Solana is replacing its consensus today. The switch from TowerBFT to Alpenglow is set for October 21, at around 14:50 UTC according to the developer house Anza, which is around 16:50 German t
Solana is replacing its consensus today. The switch from TowerBFT to Alpenglow is set for October 21, at around 14:50 UTC according to the developer house Anza, which is around 16:50 German time. The Solana Foundation most recently projected the corresponding epoch start at about 15:00 UTC; an earlier estimate named 13:20 UTC. Expect a window of roughly two hours, then, not a button press to the minute. What changes, by contrast, can be stated precisely: the time to irreversibility of a payment falls from around 12.8 seconds to a target of 100 to 150 milliseconds.
The price has been unmoved by it so far. SOL traded at $106.54 on Wednesday morning, 2.3 percent above the previous day, after a daily low of $103.72 and a high of $107.93 (data from CoinGecko, as of October 21). Where the price can point over a horizon of months is set out in our Solana price prediction. For today something else counts: for investors, a protocol switch is first a risk date and only then an argument.
TowerBFT is the procedure by which Solana's validators have until now agreed on which block holds. Each of those votes ran across the blockchain as a transaction of its own. Alpenglow takes the votes off the blockchain: validators will sign them as BLS messages, which can be bundled into a single certificate. A BLS signature scheme is one in which many individual signatures can be combined mathematically into a single one that can be verified in one step.
The changeover does not run over days. According to Unchained's report on Anza's schedule, a feature gate with the identifier SIMD-0384 flips at an epoch boundary. A counter then runs over 5,000 slots in which the blocks contain validator votes and nothing else. That sounds like a standstill and is a safeguard: because no user transactions sit in the blocks during that window, the section can be rolled back without anyone losing a payment. At the end of the counter the validators certify an Alpenglow genesis block carried by 82 percent of staked capital. Anza expects that final step to take hundreds of milliseconds. TowerBFT is then retired.
The new voting component is called Votor, and it knows two routes to completion. On the fast route, 80 percent of staked capital votes for a block and the process is finished after one round. Where that is not enough, a fallback route with a threshold of 60 percent kicks in, which needs two rounds. Both figures appear on the Solana Foundation's Alpenglow overview page. The fault tolerance is there too: Alpenglow tolerates 20 percent malicious capital plus a further 20 percent offline. By the same source, TowerBFT tolerates less than 33 percent hostile capital.
Finality is the point from which an entry can no longer be reversed. At an exchange, that is the time a deposit needs before it counts as credited. If it falls from just under 13 seconds to a tenth of a second, a waiting period disappears that card payments do not have. Whether the 150 milliseconds actually hold under load is a target, though, and not yet a measurement. The target also applies in the first instance to admitted validators, not to every route by which a wallet obtains its information.
The validators voted on Alpenglow in September 2025. Approval came from 98.27 percent of the votes cast, with 52 percent of all staked SOL participating. The second figure is the more interesting one: almost half the capital did not vote at all.
Testing then ran for months. Anza cites a bug bounty of more than 25,000 SOL, 115 participating security researchers, a simulation across 1,500 nodes and nine full-scale migrations. On testnet Alpenglow took over consensus on September 24, on devnet a day later. The testnet was restarted from scratch several times from the beginning of May, and one migration attempt failed network-wide and was rolled back before a later one succeeded. A network-wide failed attempt in testing is therefore part of the backstory to this date.
For validators the cost side changes fundamentally. Until now they paid a transaction fee for every single vote, together around 2 SOL a day. That amount had doubled after slot times were shortened to about 200 milliseconds. Alpenglow replaces the individual fees with a flat admission fee of 0.8 SOL per epoch. An epoch lasts about 24 hours. The fee is paid from the vote account one epoch in advance, and it is burned.
Work it through yourself: 2 SOL against 0.8 SOL a day is a saving of 1.2 SOL per validator, or around $128 a day at Wednesday morning's price. The 60 percent cost reduction being cited is therefore documented. The second part of the calculation is mentioned less often and matters more for the token supply: the old vote fees mostly passed on to other validators, whereas the new admission fee accrues to nobody. The amount leaves circulation. At roughly 1,000 active vote accounts that would be 800 SOL a day withdrawn from supply. That extrapolation is our own and depends on the number of validators that actually remain admitted after the switch.
The new admission fee of 0.8 SOL per epoch is burned; it accrues to no recipient.
Vote transactions historically took up around 75 percent of the room in Solana's blocks. Once they disappear from the blockchain, every statistic that simply counts transactions per second falls with them. You will see that in rankings over the coming days, and it does not mean a decline in economic activity. It means the count has until now included something no user triggered.
When reading metrics that means: from today, stop comparing Solana's TPS with the figure from two days ago. What remains reliable is the number of active addresses, trading volume on the decentralised exchanges and fee revenue. If you read a headline in the coming weeks about a collapse in Solana transactions, check whether the figure was gathered before or after October 21.
One detail that gets lost in the short reports: Firedancer and Frankendancer, the alternative validator clients, are not active during the handover window. For Alpenglow, validators have to register a public BLS key and run Agave version 4.3; version 4.3.2 is named for Agave, Jito-Solana and Harmonic Salsa. If you run your own node today and do not meet those requirements, you simply do not take part in the new consensus.
There is also a limitation that affects wallet users: some RPC providers are likely to report finality with small delays against the 150-millisecond mark. The clock on the blockchain keeps its accuracy of one second, and transaction timestamps remain approximations. The tenth of a second therefore describes the consensus and not automatically what your app displays.
While work went on at the protocol, money flowed out. The US spot funds on Solana saw net withdrawals of $6.72 million on October 19, the largest single item being the Bitwise Solana Staking ETF at $4.8816 million, with Fidelity's FSOL accounting for around $1.84 million (data from SoSoValue). How long the run now is depends on the analysis: one counts eleven consecutive trading days of net outflows, another describes the run from October 5 to 16 and so arrives at ten. The funds have lost more than $60 million in October.
More money has flowed out of the US spot funds on Solana in October than flowed in during September.
September ran the other way: inflows there came to around $272 million. Since launch, net inflows add up to about $1.55 billion, and the net assets of the spot funds stand at around $1.625 billion. October has therefore undone just under a quarter of September without seriously denting the stock. If you are considering a fund route yourself, it is worth looking at crypto ETFs in Germany and the cost structures there.
On the downside the daily low of $103.72 is the first line, with the round $100 mark below it, which the price has defended several times in October. On the upside the daily high of $107.93 comes first. Between them lies a level from our article yesterday on the treasury sale: $107.51, the line at which a listed treasury holder's sale was pinned yesterday. That line has held since, with the price running back from $103 to $106.54 and sitting just under a dollar below it.
That is the real difference from yesterday. Twenty-four hours ago the question was whether a single seller was pressing the market. Today the question is whether a technical date attracts buyers. Historically, upgrades to crypto networks have tended to be a reason to sell, because the date was known well in advance. For Alpenglow that can only be demonstrated in the days afterwards.
From this desk, Alpenglow is technically the biggest step in years and still secondary to this week's price question. Three figures from this article argue for that: ten to eleven trading days of outflows, more than $60 million out of the spot funds in October, and a price that at $106.54 sits around 64 percent below its all-time high of $293.31. Faster finality creates no buyer; it lowers friction for applications that are being built anyway.
One point argues against that, and we take it seriously: payment settlement with a tenth of a second of finality is an argument Solana can make against card networks and hardly any other blockchain can. If the 150 milliseconds hold in operation, that shifts the outlook for payment applications in 2027, not the outlook for the price in October 2026. Dissent comes from inside the project: Solana co-founder Anatoly Yakovenko wrote, "Alpenglow would have shipped in half the time if they just changed the vote program to the ag state machine", questioning the effort rather than the goal. A response from Anza is not available.
For holders in Germany the most important news is a non-event. A consensus switch is neither a sale nor a swap. Your SOL remain the same units, the holding period under section 23 of the Income Tax Act continues to run, and the day triggers no disposal. If you hold your SOL for more than a year, you still sell gains tax-free; within the year the €1,000 exemption threshold for private disposals applies. If your acquisition dates are not cleanly in order, crypto tax software helps you catch up.
With staking the case is different and has nothing to do with today: rewards from delegating are other income and must be recognised in the year they accrue. What changes for delegators is their validator's cost side, not their own tax liability. If its vote fees fall by 1.2 SOL a day, more yield stays in the pool. Whether the validator passes that on depends on its commission, which you will find in the comparisons of staking providers under review.
For users, no action is required according to the Solana Foundation. Signing, sending and approving remain unchanged; finality simply arrives earlier. That applies to you whether your SOL sit in your own wallet or are delegated to a validator. Three points are still worth a look.
First, trading venues routinely pause deposits and withdrawals during protocol switches, even when nothing goes wrong. If you are planning a transfer today between 1pm and 5pm German time, check your exchange's status page beforehand; an overview of providers is available at the crypto exchanges under review. Second, your wallet may show the new speed only with a delay, because the RPC provider sits behind it. A slow display in the first few hours is not yet evidence of a problem in the network. Third, version 4.3.2 concerns you only if you run a validator yourself.
You can put it to the test yourself this evening: send a small amount from your wallet to your exchange address and measure how long the credit takes. If it comes in well under the familiar twelve-plus seconds, the switch has reached your provider. If it comes in above that, the cause is the exchange's processing, no longer the consensus.
October 21 does not decide the price; it decides a claim. Three steps make sense today and tomorrow:
If the price holds $103.72 and runs above $107.93, the upgrade has given the market something. If it falls below $100, it was a technical date with no price effect, and the fund flows remain the story of the month.
(As of October 21, 2026. This article is not investment advice. Prices and fee structures change; check the terms with the provider before you buy.)
