Solana has switched on its Alpenglow consensus overhaul on both public test networks, the biggest protocol change in the network history. The upgrade targets transaction finality in 100 to 150 milliseconds, down from roughly 12.8 seconds today, though no mainnet date has been set.
Anza, the company building Solana core software, activated Alpenglow on public testnet on September 24 after a 17-minute countdown. Devnet followed on September 25. The Solana Foundation upgrade page now lists the new consensus system as active on both networks, giving application teams two staging environments before any move to the live blockchain.
The final TowerBFT block on testnet carried 82 percent of stake before an Alpenglow genesis block took over. Devnet completed its own transition at slot 504,148,999. Solana co-founder Anatoly Yakovenko marked the devnet switch with a short post: You are not ready for this.
How the new system works
Alpenglow replaces the TowerBFT consensus algorithm with Votor, a voting system built with researchers from ETH Zurich and approved by validators under proposal SIMD-0326. The core change is that validator votes move offchain. Under the current system, validators submit votes as onchain transactions and wait through a 32-slot confirmation chain. Those voting transactions account for roughly 75 percent of network transaction counts.
Votor collects peer-to-peer votes into compact BLS certificates instead. Two confirmation paths run at the same time. If validators holding 80 percent of stake agree in one round, finality lands around 100 milliseconds. If only 60 percent agree initially, a second round brings finality in about 150 milliseconds.
The security model changes with it. The new design tolerates 20 percent adversarial stake plus another 20 percent offline, a different tradeoff from the old system. Rotor, a planned replacement for the Turbine block propagation protocol, comes later. The Solana Virtual Machine, transaction formats, programs and fees stay unchanged, so wallet users need to migrate nothing.
Early numbers, big caveats
Solana Foundation engineer Ilan Gitter cited more than 25,000 transactions per second with 150-millisecond finality on the testnet cluster after activation. That figure comes from a test cluster, not a full public validator set handling live market activity. The 100 to 150 millisecond target itself is derived from simulations, and sustained results under real conditions remain unproven.
Client diversity is another open issue. Firedancer and Frankendancer, the independent validator clients, have not implemented Votor yet. The first public transitions depend entirely on Agave 4.3, which concentrates risk in one implementation. Anza wants developers testing integrations throughout an observation period before any mainnet-beta migration.
Anza software schedule tentatively allows feature activations to resume on September 28, but that date is not an Alpenglow launch date. No firm activation timeline for mainnet has been announced.
Even if the protocol settles in 100 milliseconds, exchanges and custodians may still process deposits on much longer confirmation schedules.
What changes for users and data providers
For most retail users, the difference between one second and 150 milliseconds is hard to feel. The practical gains show up elsewhere. An exchange could release a deposited balance almost immediately after confirmation. A payment app could tell a merchant a sale is final in a fraction of a second. Bridges and settlement layers that currently wait out reorganization risk get a shorter window to cover.
The change also reshapes network statistics. Removing validator votes from blocks will make reported transaction counts drop sharply even if actual user activity stays flat. The Foundation has told data providers to adjust their comparisons, and services that build transaction histories must keep competing candidate blocks separate until the network certifies one. Mixing their contents could produce incorrect records.
Developers using Geyser and gRPC feeds get a new bank_id identifier to handle the candidate-block period. Applications that only send transactions and read account balances need no changes at all, according to the Foundation migration guide.
Why the stakes are high
A consensus swap is the most invasive upgrade a proof-of-stake network can attempt. It touches the fundamental guarantees that every validator, exchange and bridge relies on. Solana has been running TowerBFT since launch, and the network has built its reputation on throughput and low fees. Finality measured in seconds has been the remaining gap against traditional payment rails.
If Alpenglow reaches mainnet and performs as designed, Solana would offer irreversible settlement on the same timescale as card networks, a claim no other major blockchain can make today. The competitive angle is real: other chains are pursuing similar latency targets, and institutional payment pilots have made settlement speed a selling point in recent quarters.
The risks are equally concrete. A consensus bug on mainnet would be far more damaging than a performance shortfall. That is why the two-network staging period matters, and why Anza is in no hurry to announce a date. The more time Alpenglow spends running on public testnets without incident, the stronger the case for the mainnet switch.
Exchanges face their own adjustment. Deposit crediting systems are typically built around confirmation counts measured in slots, not milliseconds. Even with fast protocol finality, custodians and bridges may keep conservative buffers until they trust the new guarantees, which would blunt the user-facing benefit at first.
SOL traded near $120 on the news, up roughly 3 percent on the day. The token has been in a recovery alongside the broader market, with stablecoin supply on the network hitting a record $17.3 billion last week and US spot SOL ETFs posting their second-best week of inflows.