HyperBFT Explained
How Hyperliquid's consensus system orders transactions and secures HyperCore and HyperEVM.
HyperBFT is Hyperliquid's proof of stake consensus system. It is based on the HotStuff family of Byzantine fault tolerant protocols and determines the transaction order that HyperCore and HyperEVM execute.
For a financial network, ordering is not abstract. If two traders compete for the same resting liquidity, the agreed transaction order helps determine which action reaches that liquidity first. Consensus therefore sits directly underneath market execution.
What consensus must guarantee
Validators need to agree on one history even when messages arrive at different times or some validators fail. Once a round commits, honest nodes should execute the same ordered transactions and arrive at the same state.
That state includes more than token balances. It can include order books, perpetual positions, vault state, staking balances, and EVM contract state. Inconsistent ordering would therefore create inconsistent financial outcomes.
Stake determines consensus weight
Hyperliquid uses delegated proof of stake. HYPE holders can delegate to validators, and stake contributes to the validator's consensus weight and reward share.
The safety model depends on a supermajority of stake following the protocol. Hyperliquid describes a quorum as more than two thirds of stake. Validator economics, delegation concentration, uptime, and operational quality therefore matter to network security.
The current documentation also separates jailing from slashing. Validators can be jailed for failing participation or responsiveness expectations, which stops them from earning while inactive. Automatic slashing is not currently implemented, so assumptions imported from other proof of stake networks may not apply.
Consensus latency and trading latency are different
Hyperliquid reports roughly 0.2 second median end to end latency and 0.9 second p99 latency for a geographically colocated client, alongside mainnet capacity of about 200,000 orders per second. Those figures describe the protocol under stated conditions.
A trader's experienced latency also includes network routing, API infrastructure, local signing, application queues, and client processing. A protocol benchmark is therefore not the same as the latency of a specific trading stack.
WebSocket and real time data systems add another layer because market data delivery and order submission can have different paths.
Why HyperBFT matters to market structure
A centralized exchange can maintain an internal sequencing engine that users must trust. Hyperliquid moves transaction ordering into a validator consensus process. That does not eliminate every form of latency advantage or infrastructure asymmetry, but it changes who determines canonical ordering and where that decision is secured.
For builders and institutions, the relevant questions extend beyond transactions per second. Validator concentration, consensus behavior, finality, networking, and the relationship between committed rounds and execution all belong in infrastructure due diligence.
Continue reading
Read What Happens When You Place a Trade on Hyperliquid? to see where consensus fits into execution, then HYPE Staking and Validators Explained for the staking side of validator participation.