How a simple formula replaced Wall Street's market makers
On traditional exchanges, market makers are specialized firms with millions in capital. They quote buy and sell prices, pocket the spread, and keep markets liquid. It's a profitable, competitive, capital-intensive business dominated by firms like Citadel Securities and Virtu Financial.
Uniswap replaced all of that with a formula: x · y = k
No order books. No market makers. No permission required. Just a smart contract holding two tokens and a math equation determining the price.
This wasn't supposed to work. It did. Uniswap alone has processed over $3 trillion in cumulative volume1, and DEXes now handle 21% of centralized exchange volume—up from 6% in January 20212.
The constant product formula didn't emerge from nowhere. On June 22, 2017, Vitalik Buterin published "On Path Independence"4—a blog post exploring how on-chain market makers could serve as "always-available counterparties." The key insight: path independence protects against reserve bleeding attacks.
Alan Lu from Gnosis implemented the first version using only token balances. Bancor launched on June 12, 2017, becoming the first AMM-based DEX5. But the design that dominated came later.
Hayden Adams was laid off from Siemens in 2017. Karl Floersch, an Ethereum developer, introduced him to Vitalik's blog post. Adams spent months building what he initially called "Unipeg"—Vitalik suggested the name "Uniswap"6. It launched November 2, 2018, funded by a $65,000 Ethereum Foundation grant.
By May 2025, Uniswap became the first DEX to surpass $3 trillion in cumulative trading volume1. The formula that a laid-off engineer built in his spare time now processes more daily volume than most traditional exchanges.
Traditional exchanges match buyers and sellers through order books. Someone posts "I'll buy ETH at $2,000," someone else posts "I'll sell at $2,001," and the exchange matches them when prices meet.
This requires market makers willing to quote prices, enough liquidity to fill orders without massive slippage, complex matching engines, and trust in the exchange operator. Liquidity pools skip all of this.
Instead, a smart contract holds reserves of two tokens—say, ETH and USDC. The price isn't set by orders. It's determined by the ratio of tokens in the pool.
x · y = k
When you buy token A, you add token B to the pool and remove token A. This changes the ratio, which changes the price. The bigger your trade relative to pool size, the more you move the price (slippage).
The formula guarantees there's always a price—no "market closed" signs. Larger trades naturally cost more through slippage. Arbitrageurs keep prices aligned with external markets. And no one can manipulate the order book because there isn't one.
It's elegantly simple. It's also good enough that it bootstrapped trillion-dollar markets.
As of January 2026, major DEXes process substantial daily volume7:
| DEX | Daily Volume | TVL |
|---|---|---|
| Uniswap v3 | $2.29B | $2.79B |
| Curve | $2.20B | $2.25B |
| PancakeSwap | $1.90B | — |
| Uniswap v2 | $1.35B | — |
| Uniswap v4 | $700M | $1.0B+ |
Total Uniswap volume across all versions: approximately $4.3 billion daily. The protocol has generated over $4.94 billion in lifetime fees8.
For context: DEX to CEX volume ratio hit an all-time high of 37.4% in June 20252. The structural shift from centralized to decentralized trading is measurable and accelerating.
Anyone can add tokens to a pool and become a liquidity provider.
You deposit both tokens in the pool's current ratio—say, $1,000 of ETH and $1,000 of USDC. In return, you receive LP tokens representing your share of the pool.
When traders swap through the pool, they pay fees (typically 0.3%). These fees accumulate in the pool, increasing the value of LP positions.
When you want out, you burn your LP tokens and withdraw your share of both assets—plus accumulated fees.
It's passive market making. You provide capital, earn fees, and let the math handle pricing.
There's a reason professional market makers charge for their services. Providing liquidity has risks that most retail participants underestimate.
Impermanent loss is what happens when token prices change after you deposit. Here's how it works:
You deposit 1 ETH ($2,000) and 2,000 USDC into a pool. Your position is worth $4,000.
ETH doubles to $4,000. If you had just held, you'd have 1 ETH worth $4,000 plus 2,000 USDC—total $6,000.
But in the pool, arbitrageurs have been rebalancing. The constant product formula means your position now looks something like 0.707 ETH ($2,828) and 2,828 USDC ($2,828)—total $5,656.
You made money ($5,656 > $4,000), but less than if you'd simply held ($5,656 < $6,000). The $344 difference is impermanent loss.
It's called "impermanent" because if ETH returns to $2,000, your loss disappears. It becomes permanent when you withdraw at diverged prices.
Impermanent loss, despite its name, isn't the full picture. A 2022 paper by Milionis, Roughgarden, Moallemi, and Zhang introduced a more precise metric: Loss Versus Rebalancing (LVR)9.
LVR captures the cost of providing liquidity to arbitrageurs who trade against stale pool prices. Unlike impermanent loss, LVR persists even if prices return to their original levels. The mechanism:
Empirical estimates put LVR at 5-7% of liquidity annually10—hundreds of millions of dollars extracted from liquidity providers each year. Many of the largest pools are unprofitable when accounting for LVR.
A study by Bancor and IntoTheBlock analyzed 17 Uniswap v3 pools representing 43% of TVL3:
| Metric | Value |
|---|---|
| Total fees earned | $199.3M |
| Total impermanent loss | $260.1M |
| Net loss to LPs | $60.8M |
| Profitable positions | 53.5% |
| Profitable wallets | 48.25% |
The conclusion was stark: "The majority of addresses providing liquidity in Uniswap v3 were not making money."
This doesn't mean liquidity provision is always unprofitable. But it means the APY numbers you see on dashboards often don't account for impermanent loss, LVR, or gas costs. If you're depositing because the yield looks attractive, you should understand what's not included in that number.
Beyond LP losses, traders face extraction through Maximal Extractable Value (MEV)—profits extracted by miners/validators and searchers through transaction ordering.
The most common form affecting AMM users is the sandwich attack:
Estimated MEV extraction11:
MEV breaks down roughly as: 60% arbitrage (arguably neutral—keeps prices aligned), 30% liquidations (beneficial for protocol health), 10-15% malicious extraction (sandwich attacks, time-bandit attacks)12.
Solutions are emerging—private mempools, order flow auctions, MEV-aware AMM designs—but the problem persists.
The basic x · y = k formula was just the beginning.
Uniswap v3 introduced concentrated liquidity, letting LPs provide liquidity within specific price ranges. This can be up to 4,000x more capital efficient in tight ranges13—but requires active position management. Over 90% of the largest LPs change their positions within a month14.
Uniswap v4 launched January 31, 202515, introducing "hooks"—custom code that executes at specific points in the swap lifecycle. This enables:
As of January 2026, v4 has over 4,600 pools and processes ~$700 million daily. The average APY across v4 pools is 56.43%16—though this figure should be interpreted cautiously given the LP profitability research above.
Liquidity pools are infrastructure. They make other things possible.
New projects can create instant markets without exchange listings. Price differences across pools create arbitrage opportunities that keep markets efficient. Other protocols can build on pool liquidity—lending platforms, derivatives, aggregators. Anyone can swap any token, any time, without accounts or approval.
The existence of liquid on-chain markets is what makes DeFi work. Without pools, there's no price discovery, no collateral valuation, no functioning ecosystem.
Liquidity pools democratized market making. Anyone with capital can participate in an activity that used to require specialized firms and exchange relationships.
But "anyone can do it" doesn't mean "anyone should."
The research is clear: roughly half of Uniswap v3 LPs lose money. Impermanent loss is real. LVR extracts value continuously. MEV taxes every trade. Position management (especially with concentrated liquidity) requires attention and skill.
If you understand the mechanics and risks—and can actively manage positions—providing liquidity can be profitable. If you're depositing because APY numbers look high, you're likely the exit liquidity for someone who understands the game better.
The infrastructure is remarkable. The formula is elegant. The risks are real.
Uniswap surpassed $3 trillion in cumulative all-time volume in May 2025, becoming the first DEX to reach this milestone. Source: Uniswap announcement, CryptoTVPlus. ↩ ↩2
DEX to CEX spot volume ratio data from CoinGecko research, November 2025. The ratio increased from 6.0% in January 2021 to 21.2% in November 2025, with a peak of 37.4% in June 2025. Perpetuals DEX/CEX ratio grew from 2.1% (Jan 2023) to 11.7% (Nov 2025). ↩ ↩2
"Uniswap v3 Impermanent Loss" study conducted by Bancor and IntoTheBlock, analyzing 17 pools comprising 43% of Uniswap v3 TVL. Published 2022. ↩ ↩2
Vitalik Buterin, "On Path Independence," June 22, 2017. vitalik.ca ↩
Bancor launched June 12, 2017 with an ICO raising 390,000 ETH (~$153 million at the time). They filed a patent application for the bonding curve mechanism in January 2017. ↩
Hayden Adams' account of Uniswap's creation: Uniswap Birthday Blog. The $65,000 Ethereum Foundation grant funded initial development. ↩
DeFiLlama DEX dashboard, retrieved January 19, 2026. Figures represent 24-hour volume snapshots and fluctuate significantly. ↩
Cumulative protocol fees across all Uniswap versions. Source: Dune Analytics, DeFiLlama. ↩
Milionis, J., Roughgarden, T., Moallemi, C.C., and Zhang, A. "Automated Market Making and Loss-Versus-Rebalancing." 2022. The paper formalizes LVR as a more accurate metric than impermanent loss for measuring LP costs. ↩
LVR estimates of 5-7% annually come from a16z crypto research and subsequent empirical validation. See: a16z Crypto - LVR Research ↩
MEV extraction data compiled from multiple sources: Flashbots, EigenPhi, and academic research. The $1.3B figure represents cumulative extraction; weekly figures vary significantly with market conditions. ↩
MEV composition estimates from ESMA (European Securities and Markets Authority) research report on MEV implications for crypto markets, 2025. ↩
Uniswap v3 documentation states concentrated liquidity can provide "up to 4000x capital efficiency" compared to v2 when providing liquidity in a tight range around the current price. Practical efficiency gains are typically lower. ↩
Analysis of Uniswap v3 LP behavior shows high position turnover. Source: Academic research on LP strategies, Dune Analytics. ↩
Uniswap v4 launched January 31, 2025 on Ethereum mainnet. The "hooks" architecture allows custom code execution during swaps. ↩
Uniswap v4 statistics from DeFiLlama and DWF Labs research, January 2026. APY figures are gross yields before accounting for impermanent loss and LVR. ↩
How a simple formula replaced Wall Street's market makers
On traditional exchanges, market makers are specialized firms with millions in capital. They quote buy and sell prices, pocket the spread, and keep markets liquid. It's a profitable, competitive, capital-intensive business dominated by firms like Citadel Securities and Virtu Financial.
Uniswap replaced all of that with a formula: x · y = k
No order books. No market makers. No permission required. Just a smart contract holding two tokens and a math equation determining the price.
This wasn't supposed to work. It did. Uniswap alone has processed over $3 trillion in cumulative volume1, and DEXes now handle 21% of centralized exchange volume—up from 6% in January 20212.
The constant product formula didn't emerge from nowhere. On June 22, 2017, Vitalik Buterin published "On Path Independence"4—a blog post exploring how on-chain market makers could serve as "always-available counterparties." The key insight: path independence protects against reserve bleeding attacks.
Alan Lu from Gnosis implemented the first version using only token balances. Bancor launched on June 12, 2017, becoming the first AMM-based DEX5. But the design that dominated came later.
Hayden Adams was laid off from Siemens in 2017. Karl Floersch, an Ethereum developer, introduced him to Vitalik's blog post. Adams spent months building what he initially called "Unipeg"—Vitalik suggested the name "Uniswap"6. It launched November 2, 2018, funded by a $65,000 Ethereum Foundation grant.
By May 2025, Uniswap became the first DEX to surpass $3 trillion in cumulative trading volume1. The formula that a laid-off engineer built in his spare time now processes more daily volume than most traditional exchanges.
Traditional exchanges match buyers and sellers through order books. Someone posts "I'll buy ETH at $2,000," someone else posts "I'll sell at $2,001," and the exchange matches them when prices meet.
This requires market makers willing to quote prices, enough liquidity to fill orders without massive slippage, complex matching engines, and trust in the exchange operator. Liquidity pools skip all of this.
Instead, a smart contract holds reserves of two tokens—say, ETH and USDC. The price isn't set by orders. It's determined by the ratio of tokens in the pool.
x · y = k
When you buy token A, you add token B to the pool and remove token A. This changes the ratio, which changes the price. The bigger your trade relative to pool size, the more you move the price (slippage).
The formula guarantees there's always a price—no "market closed" signs. Larger trades naturally cost more through slippage. Arbitrageurs keep prices aligned with external markets. And no one can manipulate the order book because there isn't one.
It's elegantly simple. It's also good enough that it bootstrapped trillion-dollar markets.
As of January 2026, major DEXes process substantial daily volume7:
| DEX | Daily Volume | TVL |
|---|---|---|
| Uniswap v3 | $2.29B | $2.79B |
| Curve | $2.20B | $2.25B |
| PancakeSwap | $1.90B | — |
| Uniswap v2 | $1.35B | — |
| Uniswap v4 | $700M | $1.0B+ |
Total Uniswap volume across all versions: approximately $4.3 billion daily. The protocol has generated over $4.94 billion in lifetime fees8.
For context: DEX to CEX volume ratio hit an all-time high of 37.4% in June 20252. The structural shift from centralized to decentralized trading is measurable and accelerating.
Anyone can add tokens to a pool and become a liquidity provider.
You deposit both tokens in the pool's current ratio—say, $1,000 of ETH and $1,000 of USDC. In return, you receive LP tokens representing your share of the pool.
When traders swap through the pool, they pay fees (typically 0.3%). These fees accumulate in the pool, increasing the value of LP positions.
When you want out, you burn your LP tokens and withdraw your share of both assets—plus accumulated fees.
It's passive market making. You provide capital, earn fees, and let the math handle pricing.
There's a reason professional market makers charge for their services. Providing liquidity has risks that most retail participants underestimate.
Impermanent loss is what happens when token prices change after you deposit. Here's how it works:
You deposit 1 ETH ($2,000) and 2,000 USDC into a pool. Your position is worth $4,000.
ETH doubles to $4,000. If you had just held, you'd have 1 ETH worth $4,000 plus 2,000 USDC—total $6,000.
But in the pool, arbitrageurs have been rebalancing. The constant product formula means your position now looks something like 0.707 ETH ($2,828) and 2,828 USDC ($2,828)—total $5,656.
You made money ($5,656 > $4,000), but less than if you'd simply held ($5,656 < $6,000). The $344 difference is impermanent loss.
It's called "impermanent" because if ETH returns to $2,000, your loss disappears. It becomes permanent when you withdraw at diverged prices.
Impermanent loss, despite its name, isn't the full picture. A 2022 paper by Milionis, Roughgarden, Moallemi, and Zhang introduced a more precise metric: Loss Versus Rebalancing (LVR)9.
LVR captures the cost of providing liquidity to arbitrageurs who trade against stale pool prices. Unlike impermanent loss, LVR persists even if prices return to their original levels. The mechanism:
Empirical estimates put LVR at 5-7% of liquidity annually10—hundreds of millions of dollars extracted from liquidity providers each year. Many of the largest pools are unprofitable when accounting for LVR.
A study by Bancor and IntoTheBlock analyzed 17 Uniswap v3 pools representing 43% of TVL3:
| Metric | Value |
|---|---|
| Total fees earned | $199.3M |
| Total impermanent loss | $260.1M |
| Net loss to LPs | $60.8M |
| Profitable positions | 53.5% |
| Profitable wallets | 48.25% |
The conclusion was stark: "The majority of addresses providing liquidity in Uniswap v3 were not making money."
This doesn't mean liquidity provision is always unprofitable. But it means the APY numbers you see on dashboards often don't account for impermanent loss, LVR, or gas costs. If you're depositing because the yield looks attractive, you should understand what's not included in that number.
Beyond LP losses, traders face extraction through Maximal Extractable Value (MEV)—profits extracted by miners/validators and searchers through transaction ordering.
The most common form affecting AMM users is the sandwich attack:
Estimated MEV extraction11:
MEV breaks down roughly as: 60% arbitrage (arguably neutral—keeps prices aligned), 30% liquidations (beneficial for protocol health), 10-15% malicious extraction (sandwich attacks, time-bandit attacks)12.
Solutions are emerging—private mempools, order flow auctions, MEV-aware AMM designs—but the problem persists.
The basic x · y = k formula was just the beginning.
Uniswap v3 introduced concentrated liquidity, letting LPs provide liquidity within specific price ranges. This can be up to 4,000x more capital efficient in tight ranges13—but requires active position management. Over 90% of the largest LPs change their positions within a month14.
Uniswap v4 launched January 31, 202515, introducing "hooks"—custom code that executes at specific points in the swap lifecycle. This enables:
As of January 2026, v4 has over 4,600 pools and processes ~$700 million daily. The average APY across v4 pools is 56.43%16—though this figure should be interpreted cautiously given the LP profitability research above.
Liquidity pools are infrastructure. They make other things possible.
New projects can create instant markets without exchange listings. Price differences across pools create arbitrage opportunities that keep markets efficient. Other protocols can build on pool liquidity—lending platforms, derivatives, aggregators. Anyone can swap any token, any time, without accounts or approval.
The existence of liquid on-chain markets is what makes DeFi work. Without pools, there's no price discovery, no collateral valuation, no functioning ecosystem.
Liquidity pools democratized market making. Anyone with capital can participate in an activity that used to require specialized firms and exchange relationships.
But "anyone can do it" doesn't mean "anyone should."
The research is clear: roughly half of Uniswap v3 LPs lose money. Impermanent loss is real. LVR extracts value continuously. MEV taxes every trade. Position management (especially with concentrated liquidity) requires attention and skill.
If you understand the mechanics and risks—and can actively manage positions—providing liquidity can be profitable. If you're depositing because APY numbers look high, you're likely the exit liquidity for someone who understands the game better.
The infrastructure is remarkable. The formula is elegant. The risks are real.
Uniswap surpassed $3 trillion in cumulative all-time volume in May 2025, becoming the first DEX to reach this milestone. Source: Uniswap announcement, CryptoTVPlus. ↩ ↩2
DEX to CEX spot volume ratio data from CoinGecko research, November 2025. The ratio increased from 6.0% in January 2021 to 21.2% in November 2025, with a peak of 37.4% in June 2025. Perpetuals DEX/CEX ratio grew from 2.1% (Jan 2023) to 11.7% (Nov 2025). ↩ ↩2
"Uniswap v3 Impermanent Loss" study conducted by Bancor and IntoTheBlock, analyzing 17 pools comprising 43% of Uniswap v3 TVL. Published 2022. ↩ ↩2
Vitalik Buterin, "On Path Independence," June 22, 2017. vitalik.ca ↩
Bancor launched June 12, 2017 with an ICO raising 390,000 ETH (~$153 million at the time). They filed a patent application for the bonding curve mechanism in January 2017. ↩
Hayden Adams' account of Uniswap's creation: Uniswap Birthday Blog. The $65,000 Ethereum Foundation grant funded initial development. ↩
DeFiLlama DEX dashboard, retrieved January 19, 2026. Figures represent 24-hour volume snapshots and fluctuate significantly. ↩
Cumulative protocol fees across all Uniswap versions. Source: Dune Analytics, DeFiLlama. ↩
Milionis, J., Roughgarden, T., Moallemi, C.C., and Zhang, A. "Automated Market Making and Loss-Versus-Rebalancing." 2022. The paper formalizes LVR as a more accurate metric than impermanent loss for measuring LP costs. ↩
LVR estimates of 5-7% annually come from a16z crypto research and subsequent empirical validation. See: a16z Crypto - LVR Research ↩
MEV extraction data compiled from multiple sources: Flashbots, EigenPhi, and academic research. The $1.3B figure represents cumulative extraction; weekly figures vary significantly with market conditions. ↩
MEV composition estimates from ESMA (European Securities and Markets Authority) research report on MEV implications for crypto markets, 2025. ↩
Uniswap v3 documentation states concentrated liquidity can provide "up to 4000x capital efficiency" compared to v2 when providing liquidity in a tight range around the current price. Practical efficiency gains are typically lower. ↩
Analysis of Uniswap v3 LP behavior shows high position turnover. Source: Academic research on LP strategies, Dune Analytics. ↩
Uniswap v4 launched January 31, 2025 on Ethereum mainnet. The "hooks" architecture allows custom code execution during swaps. ↩
Uniswap v4 statistics from DeFiLlama and DWF Labs research, January 2026. APY figures are gross yields before accounting for impermanent loss and LVR. ↩