Explainers
What Sui is, and what its on-chain data shows
Sui is a layer 1 blockchain that stores everything as objects instead of as balances inside accounts, and runs smart contracts written in Move. It launched in May 2023, built by Mysten Labs. Two years later, an exploit of the Cetus exchange ended with validators voting on a protocol upgrade that moved frozen funds without the attacker’s signature.
What Sui is for, and who maintains it
Sui is a general-purpose smart contract chain: tokens, exchanges, lending markets and other applications run on it as contracts. Its mainnet launched on 3 May 2023, according to the Sui Foundation’s retrospective, which calls Mysten Labs Sui’s original contributor; Mysten Labs marked the launch in a post dated the following day. The code is public in the MystenLabs/sui repository, and the Sui Foundation publishes the network’s announcements and post-mortems. What sets the chain apart from most others is how it stores state, and the language that changes it.
Objects instead of accounts
Sui’s documentation defines an object as the basic unit of storage: every asset, contract and piece of data on the chain is one. Where many chains keep a key-value store inside each account, Sui keeps objects addressed by unique IDs. Each has an owner, a version number that rises with every change, and the digest of the last transaction that used it.
Ownership is a field, and it decides how a transaction is processed. An object can be owned by an address or by another object; it can be immutable and owned by nobody; it can be shared so that anyone can use it; or it can be wrapped inside another object. A transaction names each input object by its ID, version and digest, so the sender and the validator agree in advance on exactly what is being changed. The documentation describes the resulting history as a directed graph of objects and transactions, rooted in a genesis transaction.
What the model buys is parallel execution. Transactions that touch different objects do not have to wait for each other, and the consensus documentation says transactions on different shared objects run on several cores at once; only those that write to the same shared object must run in sequence. Wallets have to handle this model, which is one of the reasons a wallet works on one chain and not another.
Move, and what it guarantees
Move began in 2018 inside Facebook’s Libra project, according to the Move Book, the language guide maintained for Sui, and several chains now run their own versions of it. Its type system gives every type explicit permissions, called abilities: copy, drop, key (the value can be stored as an object on Sui) and store. A type without copy cannot be duplicated, and one without drop cannot be thrown away, which is how the language models assets that must not be created or lost by accident.
Contracts are published as packages. The object model page says a published package is immutable and can never be changed or removed.
The arithmetic rules matter for what follows. The Move reference on integers says addition, subtraction and multiplication abort rather than overflow. Bit shifts are another matter: a left shift aborts only when the number of positions to shift by is too large for the type, so bits pushed off the top of a number are lost without an error. Code that needs a checked shift has to write the check itself.
How Mysticeti orders transactions
Sui uses delegated proof of stake. Holders delegate SUI to validators, and stake sets each validator’s voting power. The consensus documentation says the validators and their stakes stay fixed for an epoch of about twenty-four hours, and that a quorum is any set of validators with more than two-thirds of the voting power.
The consensus protocol, Mysticeti, builds a directed acyclic graph of blocks that several validators propose in parallel; the documentation says it needs three rounds of messages to commit a block, the same as practical Byzantine fault tolerance. A transaction goes to one validator, which checks it and puts it in its next proposed block; peers run the same checks when the block reaches them. Once consensus has ordered the block, every validator executes its accepted transactions against the same state. A transaction is final when a quorum of validators acknowledge the same effects, or when it appears in a certified checkpoint.
Validators are paid in proportion to stake, with one adjustment. During each epoch they rate one another’s performance, and the gas documentation says validators that do not process transactions promptly at the agreed reference gas price receive reduced rewards at the end of it.
SUI, gas and the storage fund
The tokenomics page gives SUI four uses: staking, paying gas, serving as an asset in applications, and voting on matters such as protocol upgrades. Total supply is capped at 10 billion. The mainnet launched without an airdrop, and a one-year cliff on early investors’ tokens ended in May 2024.
A gas fee has two parts, computation and storage. Computation is measured in coarse buckets and charged at a price per unit that validators set together: at the start of each epoch every validator states the lowest price it will accept, and the reference gas price is the two-thirds percentile of those quotes, weighted by stake. A user pays at least that, and anything above it works as a tip for priority.
Storage is paid in advance, for keeping data on the chain indefinitely. The fee goes into a storage fund, which counts as stake and earns rewards like any other stake. Those rewards go to validators to cover the cost of storing data and what remains is reinvested in the fund, so that the users who wrote data in the past pay the validators who keep it later. Deleting data returns a rebate: when read on 22 November 2026, the gas page put it at 99.99% of the original storage fee. Each epoch’s stake rewards are the computation fees collected plus a subsidy, which the tokenomics page says disappears in the long run as the SUI in circulation approaches the supply cap.
The Cetus exploit, and the vote that followed
Cetus runs concentrated-liquidity pools on Sui. On 22 May 2025, according to its incident report, an attacker exploited a flaw in an open-source library that its pool contract depends on, used it to add an enormous amount of liquidity for very few tokens, and then removed liquidity in repeated rounds to drain the pools. Its later relaunch plan adds that the attacker made a large number of swaps that disrupted pool prices.
The report traces the flaw to checked_shlw, a function in the integer-mate library that shifts a 256-bit number left by 64 bits and is meant to signal an overflow whenever any of the top 64 bits is set. As written, the report says, it signalled one only for numbers greater than 2²⁵⁶ − 2¹⁹², so a result that had overflowed passed as long as not all of those bits were set. Two commits to the library’s public repository on 22 May 2025 (first, second) change the check so that every number from 2¹⁹² upwards is flagged. That is the gap the Move reference leaves to the programmer: the language aborts an overflowing addition, but not a shift that pushes bits out.
Most of the funds did not leave Sui. The incident report says validators began voting to refuse transactions signed by the attacker’s addresses, and that once the votes passed a third of stake those addresses were in effect frozen; the two frozen Sui wallets held the main portion of what was taken, and the rest had been bridged to Ethereum. On 23 May Cetus said it was working to recover $60M of stolen funds, and asked the community to support a protocol upgrade to recover $162M of funds that had been frozen, adding that nobody could make that decision alone and that it would respect the result.
The Sui Foundation agreed to run the vote among validators, which it said represent their stakers. The proposal was an upgrade allowing one specified address to act as each of the two attacker addresses, for one pre-specified transaction each, and the Foundation said the mechanism could not be used for anything else. The funds would move to a 4-of-6 multisig wallet held in trust, with two keys each for Cetus, the Sui Foundation and the auditor OtterSec, and the Foundation signing only if the other signers could not. Votes were weighted by stake, with the Foundation’s own stake excluded, and the proposal passed if more than half of total stake, abstentions aside, voted yes or no and yes outweighed no. Voting opened on 27 May. The Foundation later added that validators representing 90.9% of stake had voted yes and that the vote had closed early.
Cetus’s plan set the relaunch for 03:00 UTC on 8 June 2025, and its report on the second quarter says the protocol was relaunched in under three weeks. The plan refilled the affected pools from the recovered assets, its available cash reserves of about $7M and a 30M USDC loan from the Sui Foundation, restoring between 85% and 99% of each affected pool’s liquidity, and covered the remainder with CETUS tokens, part claimable at the relaunch and the rest released monthly over twelve months.
DefiLlama’s hack record lists Cetus CLMM at $223M in May 2025, the second-largest entry of 2025 in the record (2025 incidents), classed as token and share accounting with the technique arithmetic error. The record does not say whether its amounts are before or after any recovery. How crypto gets hacked reads the record as a whole.
Network halts
A full halt came on 21 November 2024. According to the Sui Foundation’s post-mortem, mainnet halted completely between about 1:15 and 3:45 am Pacific time, with every validator stuck in a crash loop. The cause was an assertion in new congestion-control code, the part that limits how fast transactions can write to a single shared object, which crashed validators when a transaction’s estimated execution cost came out as zero. The fix shipped in version 1.37.4, and the network was running again fifteen minutes after its release.
The halt of 14 January 2026 was longer. A bug in consensus commit logic led validators to compute different commits, so no checkpoint could gather a quorum of signatures, and the network halted instead of finalising inconsistent state. The Foundation’s post-mortem put the disruption at about six hours and said no certified transactions were rolled back. Mysten Labs’ validators ran the fix first, as a canary, before the rest upgraded.
Another incident ran from 28 to 30 May 2026. Sui’s status page lists a critical incident titled Mainnet settlement. Its updates record a fix rolled out to validators, the network back online once more than two-thirds of stake had upgraded, and participation by stake back up to 93% when the incident was marked resolved. No post-mortem for it was found for this article.
What the on-chain data shows
As of 22 November 2026, per DefiLlama, Sui had $335.0M of total value locked, 31.8% lower than thirty days earlier, with 115 protocols listed and $584.3M of stablecoin supply on the chain (data page). The total is the value of assets deposited in DeFi protocols on the chain at the snapshot’s prices, leaving out liquid staking, bridges and deposits DefiLlama flags as double-counted. A fall of almost a third in a month can come from withdrawals, from lower prices for the tokens deposited, or from both, and the dollar figure alone cannot separate them (what TVL measures).
Lending leads the list of largest protocols, which is on a different basis from the total. Suilend at $115.2M, NAVI Lending at $85.7M and Current at $43.8M are all filed under Lending. SpringSui at $41.9M, Haedal Protocol at $23.2M and Volo LST at $16.2M are filed under Liquid Staking, a category the chain total leaves out. Ember, at $34.4M, is filed under RWA.
Two exchange entries connect the data to the rest of this article. Cetus CLMM, the exchange in the 2025 exploit, appears at $17.1M under Dexs. DeepBook V3 appears at $12.2M; the Sui documentation describes DeepBook as Sui’s native liquidity layer, a central limit order book for spot and margin trading.
Stablecoin supply, at $584.3M, is a separate measure from a separate DefiLlama feed: the supply circulating on the chain, whether or not any of it is deposited in a protocol. Where stablecoin supply lives explains how that feed differs from its own chain table. The site has no hack page for Sui, because DefiLlama’s record does not yet hold the twenty incidents a chain page needs.