Solana's Alpenglow Upgrade: What 150 ms Finality Means for Paying With SOL

Alpenglow is Solana's new consensus protocol, designed by Anza to cut the time a transaction takes to become irreversible from about 12.8 seconds to roughly 150 milliseconds, according to Anza and the Solana Foundation. It passed community governance as SIMD-0326 with 98.27% of votes in favor, per Solana Status, and mainnet feature activation is scheduled to begin on September 28, 2026 under Anza's Agave v4.3 release schedule.

If you hold SOL and use it to pay or top up a card, two Solana stories crossed your feed this summer: the network "nearly halted" on August 12, and an upgrade called Alpenglow is about to make it much faster. This guide covers what finality is, what Alpenglow changes, what happened on August 12, and what it means when you spend SOL.

Three clocks get mixed together whenever someone calls a blockchain "fast."

• Block time (a slot, on Solana) is how often a new block is produced. It says nothing about whether that block will stick.
• Confirmation is the network's early signal that a block is very likely to stick. On Solana, a transaction is "confirmed" once validators holding at least two-thirds of stake have voted for its block, a state called optimistic confirmation, per Helius.
• Finality is the point of no return. The block can no longer be replaced and the transaction cannot be undone.

For a payment app the distinction is the whole business. A merchant handing over a coffee needs a fast signal that money is very likely coming. An exchange or card program crediting a spendable balance wants finality, because crediting against a block that later gets replaced means paying out money that never arrived.

For most of Solana's history the slot time was about 400 ms, and finality under the current protocol, Tower BFT (Byzantine Fault Tolerant), took about 12.8 seconds from block creation, per Anza, because it requires roughly 32 further confirmed slots, per Helius. Slots are already shrinking: SIMD-0525 cut mainnet slot time to 350 ms on August 19, 2026 and 300 ms on August 25, per Solana's upgrade tracker. That trims the wait proportionally, but you still wait dozens of blocks, so finality remains on the order of ten seconds.

For throughput rather than latency, see our Solana transactions per second breakdown.

Anza, the Solana research and development firm spun out of Solana Labs, published Alpenglow on May 19, 2025, per Anza, retiring Tower BFT and Proof of History in favor of two new components.

Votor is the voting protocol: validators "send votes directly to each other instead of as transactions, aggregate them into certificates, and finalize a block in one or two rounds," per the Solana Foundation. If 80% of stake votes for a block it is finalized in a single round; if only 60% is responsive it finalizes after two rounds, per Anza. Timing uses local timeouts rather than the synchronized clock Proof of History provided, per SIMD-0326.

Rotor replaces Turbine, the layer that spreads block data across validators, "with a single relay layer to minimize network latency," per the Solana Foundation. It is not part of the current activation: SIMD-0326 covers only Votor, and the Foundation says Rotor is planned "for a later release once Votor is adopted, and is not yet scheduled."

The "20+20" resilience model is the part that matters for the August 12 story. Alpenglow is designed to keep operating with up to 20% adversarial stake plus an additional 20% non-responsive stake at the same time, per Anza. The SIMD is candid about the trade: the protocol "does not have the 33% byzantine security that can be achieved with two-round voting protocols," but "the one-round voting and the 40% crash failure resilience is worth the tradeoff," per SIMD-0326.

• Vote. SIMD-0326 passed on September 2, 2025 with 98.27% Yes, 1.05% No, 0.69% Abstain, and 52% of stake voting, per Solana Status and crypto.news.
• Code shipped, switch off. Agave 4.2 contains the Alpenglow code but does not activate it. The Foundation's 4.2 notes said activation "is expected in Agave 4.3, the following release, targeted for October 2026," per the Solana Foundation.
• Current target: September 28. Anza's v4.3 release schedule, published August 12, 2026, moved the target earlier: testnet and devnet feature activation began August 19 and 24; validators holding 10% and then 25% of stake were asked to upgrade on September 8 and 14; general adoption is recommended September 21; mainnet-beta feature activation begins September 28, 2026. Anza labels every date tentative, so treat September 28 as a target, not a guarantee.

As of mid-September 2026, then: approved, live on testnet and devnet, but not yet active on mainnet.

On the morning of August 12, 2026, a bad internet route originating at infrastructure provider TeraSwitch's Miami facility knocked validators holding nearly 29% of staked SOL offline. The provider fixed the route in about ten minutes, but the affected validators stayed down for the full 33 minutes because backup systems did not activate, per CoinDesk, citing staking provider Marinade Finance. Roughly 90 validators were hit; 597 of 699 kept voting, and blocks kept being produced.

Why the headlines: under Tower BFT, Solana stops finalizing if more than one-third of staked tokens go dark, and Marinade estimated the network came within about 20 million SOL of that line, per CoinDesk. The Solana Foundation gives the same picture, "nearly 29% of network stake" offline with recovery "in just over 30 minutes," and says users likely never noticed. Consensus held. There was no halt.

It is easy to read "20+20" and conclude Alpenglow would handle 29% offline stake worse than today's one-third limit. It would not, as long as the offline stake is merely offline.

• Under Alpenglow, the 80% one-round fast path would have been unavailable with about 71% of stake responsive. The 60% two-round slow path would have kept finalizing, because the protocol is designed for "40% crash failure resilience" when no stake is adversarial, per SIMD-0326. Slower for 33 minutes, not near a stop.
• The 20% offline figure in "20+20" is what remains while the protocol is also absorbing 20% adversarial stake. Against an honest-but-offline outage, the limit is 40%.

The cost is the other half of the sentence: 20% adversarial tolerance instead of Tower BFT's 33%. A TeraSwitch-style failure, though, lands on the safer side of the new math.

Per Helius, Solana mainnet has fully halted seven times: December 2020 (about 6 hours), September 14, 2021 (17 hours), April 30 to May 1, 2022 (8 hours), June 2022 (4.5 hours), September 2022 (8.5 hours), February 25, 2023 (almost 19 hours), and February 6, 2024 (almost 5 hours). Since that restart the record is clean: "since February 2024, Solana has run at 100% uptime," per the Solana Foundation, and August 12 did not break the streak because blocks never stopped. It remains a fair warning about concentration, since one connectivity provider carried more than a quarter of all staked SOL, per CoinDesk; Alpenglow raises the offline tolerance but does not fix that.

Faster: anything that waits for finality before acting. A Solana Pay merchant flow that today shows "confirmed" in a couple of seconds and "final" a dozen seconds later would see both collapse toward a fraction of a second, and exchanges or card programs that credit deposits only at finality could credit them about as fast as the transaction lands. As SIMD-0326 puts it, "optimistic confirmation is superseded by faster (actual) finality."

Unchanged: fees and wallets. Alpenglow is a consensus change; transaction fees are set by a separate fee market and wallets sign the same transactions. For how the chain is used for payments day to day, see our Solana payments guide.

For SolCard users, Solana is the main road. In August 2026, 64% of SolCard deposits arrived over Solana and 43% of all deposits were SOL itself, a pattern we break down in our August 2026 crypto payments report.

Here is where finality sits in that flow. When you top up a SolCard with SOL, the balance is credited once the deposit has been processed, and it is held in fiat from then on. Card purchases never depend on a live on-chain transaction, which is why an incident like August 12 could not have declined a payment at checkout. What faster finality could shorten is the wait between sending SOL and being able to spend it. That is the entire change; fees, limits, and how the card works at the terminal do not move with Alpenglow.

A new consensus protocol for Solana, designed by Anza and published on May 19, 2025. It replaces Tower BFT and Proof of History with Votor, which finalizes blocks in one voting round at 80% stake participation or two at 60%, targeting median finality of about 150 ms versus about 12.8 s today, per Anza.

Not as of mid-September 2026. It passed governance on September 2, 2025, shipped in Agave 4.2 with activation disabled, and Anza's v4.3 release schedule has mainnet feature activation beginning September 28, 2026, with all dates marked tentative.

About 12.8 seconds from block creation under Tower BFT at 400 ms slots, per Anza; with mainnet slots now at 300 ms the wait is proportionally shorter but still on the order of ten seconds. Optimistic confirmation arrives much sooner.

For a purely offline failure, yes. Alpenglow is designed for 40% crash-failure resilience, so with about 71% of stake responsive it would have kept finalizing on its two-round path rather than approaching a halt, per SIMD-0326. The trade-off is 20% tolerance for adversarial stake rather than 33%.

Alpenglow's promise is simple: finality in around 150 ms instead of around 12.8 seconds. It passed governance by a wide margin and is on a published, tentative schedule to begin activating on mainnet on September 28, 2026; finality today is still the old finality. The August 12 near miss showed how close Solana's one-third offline limit is when one provider carries that much stake; Alpenglow's 40% crash tolerance would have made that a slow half hour rather than a scare, at the cost of lower tolerance for malicious stake. For anyone who funds a card with SOL, the effect is a shorter wait between sending a deposit and spending it, and nothing else.

• Anza -- Alpenglow: A New Consensus for Solana
• Solana Foundation -- SIMD-0326: Alpenglow (GitHub)
• Solana Status -- SIMD-0326 governance result
• crypto.news -- Solana's Alpenglow upgrade vote passes with 98% approval
• Anza -- Agave v4.3 Release Schedule
• Solana Foundation -- Alpenglow upgrade page
• Solana Foundation -- Agave 4.2 Release Overview
• Solana Foundation -- Reduced Slot Times (SIMD-0525)
• CoinDesk -- Solana neared a freeze threshold Wednesday, Marinade Finance says
• Solana Foundation -- Solana: Building, Proving and Earning Trust in Public
• Helius -- A Complete History of Solana Outages
• Helius -- What are Solana Commitment Levels?