CeFi or DeFi? You are choosing which failure you can live with

The wrong question and the right one

"Is DeFi safer than CeFi" is the question everyone asks and it has no useful answer, because the two categories fail in ways that are not comparable on a single scale.

A centralised lender fails when its borrowers default, its risk management is poor, or its management is dishonest — and you find out in a bankruptcy filing. A DeFi protocol fails when its code contains an exploitable flaw or its price oracle reports something wrong — and you find out on-chain, immediately, along with everyone else.

But the failure that actually costs retail users money in DeFi is neither of those. It is themselves: a phishing front-end, a malicious token approval, a transfer on the wrong network, a mistyped address. All permanent, all unrecoverable, and collectively far more costly than every protocol exploit combined.

So the useful question is: which of these can you actually manage? If you can evaluate a counterparty but would panic at a MetaMask signing prompt, CeFi is safer for you. If you are comfortable with a wallet but have no way to assess a private lending book, DeFi is safer for you. Both answers are correct for different people.

Side by side

Do the rates actually differ?

Less than the headlines suggest, once you normalise properly.

CeFi's top rates are genuinely higher — Nexo's 9.5% on USDT at platinum tier and Ledn's 8.5% on USDC above $100,000 exceed anything blue-chip DeFi offers. But both carry conditions: a portfolio threshold, a native token holding, a balance tier. And both are lending rates with undisclosed counterparties.

Compare instead the no-conditions rates. A flat 4% at a registered exchange against Aave's 3.8% to 5.2% on USDC supply. Those are within noise of each other, and the Aave figure comes with a publicly verifiable, over-collateralised loan book.

On Ethereum the comparison inverts. Lido's liquid staking at 3.8% to 4.1% net of a flat 10% fee beats exchange staking at roughly 3.5% net of a 25% to 35% commission. Here DeFi pays more precisely because it charges less for the same underlying reward.

What 2022 proved, and what it did not

DeFi advocates cite 2022 constantly and they are largely right to, but the lesson is narrower than the slogan.

What happened: Celsius, Voyager and BlockFi filed for bankruptcy. Genesis halted withdrawals, freezing roughly $940 million belonging to 340,000 Gemini Earn customers. Over the same period Aave and Compound liquidated undercollateralised positions on schedule and let depositors withdraw whenever they wanted.

What that proves: over-collateralised, transparent lending survives stress better than opaque, under-collateralised lending. That is a real and important finding.

What it does not prove: that decentralisation as such is protective. Plenty of DeFi protocols failed in the same period — algorithmic stablecoins, undercollateralised experiments, insufficiently audited forks. The blue chips survived because of collateralisation discipline and audit history, and a well-run centralised lender with the same discipline would have survived too. Some did.

There is also an epilogue worth knowing. Gemini Earn customers eventually received $2.18 billion in kind — about 97% of assets owed — by mid-2024, and the SEC dismissed its case with prejudice in January 2026 citing full recovery. Many recovered more dollar value than they lost because Bitcoin had risen. That was luck and three years of litigation, not protection.

The cost nobody quotes

Gas. On Ethereum mainnet, entering and exiting a position can cost tens of dollars, and a few transactions on a $500 position will consume more than a year of yield difference.

This single factor decides the question for most small balances, and it is almost never mentioned in CeFi-versus-DeFi comparisons. Below a few thousand dollars on mainnet, DeFi is simply uneconomic regardless of the rate.

The fix is layer 2. Base, Arbitrum and others run the same protocols — Aave, Compound, Lido are all deployed across several — for cents per transaction. Starting there rather than on mainnet changes the arithmetic entirely, and it is the first thing we tell anyone trying DeFi for the first time.

How to choose

Choose CeFi if you have under a few thousand dollars, you would not be comfortable being solely responsible for a seed phrase, you want fiat on and off ramps, or you simply do not want this to be a hobby. Those are all good reasons and none of them is a failure of sophistication.

Choose DeFi if you are comfortable with a wallet, you value being able to verify what backs your position, and you have enough capital that a layer 2 position is worth setting up. Start with lending supply on a blue-chip protocol, not with liquidity provision.

Choose both if you have a meaningful balance. The failure modes are genuinely uncorrelated — a centralised lender's borrowers defaulting and a smart contract being exploited have nothing to do with each other. Splitting across both is among the cheapest diversification available in this market.

What we would not do is choose based on the rate. Once conditions are stripped out, the rates are close enough that the decision should be made on which risk you can actually manage.