Solana Alpenglow consensus upgrade enters testing: endgame time compressed to 150 milliseconds

Solana R&D team Anza announced on May 11 that the “Alpenglow” consensus upgrade has officially entered the community validators’ testing phase, aiming to compress block finality time from the current 12.8 seconds to 100 to 150 milliseconds. According to CoinDesk, this is the largest-scale consensus-layer overhaul since Solana launched.

Alpenglow has two major components: Votor and Rotor

Alpenglow includes two core components that respectively replace Solana’s current Proof of History (PoH) and Tower BFT:

Votor (voting engine): uses a “dual-path finality” design. When 80% of validators are online, blocks reach finality in a single round of voting, at about 100 milliseconds; if only 60% of validators participate in the first round, it moves to a second round, at about 150 milliseconds. Both paths run in parallel, and whichever achieves consensus first is taken as finality.

Rotor (data distribution): replaces the existing Turbine protocol, switching to direct communication “validator-to-validator” to replace multi-layer data propagation. Anza likened it to “replacing a phone tree with a direct-dial phone call.”

After replacement, PoH is replaced by a “fixed 400-millisecond block time”; the 32-layer voting structure of Tower BFT is replaced by Votor’s 1 to 2 rounds of voting. Internal Anza tests show finality time is compressed by about 100 times.

Trade-off: fault tolerance reduced from 33% to 20%

Alpenglow is not an upgrade that is “uniformly better”—it makes trade-offs in fault tolerance. The original Tower BFT could tolerate 33% Byzantine-fault nodes, and under the premise of optimizing speed, Alpenglow reduces fault tolerance to 20%. In other words, if more than 20% of validators act maliciously or go offline, the network could stall.

Anza’s approach is to “prioritize practicality over absolute security.” For Solana, compressing finality time to under 150 milliseconds effectively brings on-chain transaction confirmation speed closer to the level of traditional stock exchanges and Web2 payment systems. This is especially valuable for use cases such as high-frequency trading, prediction markets, and real-time payments.

Mainnet schedule and next steps

The SIM D-0236 proposal obtained support from about 98% of validators to pass in September 2025. This test uses Agave master validators to execute; the next step will package it into an official Agave release version, then kick off full testnet-wide testing.

Solana co-founder Anatoly Yakovenko said at an event on May 5 that Alpenglow’s mainnet could go live in the “next quarter” at the earliest. When Anza launched this test, it set the target between “end of 2026 Q3 and early Q4.” Alpenglow development is led by Anza’s chief economist, Max Resnick.

Events to watch next include: whether the community testing phase exposes performance or security vulnerabilities, the timing of the testnet’s formal launch, whether validators meet the 80% participation threshold when mainnet starts, and whether the trade-off of reducing fault tolerance from 33% to 20% introduces new attack vectors in real operations.

This article “Solana Alpenglow consensus upgrade enters testing: finality time compressed to 150 milliseconds” first appeared on 鏈新聞 ABMedia.