Solana has replaced the consensus system it had run since its launch six years ago, deploying Alpenglow on mainnet in what amounts to the network's largest architectural change since that original design was first put in place.

The new system targets transaction finality of approximately 150 milliseconds, down from a previous ceiling of up to 12.8 seconds under the old protocol. Under Alpenglow, validators exchange votes off-chain and group them into certificates that confirm blocks within one or two communication rounds, rather than relying on the slower, multi-step finalisation process Solana had used since launch.

The upgrade also changes how validators pay to participate in consensus. Previously, validators paid a voting fee of approximately 2 SOL per day. Under Alpenglow, that becomes a fixed fee of 0.8 SOL per epoch, a period of roughly 24 hours. Solana estimates the change will cut validator costs by around 60%.

A rocky path through testnet

The rollout was not without friction. According to Galaxy Digital, the testnet phase experienced six restarts and one failed migration that required reverting the change before engineers could proceed. A subsequent testnet attempt completed successfully, and none of the incidents put real funds at risk, as they occurred outside the live mainnet environment.

A shrinking, more concentrated validator set

The upgrade arrives at a moment when Solana's validator base has been contracting. Bitwise reported that the network ended September with 659 active validators, about 30% fewer than a year earlier. At the same time, 69% of SOL supply remains staked, underscoring how central validation remains to the network's economics even as the number of participants running it has fallen.

Concentration among the largest operators has also increased. Data from Solana Compass shows the combined stake share held by the ten largest validators rose from 22.2% to 24.5% over the past year, a trend that will be worth watching as the lower, fixed fee structure under Alpenglow takes effect across the remaining validator set.