Solana activated its largest consensus upgrade on 21 October and secured a governance mandate to slow future token issuance, yet SOL was down approximately 14% month to date on 28 October, trading around $101–102 against roughly $118 at the end of September. The token had recovered substantially from its June lows but remained roughly 65% below its January 2025 peak near $294. Alpenglow is operational; the monetary-policy change is not. The upgrade has also introduced a structural break in reported network activity, making the distinction between economic progress and measurement changes central to assessing Solana's fundamentals.
TL;DR
- Price: SOL traded near $101–102 on 28 October against about $118.17 on 30 September, a month-to-date decline of roughly 14%. Kiln estimates a ~70% recovery from the early-June low by 12 October, when SOL traded around $108–112.
- Alpenglow: Activated on mainnet at the start of epoch 1064 on 21 October, approximately 13:36 UTC, replacing TowerBFT with Votor and targeting ~150ms finality against the prior 12.8 seconds. Phase two, Rotor, is not yet scheduled.
- Measurement break: Vote transactions are no longer included in blocks. Aggregate transaction counts and reported TPS cannot be compared across 21 October in either direction.
- Staking reform: SGP-0002 passed with roughly 67% support, but SIMD-0550 is not live. It requires SIMD-0607 in Agave v4.4 with Firedancer alignment, and no activation date has been published.
- ETF flows: US spot funds shed about $24.8m in the week of 5–9 October, ending a record 14-week inflow streak, with further single-day redemptions reported later in the month.
- Cleaner evidence: Application revenue of $144.25m in September and record stablecoin supply of $17.51bn in the same month are more usable than headline transaction counts.
A 14% Month-to-Date Fall Through a Successful Upgrade
SOL's October illustrates why protocol delivery and token performance should be assessed separately. The network shipped its most significant consensus change since genesis and secured its first binding monetary-policy mandate, and over the same period the token fell from about $118.17 on 30 September to roughly $101–102 on 28 October, a decline of approximately 14% month to date. YCharts records $101.20 for 28 October, though that observation is a midnight-UTC reference rather than a closing price, and the month is not complete.
The longer arc requires two framings held at once. Kiln's October institutional review estimates SOL had recovered roughly 70% from its early-June low by 12 October, when it traded around $108–112. Measured against the January 2025 record near $294, a price of $102 is about 65% lower. A token can rise sharply from a cyclical floor while remaining far below an earlier peak, and percentage recoveries from depressed levels systematically overstate how much of a prior decline has been reversed: a 70% gain does not undo a 65% fall.
Price weakness coinciding with the upgrade does not establish that investors disregarded it. Broader digital-asset conditions weakened over the same weeks, and expectations of an upgrade with a long public timetable may already have been reflected in prices well before activation. What the divergence does establish is that Solana's token performance and its protocol progress moved on different timescales in October, which is why the network's economics — application usage, fee generation, issuance and the share of activity accruing to holders — need assessing on their own terms.
Alpenglow Is Live: The Catalyst Has Already Landed
Alpenglow (SIMD-0326) activated on Solana mainnet at the start of epoch 1064 on 21 October 2026, at approximately 13:36 UTC, following devnet activation on 25 September and shipping in Agave 4.3. The genesis block was slot 459,652,999, certified with votes from 446 of the network's 654 validators, with Alpenglow consensus running from slot 459,653,000. The Solana Foundation describes it as the biggest change to the core protocol since genesis.
The first phase, Votor, replaces TowerBFT. Under the previous design validators voted by submitting transactions, and a block became final once enough votes had accumulated over 32 slots, taking about 12.8 seconds. Validators now send votes directly to one another and aggregate them into certificates, finalising a block in one or two rounds, with a target of roughly 150 milliseconds. The Foundation states that confirmed and finalized now describe the same state. Execution is untouched: the SVM, transactions, programs and fees are unchanged, though Proof of History no longer paces block production.
Block time was cut separately from 400ms to 200ms across four stages completed in October, and the two measures should not be conflated — the 150ms figure describes how long a transaction takes to become final, not how often blocks are produced. Two qualifications matter for any thesis written after 21 October. Alpenglow is no longer a forthcoming catalyst and cannot be carried as prospective upside. And it remains phase one: Rotor, which replaces Turbine for block propagation, is planned for a later release and is not yet scheduled, so the full programme's benefits should not be attributed to the completed one. Lower settlement latency improves an application's technical proposition without automatically generating revenue or token demand, and the test ahead is whether applications convert faster settlement into fee-paying throughput.
One operational detail bears on the tokenomics timetable below: Firedancer and Frankendancer were not supported for the migration, with validators required to run Agave v4.3.2, Jito-Solana or Harmonic Salsa. Validators must also hold at least 0.8 SOL in their vote account above the rent-exempt minimum to pay the Validator Admission Ticket fee, charged at the start of each epoch for inclusion in the next.
The Measurement Break: Transaction Counts Are No Longer Comparable
The upgrade introduced a discontinuity into Solana's headline statistics that can be misread in either direction. Vote transactions are no longer included in blocks, and Kiln notes that reported network TPS and transaction counts will therefore drop sharply as a matter of metrics accounting rather than as evidence of reduced user activity.
The consequence is narrower than it first appears, and the honest statement is a negative one. A comparison between 120.68 million transactions before the upgrade and 98.52 million after it produces an apparent 18% decline, but the two observations do not count the same set of activities, so the comparison cannot establish the direction of underlying economic activity at all — it neither demonstrates a contraction nor rules one out. The break cuts the other way on throughput claims too: 98.52 million daily transactions is approximately 1,140 per second across 86,400 seconds, which cannot be reconciled with a figure of 5,165 per second unless the two use different windows or transaction definitions. Where published figures conflict, the reconciliation belongs in the analysis rather than the selection of whichever number suits the argument.
| Metric | Comparability after Alpenglow | Analytical use |
|---|---|---|
| Total reported transactions | Structural break | Avoid pre/post comparisons |
| Validator vote transactions | No longer in blocks | Exclude from economic analysis |
| Non-vote transactions | More informative | Consistently defined user activity |
| Application revenue | Usable on consistent methodology | Application-level demand |
| Stablecoin supply | Comparable if timestamps match | Settlement liquidity |
| DeFi TVL | Comparable with valuation adjustments | Capital committed to protocols |
| Active addresses | Requires methodology review | Participation, not unique users |
DeFi TVL and active-address counts need separate treatment, because the voting change does not explain movements in the dollar value of deposited assets or in wallet participation. TVL moves with token prices, withdrawals and liquidity rotation between protocols; address counts move with user behaviour, automation and incentives. Those series remain usable, and should not be bundled with the transaction-count break into a single narrative of either expansion or contraction. Indexing requirements also changed at the upgrade, so providers' own methodologies may shift across the same boundary.
Staking Reform: Approved, Not Activated
Solana's first binding on-chain governance process has mandated slower issuance without yet delivering it, and the distinction is load-bearing for anything written about staking returns.
SGP-0002 approved the accelerated disinflation in SIMD-0550, which doubles the annual disinflation rate from 15% to 30% and brings inflation to its unchanged 1.5% terminal floor around 2029 rather than 2032. Helius models the change as reducing cumulative issuance by roughly 18.9 million SOL over six years. But a passed proposal is a mandate, not a mainnet change: Kiln states that the faster curve requires SIMD-0607, which has to ship in Agave v4.4 with Firedancer alignment before a feature gate can be scheduled, and as of 28 October no activation date has been published. Current staking rewards still follow the existing 15% schedule. That Firedancer was not supported for the Alpenglow migration is worth holding alongside the dependency, since the alignment work sits on the critical path.
Published yield figures need handling with the same care. VanEck reported a gross network staking yield of 4.35% on 27 October. Fund distribution yields and forward return estimates circulate alongside it, and they are not the same measure: they reflect different calculation periods, validator performance, fee and MEV income, liquid-staking economics, distribution policy and annualisation method, so a fund's distribution yield is not the network's inflation reward rate and a forward estimate is not an observed return. Helius's modelled path — about 4.34% in year one, 3.00% in year two and 2.25% in year three — assumes a 68% staking rate and excludes commissions, MEV and block rewards, which its authors describe as a worst case for periods of very low network activity. Its year-one figure sits close to the current observed network yield, but that numerical similarity does not establish that the accelerated schedule is active.