Solana Alpenglow Goes Live, Cuts Finality to 150ms

Alpenglow replaces TowerBFT on Solana mainnet

Anza’s rollout page now labels the mainnet-beta migration completed. The Solana Foundation’s upgrade page also lists the mainnet feature gate as active. These are different checkpoints: enabling the feature starts the transition, while the consensus handoff follows later.

Daily Crypto Briefs additionally queried the public mainnet RPC during reporting. Its getAgGenesisCert method returned a non-null certificate identifying slot 459,652,999 as the preserved genesis block. A separate epoch query returned epoch 1064 and a current slot beyond the announced boundary.

The Foundation’s verification instructions explain that the certificate is absent while TowerBFT remains in control. The response corroborates migration; it is not an independent measurement of the 150ms speed claim. The preserved block immediately precedes the announced takeover boundary, so the two slot numbers describe different parts of the handoff.

Anza’s migration design first isolates the boundary by temporarily admitting only validator votes into subsequent blocks. Validators agree on the retained block, form a signed certificate, discard the temporary tail and resume user transactions under Votor, Alpenglow’s voting mechanism.

That sequence explains how the network can change consensus without replacing users’ balances or starting a separate token. The launch announcement says the process followed public testing, security exercises and repeated migration rehearsals. It does not provide a complete independent production-latency distribution.

Solana’s 150ms finality has infrastructure limits

Anza’s original research announcement described a median near 150ms based on simulations using mainnet’s then-current stake distribution, excluding computation overhead. Today’s launch moves the protocol into production, but the original assumptions remain relevant when interpreting the headline number.

Votor runs two voting paths concurrently. One can finalize a block in one round when 80% of stake responds; another uses two rounds with 60%. Whichever path finishes first determines the result. The design also distinguishes up to 20% adversarial stake from an additional 20% offline stake, rather than treating all failures as interchangeable.

Anza’s commitment documentation says confirmed and finalized are equivalent under Alpenglow: both describe a transaction committed to the permanent ledger. Under TowerBFT, stronger finality required additional confirmed blocks on top of the transaction’s block.

However, consensus participants and an application do not necessarily receive the result simultaneously. The operator documentation says nodes outside the admitted validator set normally learn finality from certificates included in later blocks. That can add a produced-block interval, or longer when blocks are delayed or skipped.

Operators can arrange direct delivery from admitted validators to reduce that delay. If those partners become unreachable, the node falls back to receiving certificates through blocks. Delivery architecture therefore remains part of the customer experience even after faster consensus is live.

For a wallet, exchange or checkout service, a network-level finality figure alone cannot establish its displayed confirmation time or deposit-credit policy. The useful comparison is the full path from sending a transaction to the service recognizing settlement, measured consistently before and after migration.

Validator fees and Solana dashboards change

The vote-account guide specifies a recurring 0.8 SOL Validator Admission Ticket, burned from each admitted validator’s vote account. That amount must sit above the rent-exempt minimum. The first charge at epoch 1064 pays for participation in epoch 1065.

An insufficient balance prevents voting and block production in the following epoch. Operators can direct commission revenue into the vote account, but configuration changes take time and revenue is not guaranteed to cover the charge. Funding remains an ongoing obligation after the launch.

The Foundation says ordinary users need no migration action, while execution, transaction formats and fee mechanics remain unchanged. Its infrastructure guidance also warns that validator votes leave blocks. A transaction dashboard that included those votes can show a sharp numerical decline without demonstrating a fall in customer activity.

That creates a measurement break: raw totals spanning the switch need consistent definitions. Non-vote transactions and comparable service-level latency observations will be more useful than an unexplained before-and-after total. Faster consensus also does not by itself resolve the validator concentration questions identified before deployment.

Alternative.me’s Fear & Greed Index read 63 on October 21, up from 55 a day earlier. Both readings were in Greed territory; the index measures broader market sentiment rather than upgrade performance.

Source: Alternative.me, checked October 21.

Anza lists Rotor, geographically aware leader rotations and the post-quantum Quantumglow research design as later work without committed release dates. The next evidence to watch is sustained mainnet latency, validator participation and application behavior. This report corroborated the migration but did not independently benchmark those outcomes.