An epoch is a fixed period of time used as a common checkpoint. In blockchains, it groups network activity into predictable cycles. In machine learning, it means one full pass through the training data.
In many blockchain networks, an epoch is a scheduled window during which key operations happen in an organized way. Networks use it to coordinate things like when blocks are produced, when rewards are paid out, or when protocol maintenance is rolled out. Think of it as the network’s calendar that keeps time for recurring tasks.
Within an epoch, a chain can assign who gets to create blocks, settle how and when incentives are distributed, and trigger updates that need a consistent reference point for all participants. This structure helps validators and developers plan actions around a shared timetable.
How an epoch works depends on the consensus design. On proof-of-work chains, the cadence ties back to how long it takes miners to find valid blocks. On proof-of-stake chains, epochs often define the period during which validators are selected to propose and attest to blocks. Some designs also divide time into smaller “slots,” with a designated slot leader responsible for producing the block in that slot.
Epoch length varies by network. It is usually defined by a certain number of blocks or slots rather than by wall-clock time. For example, on Ethereum, an epoch spans 32 slots of 12 seconds each, which works out to about 6.4 minutes. On Cardano, an epoch contains 432,000 slots of 20 seconds and typically lasts around five days.
Staking systems often align reward cycles with epochs. This makes it clear how long coins need to be staked to take part in a given cycle and when payouts are calculated. The epoch boundary is also a convenient place for the protocol to refresh validator assignments.