The Evolution Of DeFi
From Early Experiments to 2025
Decentralized Finance started as a weird idea on the fringes of crypto: what if you could recreate every financial service, from lending to trading to savings accounts, without a single bank, broker, or intermediary? Just code, on a public blockchain, open to anyone with an internet connection.
That idea grew from essentially zero in 2017 to a $176 billion ecosystem at its peak in 2021. As of mid-2025, DeFi's total value locked sits around $95 billion. That's not a bubble that popped and disappeared. That's a new sector of finance that got tested hard, broke some things, learned, and kept going.
Here's how it happened.
The Starting Point: Ethereum and the First Idea (2015-2017)
Bitcoin proved you could move value peer-to-peer without a bank. But Bitcoin's scripting language was intentionally limited. You couldn't build complex financial logic on top of it.
Ethereum changed that. Launched in July 2015, it brought programmable smart contracts: code that lives on the blockchain, executes automatically, and can't be stopped or altered once deployed. Developers immediately recognized what this meant for finance.
The first serious attempt was MakerDAO. Danish developer Rune Christensen had proposed the idea on Reddit in 2015: a stablecoin issued by a DAO, backed by crypto collateral. In December 2017, it went live. Users could lock ETH into a smart contract and mint Dai, a stablecoin pegged to the dollar. No bank held the dollars. The peg was maintained entirely through over-collateralization and code.
Every Dai was backed by at least 150% in ETH. If your collateral fell below that threshold, the protocol automatically liquidated it. The MKR governance token gave holders control over the system's risk parameters, and made them "lenders of last resort" -- if the system ever had a deficit, MKR would be minted and sold to cover it. That aligned incentives sharply: govern badly and your tokens get diluted.
Dai became the first proof that a decentralized, stable currency was possible on Ethereum. It immediately became a building block for everything else.
Meanwhile, people were trying to build decentralized exchanges. The first attempts were order-book based and painful: every order or cancellation was an Ethereum transaction, gas costs made market making uneconomical, and liquidity was thin. EtherDelta, the most popular early DEX, even got its frontend hacked via a DNS attack. The contracts were fine; the experience was terrible.
A better approach was needed.
The AMM Breakthrough (2018-2020)
In November 2018, Hayden Adams launched Uniswap -- a simple automated market maker using a formula almost anyone can understand: x * y = k.
Instead of matching buyers with sellers, Uniswap used liquidity pools. Anyone could deposit two tokens into a pool and become a market maker, earning 0.3% on every trade. The price was set algorithmically based on the ratio of the two tokens in the pool. No order book. No active market maker required.
It was elegant in a way most DeFi things aren't. A few dozen lines of code created a functioning exchange for any ERC-20 token. Slippage existed, but the permissionless nature meant any project could list immediately by funding a pool. Uniswap's liquidity grew from $20M to $2.9B by 2020.
In parallel, lending protocols emerged. Compound launched in 2018 with a pooled lending model: deposit an asset, earn interest. Borrow against collateral. Interest rates adjusted algorithmically based on utilization -- high demand meant higher rates, low demand meant lower. No counterparties needed. Everything was transparent and automatic.
Compound's interest-bearing tokens (cTokens) were composable ERC-20s. You could deposit them elsewhere, use them as collateral, or trade them. This composability became DeFi's defining feature: protocols could plug into each other like Lego pieces.
Then in June 2020, Compound launched the COMP governance token and started distributing it to users. Yield farming was born. People deposited and borrowed just to earn COMP on top of their interest rates. TVL exploded. Every other protocol looked at what Compound did and copied it. "DeFi Summer" had arrived.
Aave came along around the same time and took the Compound model further. Flash loans were their signature innovation: uncollateralized loans that could be borrowed and repaid within a single transaction. If you don't repay, the transaction reverts as if it never happened. Suddenly anyone with the right code could access millions in liquidity for a few seconds -- useful for arbitrage, liquidations, and collateral swaps. Useful also for attacks on vulnerable protocols, which became a recurring problem.
By September 2020, DeFi TVL had crossed $10 billion. A year earlier it was under $1 billion. The infrastructure worked.
Curve, the Wars, and Stablecoin Liquidity
While Uniswap handled general-purpose token swapping, swapping between stablecoins (USDC to DAI to USDT) was still inefficient on a constant-product AMM. Michael Egorov launched Curve in January 2020 to fix this.
The Curve stableswap invariant is a hybrid formula that behaves almost like a constant sum (meaning near-zero slippage) when tokens are close to parity, and falls back to a constant product at extreme imbalances. For stablecoin trades, this gave Curve essentially zero slippage at scale. It became the backbone of stablecoin liquidity on Ethereum.
Then came the CRV token and veCRV governance -- arguably the most influential tokenomics design in DeFi.
CRV emissions flow to pools based on gauge weights. Gauge weights are set by veCRV holders, who lock CRV for up to four years to get voting power. The longer you lock, the more power you get. LPs who also hold veCRV get up to a 2.5x boost on their CRV rewards.
This created the Curve Wars. Any protocol that needed deep stablecoin liquidity needed CRV emissions flowing to their pool. Getting those emissions required accumulating veCRV. Getting veCRV meant buying and locking CRV. Convex Finance emerged to let people pool their CRV influence, and suddenly protocols were competing to accumulate CVX, bribing veCRV holders via platforms like Votium, and treating Curve gauge votes like a geopolitical resource.
Almost every stablecoin project that mattered in 2021-2022 was fighting over Curve gauge votes. It's the most creative piece of on-chain governance design in DeFi's history.
2021: Bull Run, Concentration, and Growing Pains
DeFi TVL went from $15B in January 2021 to $176B in November. Capital poured in. Protocols launched on Layer-2s as Ethereum gas costs became prohibitive for everyday users. Arbitrum, Optimism, and Polygon all became DeFi homes for Aave, Uniswap, and Curve deployments.
Uniswap V3 dropped in May 2021 with concentrated liquidity: LPs could now specify price ranges for their liquidity rather than spreading it uniformly from zero to infinity. Capital efficiency improved 10-100x for active price ranges. Professional market makers started treating Uniswap positions like order book strategies. The AMM increasingly resembled a traditional exchange in behavior if not in design.
Ethereum's Beacon Chain launched in December 2020, beginning the transition to Proof-of-Stake. Staking, however, required 32 ETH minimum and locked funds indefinitely until future upgrades. Lido launched almost immediately to solve this: deposit any amount of ETH, receive stETH in return. stETH accrues staking rewards while remaining usable in DeFi -- as collateral in Aave, in Curve pools, as a yield-bearing base asset for strategies.
Lido grew quickly. By the time the Ethereum Merge completed in September 2022, Lido had become the dominant liquid staking provider and stETH had become the most widely-held yield-bearing asset in DeFi.
2022: Stress Tests
2022 was the year DeFi's thesis got stress-tested at maximum pressure.
In May, Terra's algorithmic stablecoin UST collapsed. UST had no hard collateral backing it -- its peg relied on a mint-and-burn mechanism with LUNA. When confidence broke, the spiral was instantaneous and total. Tens of billions evaporated in days. Protocols that had accepted UST as collateral or relied on Terra's DeFi ecosystem were hit hard.
Notably, MakerDAO, Aave, and Compound all survived without systemic failure. Over-collateralization held. Automated liquidations worked. The lesson was reinforced: transparent, over-collateralized lending with code-enforced rules is more resilient than trusting a team's claims about reserves.
In June and July, centralized lenders collapsed: Celsius, Three Arrows Capital, BlockFi. Off-chain credit, no transparency, no automated liquidations -- the failures were slow, opaque, and complete. On-chain, Aave's liquidation bots ran continuously the whole time.
Regulatory attention also arrived. The OFAC sanction of Tornado Cash smart contract addresses in August 2022 was unprecedented: the US government sanctioned code on a public blockchain. Some DeFi frontends began geoblocking and filtering. The protocol-level code remained unstoppable; the user-facing surfaces faced pressure.
2023-2025: The Current Landscape
The bear market culled unsustainable yield schemes and zombie projects. What remained was more mature.
Ethereum's Shanghai upgrade in April 2023 enabled staking withdrawals, removing the last major risk from liquid staking. stETH became a redeemable, yield-bearing Ethereum native asset. Lido's TVL climbed to $22.6B, making it the second-largest DeFi protocol. Restaking via EigenLayer extended the model: stake ETH, earn ETH staking yield, and additionally secure other protocols for extra return.
Aave continued expanding, now deployed on 13 chains with $24.4B TVL -- the largest single DeFi protocol. Its V3 improvements (isolation mode for risky assets, supply caps, efficiency mode) tightened risk management while broadening asset support. Aave launched its own stablecoin, GHO, entering MakerDAO's territory.
MakerDAO moved further into real-world assets: US Treasury bills, tokenized bonds, real estate-backed loans. Maker's Endgame plan proposed spinning parts of the system into subDAOs. The protocol that started as a pure crypto-collateralized stablecoin system now holds significant TradFi assets as backing -- a pragmatic adaptation to scaling Dai's supply.
Uniswap runs on over 30 networks with $3.8B TVL spread across them. Total DEX volume regularly exceeds $5-10B daily.
As of mid-2025, the breakdown tells an interesting story. Lending protocols hold roughly $40B in TVL. Liquid staking holds $42B. DEXes hold $18B but generate the highest fee revenue among all categories. Ethereum still hosts 52% of all DeFi TVL despite the multi-chain expansion.