang, co-founder of onchain portfolio app Glider and previously a trader at algorithmic trading firm XTX Markets, rejected that conclusion.
「At XTX Markets, I traded 4% of all US Equities volume on any given day. None of it ever went through an AMM and none of it ever will,」 he wrote. XTX publicly shares only a firm-wide figure of about $250 billion in daily traded volume across 35 countries, without a separate breakdown for U.S. equities.
Huang listed five objections covering execution through distribution. In his view, market makers need to place and cancel orders across thousands of assets and at varying depths, something an AMM cannot do. In an AMM, he said, a participant puts up liquidity in a range, accumulates impermanent loss and gets picked off by takers: 「you put up liquidity in a range and then acquire impermanent loss while getting picked off by takers.」
He also argued that managing inventory onchain means modeling network congestion and paying gas, while traditional venues rely on co-located hardware and measurable latency percentiles. AMMs do not separate order flow either, which led him to another pointed line: 「Wintermute should not be getting the same pricing as Joe in Minnesota.」
Huang then turned Adams’ democratization argument on its head. 「Retail participants should not be making markets,」 he wrote. 「The worst thing for retail to be doing is trading against other retail.」 He pointed to eye-catching yields on thin pools as evidence, saying headlines like 「452% on CASHCAT」 obscure the fact that many retail users do not understand how AMMs work.
The Defiant reported this month that Merkl campaigns on Robinhood Chain were advertising annualized yields between 1,274% and 36,540% on Uniswap v4 pools holding between $358 and $20,510.
Huang’s alternative is the structure tokenized equity issuers are already using. 「All of the tokenized equity issuers are moving to a model similar to @Ondo whereby stocks are minted/redeemed through RFQ (no AMMs),」 he wrote, followed by dark pools and single-dealer platforms that stream quotes directly.
Critics focused on demand: who wants to trade SPY against NVDA?
The most widely cited objections centered on demand rather than supply. Katia Banina, chief executive of Wintermute-incubated trading venue Bebop, called Adams’ post 「very good」 and the index fund analogy 「strong (if deceptive),」 then challenged the pair structure itself.
「From the trading side though, those correlated pairs make little sense,」 she wrote.
Her reasoning was straightforward. 「People don't trade financial assets against dollars out of necessity, but because dollars are money that buy goods and services. So most people would still want to do SPX/USD or NVDA/USD. So if the major pool for NVDA is paired with SPX, then you have to hop 2 pools and pay 2 fees.」 She also disputed the idea that the correlation is tight enough to matter, noting that 「single name vs index are very far from perfectly correlated.」
Adams anticipated part of that routing argument in the post itself, writing that 「Investors can still buy and sell everything in dollars, since routing across pools is automatic,」 but he did not directly deal with the doubled fee. Huang endorsed Banina’s framing: 「Ofc correlated pairs limit impermanent loss, but as you've said, who tf wants to trade SPY/NVDA?」
Even so, Banina did not write off AMMs entirely. 「LP-ing has been a long-standing strat for years,」 she said, adding that such pools generate arbitrage flow that both DEXs and chains want. 「So I think AMMs might linger some more.」
Supporters said correlated pools still have routing and market-design value
Derek Barrera, founder of liquidity-management protocol Steer Protocol, made a related point from another angle. Correlated pairs are where liquidity tends to form because impermanent loss is lower, he said, but 「those edges don't generate a ton of volume.」 His case for deploying them anyway was routing: keeping depth between correlated markets can capture multi-hop flow when uncorrelated pairs move out of line.
Luca Prosperi, co-founder and chief executive of stablecoin issuer M^0, read Adams’ post favorably but still landed on cost. 「If there is one thing where AMMs do not excel, is capital efficiency,」 he wrote. 「AMMs are excruciatingly capital inefficient.」 He expects intent-based frameworks providing just-in-time liquidity to sit alongside the pools, not replace them.
Guillaume Lambert, founder and chief executive of Uniswap options protocol Panoptic, answered Huang by rejecting the framing itself. 「Of course market makers will complain they can't do what makes them $ on an AMM. AMMs are not a product for market makers. Period,」 he wrote. In his view, one of the biggest mistakes made by researchers is applying spot market-making concepts to Uniswap LPs.
On Huang’s first objection specifically, Lambert wrote: 「Yes, that's the point. Not having to pay someone to update quotes every 10ms is why AMMs are better for most assets.」
The most direct supporters of Adams’ thesis came from investors and Uniswap-linked accounts. Variant founder Jesse Walden called the post 「a banger」 and said much of tokenization today is 「sustaining innovation for incumbents,」 while AMMs plus tokenization would shift 「market making from active/professional to passive/automated.」 Kenneth Ng, co-founder of the Uniswap Foundation, pointed to memecoins paired against tokenized stocks as a live example and wrote that 「AMMs enable programmable markets, but after almost $5t traded we're still early.」
Permissioning may block the model for regulated assets
Julian Kwan, chief executive of Singapore tokenization platform InvestaX, accepted the capital-cost mechanism and then pointed to where it breaks for regulated assets.
「Permissioned Pools enforce the allowlist at the pool level, and the allowlist belongs to the issuer,」 he wrote. 「So the answer to who can be the LP is whoever the issuer approves. Granted, not open.」
He also distinguished the assets Adams cited. 「The ten stocks against SPY work because those tokens are price exposure, not ownership. Private credit against a tokenized Treasury fund is a different regime.」 The Defiant has reported on that distinction in its coverage of Robinhood’s stock tokens.
Uniswap v4 hooks are part of the next step in Adams’ thesis
Adams pointed to Uniswap v4 hooks as the second half of the thesis and named the DualPool hook shipped on July 22. The hook parks pool inventory in ERC-4626 vaults between swaps so passive capital can earn lending yield. Spark co-designed it and runs the largest deployment.
Replying to a comment on Tuesday, Adams said he 「might have to do a version of the blog post that is 5x the length at some point that goes a bit deeper on some of the tradfi analogies.」
Uniswap fees, DEX share and UNI market data added context
According to DefiLlama, Uniswap generated $1.93 million in fees and $176,941 in revenue on Aug. 17, and $81.2 million in fees and $5.75 million in revenue over the past 30 days. Protocol revenue is routed entirely to UNI buyback-and-burn. The v4 fee switch was turned on July 27, nearly tripling protocol revenue.
CoinGecko data showed UNI trading at $3.30 late Tuesday, up 0.5% over 24 hours and down 11.8% over the week, with a market capitalization of $2.06 billion and 24-hour volume of $125 million.
On market structure, DEX spot volume is running at about 20% of centralized spot volume based on DefiLlama component data, and 24% on The Block’s monthly measure, which the report described as a record. Adams’ own post put the figure at 「over 20%.」