Solana is replacing the engine under its blockchain, and it is happening live this month
Alpenglow is the biggest change to how Solana reaches agreement in the network's history: it retires Tower BFT, aims to cut finality from about 12.8 seconds to roughly 150 milliseconds, and is rolling out to validators in stages through September. Here is what is actually changing, when, and the caveats behind the headline numbers.
Solana is deploying Alpenglow, which Anza calls the largest overhaul of its consensus mechanism ever, in a staged validator rollout through September 2026. It replaces the Tower BFT voting system with a new engine called Votor, targeting finality of roughly 150 milliseconds instead of about 12.8 seconds. A companion change to block propagation, called Rotor, comes later in its own separate upgrade. The change is not a single launch: feature activation on mainnet begins on 28 September, with the full transition expected to complete in October. The headline speed and resilience figures come with real conditions, and this is an engineering story, not a price one.
Most blockchain upgrades tinker at the edges. This one swaps out the core. Alpenglow is a complete redesign of how Solana's validators agree on the order of transactions and decide when that order is final, and Anza, the firm that builds Solana's main software, calls it the biggest consensus change in the network's history. It is also not a distant roadmap item: the rollout is happening on the live network right now, in stages, through September and into October. Here is what is genuinely changing and where the marketing numbers need an asterisk.
What is Alpenglow?
At its heart, this first stage of Alpenglow overhauls how validators vote to finalise blocks. It retires Tower BFT, Solana's original voting mechanism, in favour of a new engine, and it moves validator votes off-chain into compact cryptographic certificates so they no longer clog the chain or cost a per-vote fee. The promise is dramatically faster finality and a network that keeps working even when a large share of validators are malicious or simply offline. A separate change to how blocks are spread across the network, called Rotor, is planned as a later upgrade.
Is it live yet, and when does it actually happen?
This is the detail most coverage gets wrong. Alpenglow is not a single "launches on 28 September" event. It is a graduated rollout tied to Anza's Agave 4.3 software schedule: roughly 10% of validators adopted it around 8 September, the 25% mark around 14 September, and general adoption is recommended by 21 September. On 28 September the Alpenglow feature gate begins switching on on mainnet, after about a week of the new software running widely. Anza's own developers have pointed to October for the full network-wide transition to complete. So the accurate framing is "staged through September, activation begins 28 September, completion expected October," not "goes live on the 28th."
How much faster is it, really?
The target is a big jump: from Tower BFT's roughly 12.8 seconds of deterministic finality down to a median of about 150 milliseconds, with a white-paper best case as low as 100 milliseconds. Those figures are conditional, and worth reading carefully:
- The single-round fast path applies when a block gets at least 80% of stake approving it; Anza's own schedule attaches the ~150ms median to that 80% threshold.
- If round one lands between roughly 60% and 80%, a second round runs.
Anza is explicit that these numbers come from simulations on the current mainnet stake distribution and exclude computation overhead, and the measured figure is higher than the headline: test-cluster data cited in the release schedule had 96% of blocks hitting fast-path finalisation at around 214 milliseconds. So treat "150ms" as a strong target under good conditions, not an unconditional guarantee.
What are Votor and Rotor?
Alpenglow has two named components, but they arrive separately:
- Votor is the new voting and finalisation engine, and it is the part activating now, replacing Tower BFT. Votes become lightweight off-chain messages that are aggregated into a single certificate, so only a small header lands on-chain and validators no longer pay a fee per vote.
- Rotor is a planned block-propagation layer for a later, separate upgrade. It is designed to refine Solana's existing Turbine system, swapping the multi-layer relay tree for a single flat relay of erasure-coded data to cut network hops. For now Turbine is retained, and Rotor gets its own proposal down the line.
Is Solana really 40% fault-tolerant now?
Yes, with an important nuance. Tower BFT tolerated up to about one-third (33%) of stake being faulty. Alpenglow uses what Anza calls a "20+20" model: safety holds against up to 20% adversarial (actively malicious) stake, and liveness holds with an additional 20% of stake simply being offline, for 40% combined non-functional stake. So the network can keep finalising blocks with up to 40% of stake down, but the threshold for outright hostile stake is 20%, not 40%. In fact, against purely malicious stake that 20% is lower than Tower BFT's roughly 33%; what rises is the combined tolerance for malicious plus offline stake. Do not read "40% fault tolerance" as "Solana can survive 40% of validators attacking it."
What do validators have to do?
To join the new consensus set, validators must register a BLS public key on-chain (using a recent Solana command-line release). Per Solana Foundation guidance, validators without a registered key are excluded from the admitted set and stop participating in consensus. Registration has been open since July, alongside a validator admission ticket system and a cap of 2,000 admitted validators. This is a real operational step, so it is fair to note the risk that operators who skip it get left out, but that is Anza and the Foundation's stated requirement, not a prediction that validators will fail to comply.
Didn't validators already vote on this?
They did, about a year ago. The governance proposal (SIMD-0326) passed in September 2025 with 98.27% voting yes and 52% of stake participating. That vote is not this month's news. What is happening now is the actual deployment of the thing that was approved back then. It is also separate from another Solana mainnet change landing this month, a new transaction format, so do not conflate the two.
Tower BFT vs Alpenglow
| Tower BFT (old) | Alpenglow (new) | |
|---|---|---|
| Finality | ~12.8 seconds | ~150ms median target (measured ~214ms) |
| Voting | On-chain, per-vote fee | Off-chain, aggregated certificates (Votor) |
| Block spread | Turbine relay tree | Turbine retained (Rotor is a later, separate upgrade) |
| Fault tolerance | ~33% faulty stake | 20% adversarial + 20% offline (40% combined) |
| Status | Being replaced | Staged rollout Sept, activation begins 28 Sept, completes ~Oct |
Circle built a blockchain where USDC pays the fees, and the validators are names like BlackRock and Visa
Circle's Arc is scheduled to open its public mainnet on 16 September: an EVM-compatible Layer 1 where the USDC stablecoin itself pays for gas, secured by a closed set of institutions including BlackRock, Visa, Mastercard and DTCC. It is open to build on and closed to validate, and the separate ARC token is not the gas and has no confirmed trading date. Here is what is actually launching.
Cronos halted and rewound its own blockchain after a Tectonic exploit. Here is what that really means
On 30 August, validators froze the Crypto.com-linked Cronos chain and rolled it back about two hours to reverse a lending exploit. The headline says $75M, but only $6.3M actually left. The bigger story is a chain choosing to rewrite settled history, and the argument that restarts every time one does.
The SEC cancelled its own crypto vote, then approved the proposal four days later. Here is what 'Regulation Crypto Assets' actually says
After cancelling a public meeting, the SEC approved a crypto-offering proposal on 18 August by seriatim vote. The text is now public: a tiered fundraising exemption up to $75 million, a $5 million startup path, and a safe harbour that turns on ceasing 'essential managerial efforts.' A proposal, not law, unlikely before 2027.
Someone minted 4 billion tokens out of thin air on Harmony. Here is what we know
On 12 August an attacker exploited a consensus-verification flaw on the Harmony blockchain to mint roughly 4 billion ONE tokens, inflating supply about 26 percent and crashing the price by a third. The realised haul was only a few million dollars, but tokens minted from nothing is the real alarm.
Wall Street is putting bank deposits on a blockchain. The catch: it is barely 'crypto' at all
In August 2026, Wells Fargo became the latest giant to say it will tokenise customer deposits, joining JPMorgan, Citi and Bank of America in a shared network due in 2027. Wall Street calls it 'tokenisation,' but a tokenised deposit is not a stablecoin, and the change is smaller than the word suggests.
The SEC is about to start writing crypto's rulebook itself, with Congress stalled
On Friday the SEC votes on whether to propose 'Regulation Crypto,' a set of token-fundraising rules it has been sketching since 2025, now arriving as the CLARITY Act sits stalled in the Senate. Here is what tomorrow's vote actually decides (less than the headlines suggest), what the rules would do, and why nothing changes on Friday.