Fidelity's FCAT Finds Solana More Reliable, With Stake, Hosting and Client Software Still Concentrated

Fidelity's research arm has published its first health check of the Solana SOL$104.91-1.6% network, and its verdict is specific: the chain has become far more reliable, while control over it still sits with a small group of validators, two European cities and a handful of hosting providers. 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 16 October 2026. It finds that about 18 validators together hold the one-third of stake needed to stop the network finalizing blocks. Solana Compass's own stake data puts that number at exactly 18, where it has stayed every epoch since early August.

FCAT is Fidelity's technology research unit, separate from the fund business that runs Fidelity's staked SOL ETF, and the report says its opinions are the authors' own. The authors call it a baseline to measure Solana against after the Alpenglow consensus upgrade.

What Fidelity's FCAT found on Solana's reliability

FCAT's reliability findings for Solana are positive. The network has had "no reported mainnet outage across two consecutive years," vote success averaged about 99.7% during 2026 and vote latency averaged 1.08 slots, according to the full FCAT report, which takes the vote figures from Syndica. The stake-weighted skip rate, the share of scheduled blocks that leaders failed to produce, stayed between about 0.02% and 0.12% across a ten-epoch sample of Trillium data in the report. Block space has headroom too: after blocks grew to 100 million compute units in late July, FCAT found utilization between roughly 30% and 45% over a sample week.

Solana's Nakamoto coefficient: 18 validators, two countries, two ASNs

The Nakamoto coefficient counts the smallest number of validators that together control more than a third of stake. Under Solana's consensus rules, a group that size could stop everyone else from finalizing blocks, though it could not take user funds or rewrite finalized transactions. Citing Q3 2026 data from Sonda, FCAT says that one-third threshold is "concentrated across approximately 18 validators, two countries, two autonomous system numbers (ASNs), and two base-client implementations." An ASN is a network operator's routing identity, so two ASNs means two providers' networks.

Our numbers match. Solana Compass's validator rankings show a Nakamoto coefficient of 18 at epoch 1060 on 17 October 2026, unchanged in every epoch since epoch 1012 on 4 August. The ten largest validators hold 24.42% of active stake, within a narrow 24.08% to 24.66% band over those 49 epochs. Our count of validators with active stake is 678, down from 691 at epoch 1012; FCAT's Q3 figure of about 717 uses a different date and definition.

FCAT adds a caveat that cuts the other way. A validator count measures vote accounts, and one operator can run several. The report notes that Figment operates two of the top five validators in the superminority, while Kraken and Coinbase cbBTC$81,598.75-0.7% each run multiple validators. In our 17 October snapshot, Figment's own validator ranks first with 4.00% of stake and the Ledger by Figment validator ranks seventh with 2.11%, about 6.1% between them. "Validator count should realistically be treated as a measure of participation, not a complete measure of independent control or operator diversity," the authors write.

Frankfurt, Amsterdam and the TeraSwitch hosting problem

Geography is where the report is most pointed. About 59% of stake sits in Germany, the Netherlands and the United States, including roughly 40% in Frankfurt and Amsterdam alone, according to Sonda data that FCAT cites with a warning that IP geolocation may not be precise. Our validator data shows the same pull: five of the ten largest validators by stake list Frankfurt as their location.

Hosting is sharper still. The Solana Foundation Delegation Program asks validators it supports to keep any single ASN below 25% of stake and any data center below 15%. An October 2026 snapshot from Validators.app put the largest ASN at about 29%, "still above the program's ceiling," which the authors call "a gap between the network's desired risk profile and its actual infrastructure distribution." The report's chart of stake by provider puts TeraSwitch first at about 124 million SOL, ahead of Cherry Servers at 62 million and Latitude at 53 million.

That concentration has already been tested. FCAT's account of the August 2026 routing failure, which knocked about 90 validators offline for 33 minutes and left the affected stake roughly 4.5 percentage points short of the one-third halt threshold, cites our reporting on the TeraSwitch outage. The report lists mitigations including private networks such as DoubleZero 2Z$0.033-1.8%, and the Solana Foundation's STRIDE risk program and SIRN incident-response network, both introduced in early 2026.

Solana client diversity: Jito still runs most of the stake

Software is the third concentration. FCAT's chart of Sonda data puts 36.2% of stake on Jito BAM and 26.9% on the standard Jito JTO$0.468-1.1% client, about 63% combined, with Harmonic variants at about 24% and plain Agave at 1.2%. The report's argument is that most of these share code. Jito's client is a fork of Anza's Agave, and products such as Harmonic, Rakurai and BAM change how blocks are scheduled and built while leaving the shared consensus and execution code in place. "Adoption of two client configurations did not truly eliminate reliance on common Agave code," the authors write.

Firedancer, built by Jump Crypto on a separate C++ codebase, is the client FCAT counts as real independence. Configurations labelled as Firedancer variants in the same chart (Firedancer, Harmonic-Firedancer and FireBAM) add up to about 15% of stake. Overclock's Mithril client, written in Go, produced blocks on the Alpenglow community test cluster in June 2026 but has no material mainnet share yet, according to the report.

Validator economics: the 200,000 SOL break-even

FCAT's economic section asks whether smaller Solana validators can stay in business. The report puts annual running costs at around $60,000 and up, and estimates break-even at about 200,000 SOL of delegated stake, assuming a 5% commission on inflation and MEV rewards. Costs are paid in dollars while revenue arrives in SOL, and the authors write that sustained price falls "could leave smaller validators insolvent."

Software choice shows up in revenue. Trillium data in the report shows Harmonic validators earned a median 0.008 SOL in priority fees per 10 million compute units over the seven days to 15 October, double the 0.004 SOL earned by plain Agave. FCAT attributes the gap to a 50-millisecond delay in Harmonic's transaction ordering that captures more priority fees, one of the small trade-offs between revenue and speed the report says block-building systems make.

Three governance changes bear on those margins. SIMD-0550 would double the yearly cut in Solana's inflation rate from 15% to 30%, and FCAT notes it was approved in the SGP-0002 governance vote; the authors model about 22.3 million SOL less issuance over six years. It is not yet in effect: the inflation rate the Solana RPC reports at epoch 1060 still uses the 15% taper, according to the inflation data Solana Compass tracks. SIMD-0357's Validator Admission Ticket would charge 1.6 SOL per epoch against roughly 2.14 SOL in current voting costs, about 25% cheaper overall, though as a flat fee it weighs more heavily on small validators.

SIMD-0123 would pay block revenue to delegators natively, which FCAT says could start a "race to zero" in commissions that larger operators are better placed to win. Commissions are already low across much of the network. As of 17 October, 30.4% of active stake sat with 233 validators charging 0% commission on inflation rewards, according to Solana Compass commission data. A further 23.7% sits at 100%, mostly exchange and custodian validators staking their own or customers' SOL, which keep all rewards by design.

What to watch on Solana after Alpenglow

FCAT's overall view is that "the broader trajectory remains positive." It also names the test ahead: "The next phase of Solana's development will test whether these advances will come at the expense of decentralization or network security." On speed, the report notes that 200-millisecond slots have raised vote latency in some regions, and calls for a cautious approach to further cuts.

The baseline points to what to track next: the largest ASN against the Foundation's 25% ceiling, Firedancer's stake against Agave-based clients, and the validator count as lower issuance and flat admission fees arrive. On our data, the Nakamoto coefficient has sat at 18 for ten weeks, so the stake map is stable. Whether that is good news depends on which of those markers moves first.