Fidelity Flags Solana Concentration Risk After Two Outage-Free Years

Solana’s uptime record faces a concentration test

The full October report estimates that about 20 validator identities held more than one-third of active stake in Q3. A coordinated group above that threshold could obstruct finality, but would not gain unrestricted access to balances or the ability to rewrite finalized transactions.

Identities are not necessarily separate operators. The report also estimates roughly 40% of stake in Frankfurt and Amsterdam, while warning that IP-based location estimates are imperfect. Its October infrastructure snapshot places a single network provider near 29% of stake.

The software distinction is similarly important: Jito-Solana incorporates Agave code, so separate client labels do not establish independent protection from shared defects. FCAT identifies independent implementations such as Firedancer as a way to reduce that dependency.

Taken together, those observations suggest that a long list of validator accounts can overstate the number of separate failure paths. Moving a machine to another city would address location exposure, for example, without necessarily changing its software or network provider.

That is an operational distinction for businesses choosing infrastructure. Two service contracts may offer less redundancy than expected if both ultimately depend on the same upstream connection. Assessing that overlap requires looking beneath the number of providers visible to an application.

August’s routing failure shows Solana’s remaining exposure

The Solana Foundation’s September account of the incident says an August 12 routing failure took nearly 29% of stake offline. Blocks continued, and the affected provider recovered in just over 30 minutes.

Foundation technology executive Jacob Creech said earlier work had reduced TeraSwitch’s hosting concentration from 38% to below 30%. The episode illustrates both sides of FCAT’s assessment: the chain absorbed a substantial failure, while a single provider still represented a significant dependency.

It also explains how an outage-free record can coexist with operational incidents. A failed hosting service and a stopped blockchain describe different events. Successful block production during one disruption does not show that every conceivable combination of failures would be tolerated.

FCAT’s summary reports roughly 99.7% vote success and financial activity including more than $2 trillion in first-quarter stablecoin volume. It puts tokenized real-world assets using Solana as a distributed asset layer at about $4.4 billion, up from roughly $1 billion a year earlier.

These quantities should be read separately. Transfer volume measures activity over a period, while an asset-value estimate measures a stock of value. Adding them together would produce a misleading measure of adoption or capital at risk.

For settlement users, the practical inference is that greater financial activity increases the consequence of interruptions. It does not, by itself, prove that an interruption has become more likely. Nor does a reliable chain guarantee uninterrupted access through every exchange, wallet or application.

Alpenglow changes finality, not every failure dependency

The research lands ahead of the planned Alpenglow mainnet transition. The upgrade provides the next opportunity to compare an operating network with the baseline FCAT has documented, rather than assume projected improvements are already established results.

Anza’s operator documentation describes replacing TowerBFT’s voting and finality logic with Votor in the Agave 4.3 release cycle. It targets finality around 150 milliseconds under normal conditions, but the time at which other infrastructure learns that result can be longer.

Nodes outside the admitted validator set can receive finalization certificates in subsequent block footers. That route typically adds up to one produced-block interval and can take longer if blocks are skipped or delayed. Direct delivery through an admitted partner introduces connection latency and depends on that partner remaining reachable.

Anza therefore advises considering multiple partners for resilience. The documentation shows why a consensus speed target and the confirmation time displayed by a trading or payment application need separate measurement. It also leaves operators with choices about connectivity, funding and infrastructure placement.

The wider market’s mood offers another separate reading. Alternative.me’s Bitcoin-focused sentiment page displayed 55, or Greed, against 63 in its previous-day field at the check. That index does not measure Solana’s technical readiness or infrastructure concentration.

Source: Alternative.me. This is the displayed reading at the check, rather than a claim that a new October 18 daily observation had been published.

The next test is production performance after the consensus transition, including continued participation and delivery of finality information to applications. FCAT’s figures are dated snapshots; current operator-level independence and the eventual distribution of infrastructure after migration remain questions for subsequent measurement.