Mastodon Skip to content
LIVE - NYSE/-/- CRYPTO/OPEN/24/7
BTC$84,169▼ 2.55%ETH$2,688▼ 2.40%SOL$115.09▼ 3.00%TOTAL CRYPTO$2.87T▼ 5.32%S&P 5007,706.03▲ 0.41%NASDAQ26,936.04▲ 2.89%DOW51,511.59▼ 3.31%GOLD4,316.20▼ 8.12%WTI92.18▲ 8.43%BRENT98.24▲ 6.59%EUR/USD1.1391▼ 2.49%USD/JPY158.38▼ 0.33%DXY101.10▲ 2.12%
Crypto

Solana Starts Testing Alpenglow, Chasing 150ms Finality

Solana moved its Alpenglow consensus upgrade to public testnet. The target: cutting transaction finality from 12.8 seconds to 150 milliseconds.

Pexels – Willian Justen de Vasconcellos

Solana has moved its biggest consensus overhaul in years onto a public testnet, where developers will try to cut the time a transaction takes to become irreversible from about 12.8 seconds to 150 milliseconds. The upgrade, called Alpenglow, replaces the network’s TowerBFT system with a new component named Votor, which lets validators finalize blocks after one or two rounds of direct voting.

The change matters for anyone building on Solana. Exchanges, bridges and merchants today wait roughly 13 seconds before a payment counts as final. If Alpenglow works as designed, that wait shrinks to a fraction of a second, putting Solana’s settlement speed closer to card networks than to existing blockchains. Solana’s current setup already offers fast pre-confirmations, around 400 milliseconds, but true finality has stayed slow, and finality is what payment processors and trading venues actually care about. A pre-confirmation can be rolled back; finality cannot.

How Votor changes the vote

Under the old system, validators attach their votes to transactions on the chain itself, which adds load and delays. Alpenglow drops vote transactions entirely. Votes move directly between validators, and observed votes and certificates are kept in a separate data structure called Pool, which feeds Votor and is stored alongside the network’s block data. Voting runs in two rounds, and a block is final once enough stake signs off in either.

The design also assumes a specific threat model: the network should stay safe even if 20 percent of stake acts adversarially and another 20 percent goes offline, the so-called 20+20 model the foundation cites in its documentation. That is a stronger resilience assumption than TowerBFT offered in its simplest form, and it reflects lessons from past Solana outages where stalled consensus took the network down for hours.

The Solana Foundation describes the change as the first major consensus upgrade since TowerBFT. It also changes how the network thinks about rewards, blockspace and voting itself. Once the migration finishes, Solana will be a different protocol under the hood, even if users see nothing change.

That last part is deliberate. User-facing behavior stays the same: wallets and infrastructure handle the swap automatically, and people sending funds do nothing. The initial testnet migration requires the Agave 4.3 client. One caveat for operators: the first phase does not support Firedancer or Frankendancer, the independent validator clients many large operators run, so those validators will sit out the first testnet round.

What changes and what does not

Metric TowerBFT today Alpenglow target
Finality about 12.8 seconds about 150 milliseconds
Pre-confirmation about 400 milliseconds unchanged
Voting on-chain vote transactions direct validator messages via Pool
Block propagation Turbine tree Turbine now, Rotor later

A mainnet update is penciled in for September 28, but the foundation is careful to call it tentative. It is a scheduled network upgrade, not a confirmed Alpenglow launch date. A later phase, Rotor, will replace the Turbine block propagation system with a single relay layer designed to cut network latency further. Rotor keeps Turbine’s erasure-coding idea but changes the delivery plan from a tree of nodes to one relay layer, trimming the hops a block takes between validators.

Strong timing for the ecosystem

The engineering milestone lands during a strong stretch for Solana’s ecosystem. Spot Solana ETFs in the US have now recorded ten consecutive weeks of inflows, and the network saw 348 million dollars in real-world asset inflows over the past day, lifting total value locked to 4.23 billion dollars, according to CoinGecko data. On the trading side, Raydium said xStocks, its tokenized equities venue, passed 1.3 billion dollars in cumulative volume within 23 days of launch.

Faster finality feeds directly into those numbers. Tokenized stock trading and perp venues both benefit when trades settle almost instantly, because risk windows shrink and capital recycles faster. Institutional desks that tested Solana and balked at 13-second finality have a concrete reason to retest once mainnet activation is real. Market makers in particular price the risk that a fill gets reversed, and that risk scales with finality time.

The market reaction has been muted so far. SOL traded near 104 dollars, down about 4 percent over 24 hours, in line with a broader pullback across major tokens as bitcoin slipped back under 80,000. Traders appear to be treating Alpenglow as an engineering milestone rather than a price catalyst, at least until mainnet activation gets a firm date.

Still, the stakes are real. Solana’s pitch to institutions has always been speed, and rivals have been closing the gap. Newer chains advertise sub-second finality, and payment-focused networks market hard on settlement guarantees. If Votor delivers 150-millisecond finality on mainnet, Solana regains a clear technical edge in payments and trading use cases where finality time is the product.

The risk is operational. Consensus rewrites are the hardest kind of upgrade in distributed systems, and public testnet is where validator operators start breaking things. Client diversity adds another wrinkle: with Firedancer excluded from phase one, most of the network will run Agave through the transition, concentrating risk in a single client until the second client catches up. If the testnet surfaces problems, the September 28 date slips and the network waits. Developers have run Alpenglow through months of internal testing before this point, and the foundation has given no backup date if the tentative slot is missed.

There is also a fee question nobody has fully answered. Dropping vote transactions removes a large share of network traffic, which changes validator economics in ways that are still being modeled. Validators earn rewards partly through voting, and Alpenglow rewrites how those rewards are computed. The foundation says the new design makes rewards simpler, but validators will want to see the numbers before they vote on mainnet activation.

SourcesCoinDesk; Solana Foundation; KuCoin market report; CoinGecko
Share: X