Solana Launches Alpenglow Upgrade to Reduce Confirmation Time to 150 ms
The Solana network will launch the Alpenglow upgrade on October 21, replacing the TowerBFT validator voting mechanism used since 2020, aiming to shorten the transaction final confirmation time from about 6 seconds to 150 milliseconds.
The core of the upgrade is to move validator voting off-chain and aggregate votes into certificates, allowing transactions to achieve final confirmation in one to two rounds of consensus. If the mainnet performs as expected, the confirmation experience will be close to that of credit card payments or centralized trading platforms, directly improving meme coin trading, stablecoin payments, and on-chain trading platform usage.
Currently, validators pay about 2 SOL in voting fees daily. After the upgrade, voting fees will be eliminated, replaced by a fixed access fee of 0.8 SOL per epoch, approximately every 24 hours, with expected related costs decreasing by about 60%. This mechanism has been validated in testing and will not change user transaction formats or fee structures.
During testing, the network was cleared and restarted from scratch 6 times, with one migration attempt causing the entire test network to fail and roll back, while subsequent migrations were ultimately successful. These issues occurred on the test network and did not involve real funds, paving the way for this mainnet switch.
The upgrade will start at the beginning of epoch 1064, transitioning over 5000 slots and completing the switch at slot 459653000. Validators will need to run a client based on Agave 4.3, while some other clients do not yet support the migration phase.
Faster confirmations will attract capital inflows from high-frequency trading and payment scenarios into the Solana ecosystem. After the reduction in validator costs, some small and medium nodes may increase participation, while nodes that fail to upgrade in time may temporarily exit consensus and incur loss of earnings.
Source: Public Information
ABAB AI Insight
Since its mainnet launch in 2020, Solana has relied on the TowerBFT voting mechanism and approved Alpenglow in 2025 with a 98.27% support rate through the SIMD-0326 proposal. The Anza team is responsible for developing the Agave client, which has undergone 18 months of preparation, demonstrating migration multiple times in the test cluster, and completed the switch from testnet to development network in September 2026.
Funding and resources are primarily directed towards optimizing the Agave client, implementing BLS aggregate certificates, and the Validator Access Ticket (VAT) mechanism. Anza has changed the validator voting fee from an on-chain consumption of about 2 SOL daily to a fixed fee of 0.8 SOL per epoch, lowering the operational threshold for nodes while stripping consensus from the burden of on-chain transactions, freeing up bandwidth for high-frequency trading and payment scenarios.
This is similar to Ethereum's consensus layer reconstruction during the transition from PoW to PoS, but Solana chose to switch all at once on a running network rather than forking. Currently, Solana is in a performance control phase after rapid expansion, trying to bring final confirmation speed closer to traditional payment systems to solidify its share in meme coin and stablecoin settlements.
Essentially, this is a technological substitution: by moving voting off-chain and introducing certificate aggregation, Solana compresses consensus delays from seconds to milliseconds, changing the structure of on-chain transactions where "confirmation still requires waiting for finality." Mechanically, this reduces the cost of validators continuously burning SOL while increasing the network's capacity for high-frequency applications, but in the short term, some nodes may experience reduced earnings due to client compatibility and migration window issues.
ABAB News · Cognitive Laws
Consensus speed determines transaction experience; the shorter the delay, the more concentrated the funds.
Node costs decrease, but the network participation threshold may actually rise.
Upgrade failures are free on the testnet; mainnet failures cost the entire network.