SMS
Solana’s new consensus protocol, Alpenglow, has gone live on mainnet beta, replacing TowerBFT. The core development team Anza stated that this switch occurred at slot 459,653,000.
Transactions are now irreversible in about 150 milliseconds, down from the previous 12.8 seconds. Anza believes this is significant for exchanges, banks, and merchants.
The largest consensus upgrade since Solana's network launched in 2020 went live on Wednesday, reducing the time required for a transaction to become irreversible from approximately 12.8 seconds to about 150 milliseconds, according to Anza.
Validators retired Solana's TowerBFT consensus protocol, which has been in use since March 2020, at slot 459,653,000 around UTC 14:00. They switched the system live on the running network while preserving the existing ledger, with coordination of a common handoff point by independently operating parties.
This change targets finality—the point at which transactions can no longer be reversed and when exchanges, banks, or merchants wait before confirming that assets have been transferred. Under TowerBFT, Solana provides conditional confirmation within seconds, but full finality takes approximately 12.8 seconds. Alpenglow combines these two steps into one.
“This is the largest and most ambitious blockchain upgrade since ‘The Merge,’ and Anza turned it from a whitepaper into a live system on mainnet in less than 18 months,” said Anza CEO Brennan Watt, referring to Ethereum’s switch to proof-of-stake in 2022.
Before launching on mainnet, Alpenglow underwent public testing, including red teaming, penetration testing, repeated migration drills, and a security competition, followed by deployment on Solana’s testnet and devnet. Anza integrated the protocol into Agave, its maintained validator software, and activation was completed by independent validators.
Anza’s research team, led by Professor Roger Wattenhofer from ETH Zurich, includes Quentin Kniep and Kobi Sliwinski, who released the Alpenglow design specification in May 2025.
Under Alpenglow, validators no longer write votes into subsequent blocks; instead, they send votes directly to each other, with vote weights calculated based on staked amounts. Wattenhofer said in a recent interview last month with Jacob Creech, Vice President of Technology at the Solana Foundation: “If 80% of the votes say yes, it’s immediately finalized. You’ll know right away that the block has been settled.”
He explained that the second path involves two rounds of voting, each requiring 60% stake support, and for validator clusters in closer geographic regions, such as Europe, the process could be even faster. The protocol is designed to continue operating even if less than 20% of the stake is malicious and another 20% is offline. If block finality is not achieved within 10 seconds, validators gradually increase block times until messages can be delivered. “Eventually, even a poor internet connection will get us another block,” Wattenhofer said.
More upgrades are still in the works. Rotor will change how block data propagates across the network, while geography-aware leader rotation aims to reduce latency when block production switches between validators. Quantumglow is a research design planned to replace Alpenglow’s elliptic curve cryptography with hash-based signatures, enabling resistance to quantum computers while maintaining the same finality speed. However, none of these upgrades have confirmed dates.
Alpenglow's Path to Mainnet
Validators approved the upgrade in September 2025, passing the SIMD-0236 proposal to implement the scheme with 98% support. At the time, Anza's Chief Economist Max Resnick told Decrypt: "The most immediate change users will feel will be confirmation delays reduced to approximately 150–200 milliseconds."
Alpenglow launched on Solana's community testnet in May 2026, at which time Resnick stated the team aimed to activate the mainnet by the end of Q3 or early Q4. On October 13, Anza confirmed the date of October 21.
This upgrade also paves the way for other changes. Developers from Jump Crypto’s Firedancer team have proposed removing Solana’s current limit of 60 million compute units per block after Alpenglow launches, allowing block sizes to expand to the maximum speed that the fastest validators can handle, while slower validators can use their “skip vote” feature to abstain. Critics have warned that this change could favor better-funded operators and exacerbate centralization.