Fidelity's Solana health check: reliable, but too concentrated

Fidelity's research unit has given Solana a mostly clean bill of health on reliability, while warning that control over the network still sits with too few validators, data centres and software stacks. The report, "The State of Solana: Network Health and Resilience", was written by Mia Galvin and Kevin Kelly of the Fidelity Center for Applied Technology (FCAT) and published on October 16. Its central argument is that counting validators is not enough: stake distribution, hosting, geography and client software have to be read together. Here are the five findings that matter most.

Reliability: two years without a major outage

FCAT says Solana has not had a major mainnet outage since February 6, 2024 — a big change for a chain once known for halts. According to figures from the report cited by Crypto Briefing and Solana Compass:

• Vote success averaged about 99.7% in 2026, with average vote latency of 1.08 slots.
• Tokenized real-world assets on Solana grew from about $1 billion to $4.4 billion year on year.
• The network handled more than $2 trillion of stablecoin volume in the first quarter of 2026 alone.

Stake: about 18 validators could halt finality

Solana stops finalizing blocks if one-third of stake goes offline or acts together. FCAT counts roughly 18 validators holding that much stake, based on third-quarter 2026 data. That number — often called the Nakamoto coefficient — is the report's clearest warning sign.

Hosting: one provider holds nearly 29%

A single hosting provider accounts for close to 29% of total stake, which FCAT says is above the limits the Solana Foundation itself considers acceptable. The risk is not theoretical: staking firm Kiln notes that an August 12 routing fault at one provider briefly knocked about 28.8% of staked SOL offline, within a few points of the halt threshold.

Geography: Frankfurt and Amsterdam carry about 40%

Roughly 40% of stake sits in just two European cities, according to geolocation data FCAT cites — with the caveat that IP-based location is not always precise. A regional network or power problem would therefore hit a large slice of validators at once.

Software: most stake runs code from one lineage

About 63% of stake runs Jito-Solana and about 26% runs Harmonic, the report says, and both are tied to the dominant Agave codebase. Different client names do not guarantee real diversity if a single bug in shared code can affect most of the network.

What to watch next

FCAT calls the overall trajectory positive but says the next phase will test whether speed gains come "at the expense of decentralization or network security." The first big test is the Alpenglow consensus upgrade, which aims to cut finality to about 150 milliseconds. For background on how validators secure chains like Solana, see our guide to proof of work vs proof of stake.