What is Sui (SUI)?

Sui's central idea is that most transactions have nothing to do with each other, so making them queue is waste. A coin transfer from you to a friend does not touch the same state as a swap between two strangers, and a chain that processes them one after another is throwing away throughput for no reason.

So Sui does not have accounts with balances. It has objects, each owned by an address, and a transaction declares which objects it touches. Two transactions touching disjoint objects run in parallel with no ordering between them at all. Simple transfers skip full consensus entirely and finalise in milliseconds; only transactions touching shared objects — a DEX pool, an auction — go through the consensus path.

That is a real architectural insight rather than a parameter tweak, and it is why Sui feels fast in a way that is not just marketing. It comes from Mysten Labs, founded by engineers who worked on Meta's Diem before it was abandoned, which is also where Move came from.

Move, and why it matters for safety

Move treats assets as first-class types the language itself protects: a coin cannot be duplicated or accidentally discarded because the type system forbids it. A whole family of Solidity bugs is unrepresentable. Sui's dialect extends this with the object model, which is more expressive than Aptos's account-based Move and also further from anything a developer has used before.

The outages

Sui halted three times during 2026. The most serious was on 28 May: a gas charging logic bug in release v1.72 sent every validator into a crash loop, stopping the chain for six hours and 44 minutes and freezing roughly $1 billion of on-chain assets. SUI fell about 8% during the incident. It was the second outage of the year at that point, with another following.

The pattern is the one Solana had: a young, complex, highly optimised codebase with limited client diversity, where a defect in a release takes the whole network down together. Solana's answer took four years and a second independent client. Sui does not have one yet.

• Gas for transactions, in a fee model where storage is paid for up front and partially refunded when objects are deleted.
• Staking to validators, which secures the network and pays about 1.49%.
• Collateral and liquidity across Sui DeFi, which grew quickly from a small base.
• Gaming and consumer applications, the category the object model genuinely suits — a game item is an object, which maps to Sui's data model more naturally than to a token standard.

Sui versus Solana

Both chose one fast chain over rollups, and both parallelise. Solana discovers conflicts at runtime; Sui has transactions declare their objects up front, which is cleaner in theory and demands more of the developer. Solana has four more years of production hardening, far deeper liquidity, an independent second client in Firedancer, and its outages are now history rather than current events. Sui has the more elegant model and the operational record of a younger chain — which is precisely what 2026 demonstrated.

10 billion SUI is the maximum. A substantial portion was allocated to early contributors, investors and the Mysten Labs treasury at launch in May 2023, on vesting schedules that have released steadily since — the standard venture-backed structure, and the main supply consideration through the token's early years.

Staking rewards come from a subsidy fund that pays out on a declining schedule, plus transaction fees. The storage fund is the more interesting mechanism: storage costs are paid up front and held to compensate future validators for holding that data, with a partial refund when objects are deleted. Charging for state rather than only for computation is a better answer to state bloat than most chains have, and it is one of Sui's genuinely good ideas.

A portion of fees is burned, though at current volumes the effect on supply is small against unlock flow.

Staking SUI means delegating to a validator. Rewards are distributed at epoch boundaries — 24 hours — and the current rate is about 1.49% APY, which is low by the standards of comparable proof-of-stake chains and reflects both the subsidy schedule and how much of the supply is already staked.

There is no slashing implemented. A poor validator costs you rewards rather than principal. Unstaking takes effect at the next epoch boundary, so the wait is roughly a day rather than the weeks that Polkadot or Avalanche require.

Worth stating plainly: at 1.49%, staking SUI is not a yield strategy. It is a way to avoid being diluted by the holders who do stake, and it should be weighed as that rather than compared against Solana's 6% as though they are the same product.

Reliability is the live risk, not a historical one

Three halts in 2026, including 6 hours 44 minutes with roughly $1bn frozen, is a current operational record rather than growing pains from launch year. A chain that stops cannot be liquidated on, cannot be arbitraged and cannot be exited — every DeFi position on it is exposed to the outage rather than merely inconvenienced by it. Sui runs a single client implementation, which is the condition that turns one release bug into a network halt.

Validator concentration and hardware

High throughput demands capable validators, which means fewer of them and a professional operator base. The same trade Solana made, without Solana's years of adversarial testing behind it.

Unlock pressure

Large allocations to early investors and contributors continue to vest. That is a mechanical supply flow independent of anything the network achieves.

Developer scarcity

Sui's Move dialect is specific to Sui — not the same as Aptos's Move, and nothing like Solidity. The object model is powerful and is a genuine relearning cost, which limits how quickly the ecosystem can grow regardless of funding.

Low staking yield

1.49% is thin compensation for holding a high-volatility asset through outages. It does not change the risk profile and should not be counted on to offset it.