DeFi TVL, mid-2026
~$70B
Down from ~$115B in January.
DeFi, past the buzzwords.
Liquidity, lending, yield. Three words that hide a lot of machinery and a lot of ways to lose money. This is what they actually mean, with the math worked out, the honest question to ask about any yield, and a graveyard of real losses so you learn on someone else's dollar instead of yours.
not advice · plain-English map · verify before you deposit
Why this lesson exists
DeFi is finance with the middlemen replaced by code. That is the whole pitch, and it is real. But "no middleman" also means no one to call, no chargeback, no deposit insurance, no reversing a bad transaction. The rules are whatever the contract says, and the contract does not care that you didn't read it. This page teaches you to read it.
The vocabulary, before anyone can sell it to you.
Most DeFi pitches work because you don't want to admit you don't know a word. Here is every one you keep hearing, in the plainest version that is still true. Come back to this table when a thread starts throwing terms at you.
| The word | What it actually means | The catch |
|---|---|---|
| DeFi | Financial services (trading, lending, borrowing) run by public smart contracts instead of a company. | No support desk, no undo button, no insurance. |
| Liquidity | The pile of tokens sitting in a contract so people can trade or borrow against it. | It is other users' money, and it can leave fast. |
| Liquidity pool | A shared pot of two tokens that traders swap against. Providers deposit, traders pay fees. | You take on impermanent loss (section 04). |
| AMM | Automated market maker. The formula that sets the price inside a pool. No order book, just math. | The math can be arbitraged against you. |
| LP | Liquidity provider, or the receipt token proving you own a share of a pool. | Your share value moves with the pool, not your deposit. |
| TVL | Total value locked. Dollars sitting in a protocol right now. | High TVL is not safety. It is a bigger target. |
| Yield / APY | The return you are quoted, annualized. | Ask who pays it. Section 06 is the whole answer. |
| Staking | Locking a token to help run a network (or a protocol) and earning a reward for it. | Lockups, slashing, and "fake" staking that is just emissions. |
| Collateral | Assets you post so you can borrow against them. | Falls too far and it gets sold out from under you. |
| Liquidation | The protocol force-selling your collateral when the loan gets too risky. | Fast, automated, and it charges you a penalty. |
| Oracle | The feed that tells a contract what an asset is worth off-chain. | Fool the oracle, drain the protocol (section 07). |
| Stablecoin | A token designed to hold a fixed value, usually $1. | "Designed to" is not "guaranteed to." Ask what backs it. |
| Bridge | A contract that moves value between two chains. | The single most-hacked thing in all of crypto. |
Takeaway: every one of these is a normal idea wearing a jargon costume. A pool is a shared pot. Yield is someone paying you. An oracle is a price feed. Strip the costume off before you decide.
A pool is a vending machine that reprices itself.
Forget order books. Most DeFi trading runs on a formula. The classic one, from Uniswap, is x times y equals k. You put two tokens in a pot, say ETH and USDC. Multiply the two balances together and you get a number, k, that the contract refuses to let drop. Every trade has to keep that product the same.
Someone wants ETH? They add USDC and take ETH out, but they must leave the product at k. So the more ETH they pull, the more USDC each one costs. The price moves along a curve automatically. No market maker, no quotes, just arithmetic. That is the "automated" in automated market maker.
Source: Uniswap v2 constant-product model (x·y=k). https://docs.uniswap.org/contracts/v2/concepts/protocol-overview/how-uniswap-works
Where your fee income comes from
Every swap pays a fee (Uniswap v2 charges 0.30%) that gets added to the pool and split among providers. That is your paycheck for supplying liquidity. On a busy pool it adds up. The catch is what the traders and arbitrageurs quietly take from you at the same time, which is the next section.
Impermanent loss. The tax on being the pool.
This is the one that gets people. When you provide liquidity, you are the vending machine. If the price of your tokens moves, arbitrageurs trade against your pool to keep it in line with the rest of the market, and they pay you less than the move was worth. You end up with less than if you had just held the two tokens in your wallet. That gap is impermanent loss. It only becomes permanent when you withdraw.
"Impermanent" is a soft word for a real cost. Let's put actual numbers on it. No hand-waving, every line checks.
You deposit into an ETH / USDC pool, ETH at $2,0001 ETH + 2,000 USDC
Starting value$4,000
The constant, k = 1 × 2,0002,000
ETH rises 50%, to $3,000. Pool rebalances so ETH/USDC ratio = price—
New ETH in pool = √(k ÷ 3,000)0.8165 ETH
New USDC in pool = √(k × 3,000)2,449.49 USDC
Check: 0.8165 × 2,449.49= 2,000 ✓
Your pool is now worth (0.8165 × $3,000) + 2,449.49$4,898.98
If you had just held 1 ETH + 2,000 USDC$5,000.00
Impermanent loss−$101.02 (−2.02%)
The arbitrageurs took 0.1835 ETH (worth ~$550) and left you ~$449 in USDC. The ~$101 difference is the money the pool paid, piece by piece, to stay priced correctly. Formula check: 2·√1.5 ÷ (1 + 1.5) − 1 = −2.02%. Matches.
Notice the shape of it: the loss depends only on how far the price moved, not which direction. Up 50% or down 33% (same ratio the other way) costs the same. And it gets worse fast as the move gets bigger.
Source: closed-form IL for a 50/50 constant-product pool, IL = 2·√r/(1+r) − 1. https://speedrunethereum.com/guides/impermanent-loss-math-explained
| Price move | Impermanent loss | What it means for you |
|---|---|---|
| 1.25× / 0.8× | −0.6% | Barely noticeable. Fees usually cover it. |
| 1.5× / 0.67× | −2.0% | The example above. Real, small. |
| 2× / 0.5× | −5.7% | Now fees have real work to do. |
| 4× / 0.25× | −20.0% | You'd have been far better off just holding. |
| 10× / 0.1× | −42.5% | The pool kept almost none of the upside. |
Takeaway: providing liquidity is a bet that fees earned > impermanent loss. That is why the calmest pairs (two stablecoins, or an asset against its own staked version) are the popular ones. They barely move, so IL stays tiny and the fees are close to free money. Pair two volatile tokens and you are fighting that red curve with every price swing.
Lending markets. Post more than you borrow, or get sold.
The other big DeFi primitive is the lending market: Aave on Ethereum, Kamino and marginfi on Solana. It is a shared pool again, but for loans. Suppliers deposit assets and earn interest. Borrowers post collateral and take a loan against it. No credit check. The only thing keeping the lender safe is that every loan is over-collateralized, you have to post more than you borrow.
- Utilization is how much of the supplied pool is currently borrowed. Rates float with it: an empty pool pays suppliers almost nothing; a heavily borrowed pool pays a lot, to lure more deposits in. On Kamino, USDC supply has paid roughly 4–9% across 2026 depending on that demand.
- Borrow rate is what borrowers pay, and it climbs steeply once utilization passes ~80%. marginfi's SOL and USDC rates can jump from single digits to double digits in hours when demand spikes.
- Health factor is the number that decides if you live. It is your collateral times its liquidation threshold, divided by your debt. Above 1, you are fine. Below 1, you get liquidated.
You supply SOL as collateral$10,000
Liquidation threshold for SOL80%
You borrow USDC against it$6,000
Health factor = (10,000 × 0.80) ÷ 6,0001.33
Above 1.0, so the position is safe… for now✓
SOL drops. Collateral falls to $7,500−25%
Health factor = (7,500 × 0.80) ÷ 6,0001.00
One more tick down → liquidation firescollateral sold + penalty
A 25% drop in your collateral wiped out your safety margin, and you never touched the loan. Liquidation is automated and merciless: a bot repays part of your debt, seizes your collateral, and keeps a bonus (often 5–10%) for doing it. In a fast crash, thousands of positions liquidate in the same minute.
The looping trap
The popular Solana move in 2026 is "looping": supply a staked-SOL token, borrow SOL, buy more staked-SOL, repeat, to lever up a yield. It works until the staked token trades even slightly below SOL (a depeg) or rates spike. Then your health factor craters on a move you didn't think was possible, and the liquidation cascade does the rest. Leverage does not create yield. It rents it, and the rent comes due all at once.
Takeaway: borrowing in DeFi is not "get money and forget it." It is a live position you have to babysit. Know your liquidation price before you borrow, keep a fat margin, and remember the collateral, not the loan, is what moves you underwater.
Where the yield actually comes from.
This is the section that would have saved most people who got wrecked. Any time a protocol offers you yield, there is exactly one question that matters: who is paying me, and why would they keep doing it? There are only a few honest answers, and one dishonest one that keeps coming back in a new wrapper.
Source: yield-source taxonomy synthesized from AMM fee mechanics and the Terra/Anchor case (section 07).
- Trading fees (real). Traders pay to swap; liquidity providers collect. As long as people trade, this pays. It is the cleanest yield in DeFi. Just remember it comes with the impermanent-loss curve attached.
- Lending interest (real). Borrowers pay to borrow; suppliers collect. It rises and falls with how badly people want leverage. Cyclical, but real money from a real counterparty.
- Staking / network rewards (real, but issuance-funded). The chain pays validators new tokens for securing it. Genuine, but it dilutes every holder to pay the stakers. Look at the yield after inflation, not before.
- Token emissions / "farming" (dilutive, temporary). The protocol prints its own governance token and hands it to you to attract deposits. It can be huge and it can be fine, but it is marketing spend, not profit. When emissions stop, so does the yield, and everyone who farmed it is selling the token on your head.
- Ponzi / reflexive (it ends). The yield is paid out of new deposits, not any real activity. It always looks incredible right up until the inflows stop. If nobody can explain the paying counterparty in one sentence, you are the counterparty.
The 20% that ate $40 billion
Terra's Anchor protocol paid a flat 20% on a "stablecoin." Where did it come from? Not from borrowers, borrow demand covered a fraction. The rest was subsidized from a reserve, topped up by the founders and by new deposits. It was the fourth and fifth arrows dressed as the second. When people finally asked "who is paying this?" out loud, the answer was "nobody, for much longer," and the whole thing unwound in three days. Full autopsy in the next section.
Takeaway: a number without a payer is a trap. Before you chase any APY, say the sentence out loud: "I am earning this because ___ is paying me because ___." If you can't finish it with real economic activity, the yield is emissions or worse, and you are on the clock.
Where the risk hides. The graveyard.
Smart-contract bugs, oracle manipulation, depegs, admin keys, bridges. The abstract risks are easier to remember as tombstones. Here are three that each teach a different failure, with real dollars and real dates. None of these were bad luck. Every one had a warning that reads as obvious in hindsight.
Terra / UST
May 2022 · algorithmic depeg
~$45B
UST was an "algorithmic stablecoin," held at $1 not by cash in a bank but by a trading loop with its sister token LUNA. Anchor's 20% yield pulled in roughly 75% of all UST. In early May 2022, large withdrawals broke the peg, the loop went into reverse, and minting LUNA to defend the dollar crushed LUNA's price, which broke the peg further. A textbook death spiral. LUNA fell from over $80 to fractions of a cent in days; about $45B in market value evaporated. The lesson: a stablecoin is only as good as what backs it, and "an algorithm" is not backing.
failure: no real collateral + unsustainable yieldRonin bridge
March 2022 · bridge / key compromise
~$625M
Ronin was the bridge behind the game Axie Infinity, moving funds between its chain and Ethereum. Moving money only required sign-off from 5 of 9 validators. Attackers (later attributed by the FBI to North Korea's Lazarus Group) got hold of five validator keys and simply approved their own withdrawal: 173,600 ETH and 25.5M USDC, ~$625M. Nobody noticed for six days. The lesson: bridges concentrate enormous value behind a handful of keys, which is exactly why they are the most-hacked thing in crypto.
failure: centralized keys + no monitoringMango Markets
October 2022 · oracle manipulation
~$117M
A Solana lending and trading protocol that valued collateral using the live market price of its own thinly-traded token, MNGO. An attacker took a large position, then spent ~$4M buying MNGO across exchanges to spike its oracle price ~2,300% in minutes. Against that inflated "collateral" he borrowed out everything in the protocol, ~$117M. The code worked perfectly; the price feed was the weak point. He even argued in court it was a legal trade, and a judge later overturned the criminal convictions. The lesson: a protocol is only as honest as its oracle, and an oracle reading a manipulable market is a loaded gun.
failure: manipulable price oracle
The pattern under all three
None of these were "hackers broke unbreakable math." Terra trusted an incentive that couldn't last. Ronin trusted five keys. Mango trusted a price feed. Every DeFi disaster is a trust assumption someone forgot they were making. Your job before depositing is to find the assumption, because it is always there, and decide if you believe it.
The wider count is not comforting either: DefiLlama logged 121 hacks and roughly $942M lost across 2026 alone. This is not a solved space. It is a frontier with real money on it.
How to read a protocol before you trust it.
You don't need to audit Solidity. You need to ask the questions that would have flagged every graveyard entry above. Run this before any deposit. If you can't answer three of them, that is the answer.
01Who is paying the yield, in one sentence?Trading fees, borrower interest, or network issuance are real. "Emissions" or a number nobody can source is a countdown. If the sentence doesn't finish, walk.
02What backs the stablecoin?Real cash and treasuries you can see, over-collateralized crypto, or "an algorithm"? Terra answered this wrong. Ask it every time, even for the boring $1 tokens.
03Where does the price come from?The oracle. Is it a robust feed (Chainlink, Pyth) with sanity checks, or the spot price of a thin token that Mango-style money can shove around?
04Who can touch the money?Look for admin keys, upgradeable contracts, and multisigs. Is it 5-of-9 like Ronin, or a real timelock and a wide signer set? "Can the team drain this?" is a fair question with a real answer.
05Has it been audited, and by whom?An audit is not a guarantee, plenty of audited protocols got drained. But no audit at all, on a contract holding millions, tells you how seriously they take your money.
06Is there a bridge in the path?If your funds cross chains, a bridge holds them somewhere. That is the highest-risk leg in crypto. Know when you're using one.
07What's my liquidation price?If you're borrowing, know the exact collateral price that wipes you, and keep a margin fat enough to survive a bad day. Then check it, because a bad day always comes.
08Can I afford the whole thing going to zero?The honest floor. Smart-contract risk means any single protocol can go to zero regardless of your skill. Size every position as if it might.
Takeaway: DeFi is not a scam and it is not free money. It is a real financial system with the guardrails removed and the source code published. That is the trade: total transparency, zero safety net. Read the code's promises, find the trust assumption, size for the worst case, and it is a tool. Skip that and it is a slot machine that publishes its odds and hopes you don't check.
Start here first
← Back to the education portal
If any of this felt like a second language, do the wallet lesson before you touch a protocol. Back to the education portal → for crypto from your first wallet to your first swap, plus the trading and tax tracks.