What actually goes wrong in crypto earn

There is no insurance. Anywhere.

This is the sentence to internalise before anything else, because a great deal of the marketing in this category is designed to make you forget it.

A bank deposit in the United States is covered by the FDIC up to a statutory limit. In the United Kingdom, the FSCS does the equivalent. Most developed markets have a comparable scheme. If the bank fails, you are made whole up to the limit, by a government-backed fund, quickly.

No such scheme covers a crypto earn balance anywhere in the world. If the platform fails, you are an unsecured creditor in a bankruptcy proceeding. You may recover some, most or none, and it may take years.

Some platforms advertise insurance. Read what it covers. Almost always it is custody insurance against theft or a security breach — genuinely useful, and entirely unrelated to the platform lending your coins to someone who does not repay, or to the platform itself becoming insolvent. Those are the events that have actually cost people money.

Counterparty risk: the one that has actually cost people money

When you deposit into a crypto savings account, your coins leave your control. You hold a claim on the platform, and that claim is only as good as the platform's ability to honour it.

The platform then lends your coins to someone. In the disciplined version, the borrower posts collateral worth substantially more than they borrow, and the system sells that collateral automatically if its value falls towards the loan amount. Your protection is arithmetic.

In the undisciplined version, the platform lends to large counterparties on relationship terms with thin or no collateral, because those counterparties pay more and volume looks good in investor updates. When one of them fails, the loss lands on depositors who had no idea the loan existed.

You generally cannot tell which version you are in from the outside. That is the central problem with centralised lending, and it is why proof-of-reserves reporting and a published collateral policy carry so much weight in our scoring.

The 2022 case history, in detail

The standard retelling is "crypto lending blew up". The detail is more useful than the slogan, because the failures were not uniform.

Celsius, Voyager and BlockFi all filed for bankruptcy in 2022. Each had taken on risk that customers could not see, in books that were not disclosed, funded by retail deposits attracted with high advertised rates.

Gemini Earn is the most instructive case because Gemini itself did not fail. The programme launched in February 2021 offering up to 7.4% APY, and it worked by lending customer assets to Genesis Global Capital. When Genesis halted withdrawals in November 2022 amid the post-FTX credit crisis, roughly $940 million belonging to 340,000 Earn users was frozen. The SEC charged Genesis and Gemini in January 2023 with the unregistered offer and sale of securities through the programme, and Genesis later agreed to a $21 million penalty.

Aave and Compound, meanwhile, kept operating. Loans were liquidated on schedule, depositors withdrew whenever they wanted, and neither protocol missed a payment. The difference was not centralised versus decentralised in the abstract — it was over-collateralised versus not, and transparent versus opaque.

The epilogue matters too. Through the Genesis bankruptcy between May and June 2024, Earn customers received $2.18 billion in kind, representing about 97% of digital assets owed and roughly a billion dollars more in value than when withdrawals halted — because Bitcoin's price had risen sharply in the interim. A customer who had lent one Bitcoin got one Bitcoin back, worth far more than in January 2023. The SEC dismissed its case against Gemini with prejudice in January 2026 citing full investor recovery, and in August 2026 an arbitrator found Gemini not at fault for the collapse.

The MiCA gap almost nobody mentions

This is the most consequential thing on this page that is not widely understood, and it applies to anyone in the European Economic Area.

MiCA's transitional period closed on 1 July 2026. Any crypto-asset service provider without CASP authorisation cannot legally serve EU clients from that date. Several major platforms now hold those authorisations and advertise them prominently, and rightly so — it is a real framework with real supervision.

But MiCA does not address crypto-asset lending and borrowing. Its safeguarding arrangements explicitly do not apply to assets used in lending programmes. The European Commission has identified DeFi, staking, lending and borrowing as areas beyond MiCA's original scope, with a targeted consultation extended to 30 September 2026 that may eventually lead to amending legislation.

The practical consequence is a trap. A platform can hold a genuine MiCA authorisation, display it accurately, and offer you a savings product that sits entirely outside its protection. Customers reasonably assume that everything offered by an authorised firm enjoys MiCA protection. It does not.

MiCA also prevents stablecoin issuers from paying interest on their tokens, since the design intent is to stop e-money tokens competing with bank deposits. That constrains issuers rather than third-party platforms, but it shapes how stablecoin yield is structured inside the EU.

The other five risks

Price risk. The largest by a wide margin and the least discussed on earn landing pages. A 10% reward on an asset that falls 40% is a 34% loss. Yield never compensates for direction, and staking a volatile asset does not reduce your exposure to it.

Depeg risk on stablecoins. USDC traded meaningfully below a dollar in March 2023 when part of its reserves sat at a bank that failed. It recovered within days, but anyone needing liquidity during that window took a real loss. A 5% annual yield does not compensate for a 10% depeg you cannot exit.

Smart-contract and oracle risk in DeFi. Code can contain exploitable flaws and price feeds can report wrong values. This is real and has cost the sector billions, though it is heavily concentrated in new and unaudited protocols rather than in the blue chips.

User error. In DeFi specifically, this causes more retail losses than every protocol exploit combined. Phishing front-ends that clone real interfaces, malicious token approvals, wrong-network transfers, mistyped addresses — all permanent, all avoidable with discipline.

Tax surprise. Rewards are generally ordinary income at the value on the day received, which can leave you owing tax on a value that has since fallen. Our tax guide covers this properly.

A checklist that catches most of it

Twenty minutes per platform. We run this before every position we take and it has kept us out of more than one venue that later had problems.

Beyond the checklist, three habits matter more than any single check. Size positions against risk, so a high-yield lender is always a smaller allocation than a conservative one. Split across counterparties if the balance is meaningful — it costs a fraction of a percentage point and removes single-point-of-failure exposure. Re-check quarterly, because the warning signs in this market — falling rates, changing terms, slower withdrawals, quiet feature removals — appear well before the problems do.

None of this makes crypto earn a bad category. Over-collateralised lending worked through the worst crisis this market has had, staking rewards were paid without interruption throughout, and the surviving operators are meaningfully better run than their predecessors. The point is that the protections you are used to from banking do not exist here, and the diligence has to come from you instead.