Transactions per second measures a network's sustained throughput: how many transfers or smart-contract calls it can confirm and finalize each second under real-world load, not just in a short burst. It is one of the main yardsticks used to compare blockchains, since it directly affects confirmation times and fees during busy periods.
TPS is set by two design choices working together: how much data fits in each block, and how fast blocks are produced. Bitcoin's roughly 1MB blocks and ten-minute average block time cap its base layer at around three to seven TPS, a ceiling that has barely moved since launch. Ethereum's base layer typically handles somewhere in the low double digits, bounded less by a fixed transaction count and more by a per-block gas limit that varies with how complex each transaction is. Newer proof-of-stake chains such as Solana are engineered for far higher throughput, often thousands of TPS, by using faster block times and a smaller, more powerful validator set.
For context, card networks like Visa advertise capacity in the tens of thousands of TPS, a gap regularly cited when arguing that public blockchains still need scaling before they can handle mainstream payment volume.
Pushing TPS higher at the base layer usually trades away some decentralization or security, since fewer, more powerful nodes are needed to keep up. Most networks scale around this instead, using Layer 2 rollups, sidechains, or payment channels that batch transactions off-chain before settling the result back on the main chain.