- Solana is moving a consensus upgrade called Alpenglow onto its public testnet, where experiment with that have no monetary value.
- The design target is to cut finality, the point at which a transaction can no longer be reversed, from roughly 13 seconds to about 0.15 seconds.
- Nothing has launched on mainnet, the figure is a target rather than a measured result, and two of the network’s validator clients cannot yet take part.
Somewhere between pressing send and the money arriving, there is a moment when a crypto transaction stops being provisional and becomes permanent. That moment has a name, finality, and Solana is testing a change that would bring it forward by about 12.8 seconds.
Finality is the reason an exchange makes you wait before a deposit shows up as spendable, and the reason a merchant accepting crypto can’t hand over the goods the instant a payment appears. It’s the hidden delay inside almost every crypto payment you’ve ever made. Understanding it tells you far more about whether a is usable for everyday money than any figure about transactions per second.
What’s worth knowing here is also what the headlines skip: this is a test on a network where the tokens are worthless by design, and the number everyone is quoting is a goal rather than a result.
Why a transaction in a block isn’t finished
Start with what a blockchain actually is: a shared list of transactions that thousands of computers, called validators, agree on. Those computers are scattered around the world and none of them is in charge. Agreement has to be manufactured, and manufacturing it takes time.
When you send a payment, it gets bundled with others into a block and added to the list. Your wallet will usually show it straight away. But at that moment the network has only proposed that the block is correct. Validators still have to vote on it, and until enough of them have, there remains a theoretical possibility that the chain reorganises and your block gets dropped in favour of a competing version of history.
Finality is the point at which that possibility closes. After finality, undoing your transaction would require breaking the network’s core security assumptions rather than simply waiting for a different branch to win.
On Bitcoin, finality is probabilistic and never technically absolute, which is why exchanges ask for a number of confirmations before crediting a bitcoin deposit. Each confirmation is another block stacked on top, making a reversal progressively more absurd to attempt. On Solana, the design is different: there’s a defined moment when a block is considered finalised, and today that takes somewhere around 12.8 to 13 seconds.
What Alpenglow changes
Solana currently reaches agreement using a mechanism called TowerBFT, where validators record their votes on the chain itself, building up a long chain of votes that eventually confirms a block. Alpenglow replaces that with a called Votor, under which validators send their votes directly to one another and a block can be finalised after one or two rounds of voting rather than the extended vote history used today.
According to CoinDesk, citing the feature-gate tracker maintained by Anza (the software company behind Solana’s main validator client), that redesign is what produces the 150-millisecond target.
Validators joining the testnet migration have to be running Agave 4.3, the current version of that client. Anza recommended 4.3 to mainnet validators on 21 September after first rolling it out to operators representing roughly 10% and then 25% of staked SOL, a staged approach that limits the damage if something turns out to be broken. CoinDesk also reports that Alpenglow ran for more than four months on a separate, smaller network built specifically for it before this step.
What waiting for finality costs you in practice
You’ve felt this without being told what it was. Move SOL from one exchange to another and the funds sit in limbo for a while before they become tradeable. That gap isn’t the exchange being slow. It’s the exchange declining to treat your deposit as real money until the network says it can no longer be undone, because if it credited you early and the transaction vanished, the exchange would eat the loss.
The same logic applies to bridges, the services that move assets between blockchains, and to any merchant taking crypto at a till. Nobody who stands to lose money wants to release goods against a payment that might still evaporate.
Thirteen seconds is not a long time in the abstract. It is an eternity at a card terminal, and it’s long enough that most payment businesses build a waiting state into their product rather than pretending the money is there. Cut that to 150 milliseconds and the wait stops being something a user notices at all, which is the actual argument for the upgrade.
The distance between a testnet and your money
A testnet is a full working copy of the blockchain running on tokens that are deliberately worthless. Validators can crash it, misconfigure it and break it without anyone losing a penny. That’s its purpose, and it’s also its limitation: a network nobody is trying to attack and nobody is depending on behaves differently from one securing billions of pounds.
Three gaps sit between this test and a change that affects a deposit you’re waiting on.
First, it may not have started. At the time CoinDesk published, Anza’s own tracker still listed the testnet switch as pending. Second, the 150-millisecond figure is a design target measured in controlled conditions, not a result observed on a live network carrying real value. Solana has a history of congestion and outages under heavy load, and a consensus mechanism’s honest performance is the one it manages on its worst day.
Third, Firedancer and Frankendancer, the alternative validator clients built by Jump Crypto, don’t yet support the Alpenglow test. Running more than one client is how a blockchain avoids a single software bug taking down the whole network, so their absence means the initial results tell you nothing about how the upgrade behaves across the full validator ecosystem.
There’s also a date circulating. 28 September appears in Anza’s schedule, but as a tentative point for activating Agave 4.3 features on mainnet generally, not as a confirmed Alpenglow launch. They are different things, and expect to see them reported as one.
The claim to hold lightly
Coverage of this has framed the upgrade as something that would let exchanges, bridges and merchants treat payments as irreversible almost immediately. That’s a reasonable description of the intent. It is not a description of anything currently happening: no exchange has announced shorter deposit waits, and none is likely to until the change has run on mainnet long enough to be trusted.
Faster finality also doesn’t touch the reasons your money most often gets stuck. None of the usual hold-ups are consensus problems: a withdrawal held for compliance review, a bank blocking a card payment, an exchange’s own queue. Our guides on delayed withdrawals and UK banks and crypto cover the friction that actually delays UK holders, and it’s almost never the blockchain.
What to watch
Whether the testnet migration actually activates, and what the measured finality looks like when it does rather than what the design promises. Whether Firedancer adds support, because an upgrade that only runs on one client isn’t ready regardless of how fast it is. And whether any exchange, once this reaches mainnet, publicly shortens its SOL deposit confirmation time. That last one is the only signal that the change has reached a reader’s account rather than a specification document.
Next time a headline gives you a precise millisecond figure for a blockchain, the useful question is where it was measured. Lab, testnet and mainnet under stress are three different answers, and only the third one is about your money.