Solana Fees and Transaction Speed Explained

Two numbers get repeated about Solana so often that they risk becoming background noise: 400 milliseconds and a fraction of a cent. One describes how quickly the network produces blocks, the other what it charges to include your transaction in one. Both are real, both are measurable, and both are frequently misunderstood — sometimes by critics, sometimes by fans who quote them without knowing what they actually mean.

This article unpacks the machinery behind those numbers. We will look at how Solana keeps time with Proof of History, why its runtime can execute thousands of transactions in parallel, what the difference is between a confirmed transaction and a finalized one, and how the fee model works down to the lamport. We will then put the costs next to the systems most Europeans use every day — card networks and SWIFT wires — and be honest about where Solana has stumbled in the past.

If you bank with a Solana-based institution, or are considering it, this is the technical foundation underneath your account. At NGIBANK, the world's first Solana CBDC bank, every NGI transfer ultimately rides on the mechanics described below. Understanding them helps you judge the claims — ours included — with a clear eye.

What "400 milliseconds" actually means

Solana organizes time into slots of roughly 400 milliseconds. In each slot, one validator — the leader for that slot — assembles transactions into a block and streams it to the rest of the network. So when people say Solana settles in 400ms, they are describing the block production rhythm: your transaction is typically included in a block well under a second after you submit it.

But inclusion is not the whole story. Payment engineers distinguish between two milestones. The first is optimistic confirmation: a supermajority of validators, weighted by stake, has voted on the block containing your transaction. This usually happens within about a second, and once reached, the transaction will not be rolled back under any conditions the protocol considers normal. Wallets, exchanges and payment processors treat optimistically confirmed transactions as done.

The second milestone is finality in the strict consensus sense: the block has been rooted, which happens after roughly 32 subsequent slots — call it 13 seconds. At that point reversal is not merely improbable but impossible without violating the protocol itself. For context, a card payment feels instant at the terminal but can be charged back for months, and a SEPA credit transfer is typically irrevocable only once the receiving bank posts it. Thirteen seconds to absolute finality is, by the standards of money movement, remarkable. Under a second to practical finality is the number that changes user experience.

Proof of History: a clock, not a consensus mechanism

The headline innovation behind Solana's speed is Proof of History, and it is widely misdescribed. PoH is not how Solana reaches agreement — that job belongs to its proof-of-stake consensus, Tower BFT. PoH is a cryptographic clock.

Here is the problem it solves. Distributed systems struggle with time. If a thousand validators around the world each use their own clock, they will disagree about the order of events, and resolving those disagreements through messaging costs precious seconds. Bitcoin solves ordering by making block production deliberately slow. Most proof-of-stake chains solve it with rounds of communication before each block.

Solana instead makes time itself verifiable. A validator runs SHA-256 repeatedly, feeding each hash output into the next hash input. Because each hash depends on the previous one, the chain of hashes cannot be computed in parallel or faked in advance — producing hash number one million provably took a certain amount of sequential work, and therefore a certain amount of real time. Transactions and events get stamped into this hash stream, which means every validator can look at the stream and agree on the order and approximate timing of everything in it — without exchanging a single message about clocks.

The consequence is profound: the leader can stream a block continuously as transactions arrive, rather than proposing it and waiting for rounds of votes about ordering. Consensus still happens, but it happens after the fact and in parallel with block production, not as a gate in front of it. That is where most of the latency savings come from.

The rest of the pipeline: Sealevel, Gulf Stream and Turbine

A fast clock alone does not make a fast network. Three other components matter just as much.

Sealevel: parallel execution

Every Solana transaction must declare, up front, exactly which accounts it will read and which it will write. This sounds like a chore for developers — and it is — but it gives the runtime, called Sealevel, a gift: it can see instantly which transactions touch overlapping state and which do not. Two transfers between entirely different pairs of accounts cannot conflict, so Sealevel executes them simultaneously on different CPU cores. On modern hardware with dozens of cores, thousands of non-conflicting transactions run in parallel. Ethereum's virtual machine, by contrast, processes transactions one at a time; the chain's throughput is capped by a single thread of execution. If you want a deeper comparison of the two architectures, we have written one in Solana vs Ethereum for payments.

Gulf Stream: no waiting room

Most blockchains keep a mempool — a public waiting room where transactions sit until a block producer picks them up. Solana skips it. Because the leader schedule is known in advance, your wallet or RPC node forwards transactions directly to the current leader and the next few in line. Transactions spend milliseconds in transit instead of seconds in a queue, and validators are spared the memory burden of a large shared pool.

Turbine: spreading blocks like BitTorrent

A leader producing blocks every 400ms cannot upload each full block to thousands of validators directly; the bandwidth math does not work. Turbine breaks each block into small packets called shreds and transmits them through a tree of validators, each layer forwarding to the next — conceptually similar to how BitTorrent distributes files. Propagation load is shared across the network, so block size can grow without the leader's uplink becoming a bottleneck.

The fee model: lamports, priority fees and local markets

Solana's fee system is refreshingly simple at the base layer. The unit of account is the lamport — one billionth of a SOL. Every transaction pays a base fee of 5,000 lamports per signature, and most transactions carry exactly one signature.

Do the arithmetic at a SOL price of €150: 5,000 lamports is 0.000005 SOL, or about €0.00075. Not three-quarters of a cent — three-quarters of a thousandth of a euro. You could pay the base fee on more than 1,300 transactions before spending a single euro cent.

On top of the base fee sits an optional priority fee, priced in micro-lamports per compute unit. A compute unit measures how much work your transaction demands of the runtime; a simple transfer needs few, a complex DeFi interaction needs many. By attaching a priority fee, you bid for earlier placement in the leader's schedule when blocks are contested. During calm periods the priority fee needed is effectively zero; during genuinely busy moments, a well-chosen priority fee of a few hundred thousand micro-lamports per compute unit might add a few tenths of a cent to a transfer.

The most elegant part of the design is local fee markets. Because every transaction declares the accounts it touches, congestion can be priced per account rather than network-wide. When a popular token launch turns one account into a battlefield, priority fees spike for transactions writing to that account — while an unrelated payment between two other accounts proceeds at the ordinary price. Compare this with Ethereum's historical dynamic, where a single popular NFT mint raised gas prices for every user on the chain, or with card networks, where fees never respond to congestion because the settlement layer is nowhere near real time anyway.

What a payment actually costs, in euro cents

Abstract lamports are less useful than concrete examples, so consider three:

• Sending €2,000 of digital euros to a supplier. One signature, base fee only in normal conditions: roughly €0.0008. As a percentage of the amount: 0.00004%.
• A €4.50 coffee paid in USDC. Same fee, roughly €0.0008 — about 0.02% of the purchase. A card acquirer would typically charge the merchant ten to forty cents on that coffee once fixed fees are counted.
• A busy-hour transfer with a healthy priority fee. Base fee plus, say, 100,000 lamports of prioritization: about €0.016. Under two cents, for guaranteed fast inclusion during peak demand.

Notice what is absent from these examples: any relationship between fee and transfer amount. Solana charges for computation and bandwidth, not for value moved. Sending €2 million costs the same as sending €2. Every legacy payment rail — cards with their percentage-based interchange, correspondent banks with their tiered wire fees — breaks that principle.

How this compares with cards and SWIFT

Card payments feel instant, and for the shopper they are. Under the surface, the economics and timing look different. In the EU, interchange fees are capped at 0.2% for consumer debit and 0.3% for consumer credit; add scheme fees and acquirer margin, and European merchants commonly surrender 1% or more of each transaction, with 2–3% typical outside the regulated caps. The merchant then waits one to three business days for settlement — the authorization was instant, the money was not. And the payment can be disputed for months afterward.

International wires are harsher still. A SWIFT transfer routinely costs €15–50 once sending fees, correspondent bank deductions and FX margins are tallied, and takes one to five business days as the message hops between intermediary banks — none of which operate on weekends. We have broken down the full anatomy of these costs in wire transfer fees explained, and compared the two big rails in SEPA vs SWIFT transfers.

Put side by side in words rather than a table: Solana settles in under a second at well under a tenth of a cent, around the clock; SEPA settles same-day-to-next-day cheaply but only for euros within Europe, with instant SEPA covering a growing but incomplete share; cards settle to merchants in days at 1–3% all-in; SWIFT settles in days at €15–50. The gap is not incremental. It is two to four orders of magnitude on cost and three to five on speed.

Firedancer and the upgrade path

Solana's performance story is not static. The most consequential development is Firedancer, an independent validator client built from scratch in C by Jump Crypto. Its significance is twofold. First, raw performance: Firedancer's networking and execution pipeline has demonstrated throughput far beyond the original client in test conditions — over a million transactions per second in isolated demos, though real-world gains are bounded by consensus and hardware diversity. Second, resilience: a bug in a single client once meant the whole network halted together. With two independent implementations validating side by side — the hybrid "Frankendancer" deployments came first, full Firedancer following — a flaw in one client no longer takes down the network.

Earlier upgrades already reshaped daily behaviour. The adoption of QUIC for transaction ingestion, stake-weighted quality of service, and the priority fee market itself were all responses to congestion incidents — evidence that the network's weak points get engineered away rather than explained away.

What this means for banking use cases

Why should a bank care about sub-second finality and sub-cent fees? Because they dissolve constraints that banking has treated as laws of nature.

Cut-off times exist because batch settlement systems close. On Solana there is no closing time: a transfer at 03:00 on New Year's Day settles exactly like one at 11:00 on a Tuesday. Minimum viable payment sizes exist because fixed fees make small transfers uneconomical; at €0.0008 per transaction, paying a freelancer €9 for a fifteen-minute task, or streaming salary daily instead of monthly, becomes rational. Pre-funding exists because slow settlement forces institutions to park liquidity in advance along every corridor; sub-second settlement lets money arrive when it is needed rather than days before.

NGIBANK's architecture is a direct application of these properties. NGI, a central-bank-grade digital currency issued on Solana, moves between accounts with the finality characteristics described above, and incoming SEPA or SWIFT wires can arrive directly as digital euros on Solana or USDC — wire to crypto, crypto to wire. Each customer still gets a Dutch IBAN and a payment card, because the point is not to replace the interfaces people know but to replace the settlement layer underneath them. NGIBANK B.V. operates under MiCA authorization from the Dutch AFM, which matters here for a technical reason too: regulated custody means the bank, not the customer, worries about key management while the customer keeps the speed.

Honest caveats: where the story is complicated

No serious analysis of Solana's performance can skip its failures. The network suffered several full outages in 2021 and 2022, most infamously when bot traffic during token launches overwhelmed transaction ingestion. In April 2024, a flood of spam transactions caused severe congestion: the chain kept producing blocks, but ordinary transactions were dropped at high rates for days until patches landed. Fees stayed low throughout — availability, not cost, was the casualty. The record since then is much stronger, and each incident produced structural fixes, but a payments professional plans for the tail risk, not the average day. This is one reason NGIBANK maintains classical rails alongside the chain: if one path degrades, value still moves.

The second caveat is hardware. Running a Solana validator requires serious equipment — server-grade CPUs with many cores, 256 GB or more of RAM, fast NVMe storage and datacenter bandwidth. That is thousands of euros of hardware plus ongoing costs, far beyond the consumer laptop that can validate some other chains. Fewer people can afford to participate, which concentrates validation among professional operators. Solana's counterargument — that meaningful decentralization lives in the number and independence of validators, currently over a thousand, rather than in hobbyist accessibility — is defensible, but the trade-off is real and worth naming.

Third, fees are denominated in SOL, so their euro value floats with the SOL price. At €150 per SOL the base fee is €0.00075; if SOL traded at €1,500, it would be €0.0075. Still trivial, but the point generalizes: a network whose token appreciates a hundredfold without fee-schedule adjustments would need, and historically networks have made, parameter changes to keep fees rational.

This article is for informational purposes only and does not constitute financial, tax, investment or legal advice.

Frequently asked questions

Is Solana really faster than Visa?

They are different layers, so compare carefully. Visa authorizes in about a second but settles funds to merchants one to three business days later, with chargebacks possible for months. Solana reaches optimistic confirmation in about a second and irreversible finality in roughly 13 seconds — authorization and settlement collapse into one event. On raw throughput, Visa's network handles a claimed capacity of tens of thousands of transactions per second; Solana routinely processes thousands per second in production with much higher demonstrated ceilings. For settlement speed, Solana wins decisively; for authorization latency, they are comparable.

Why are Solana fees so much lower than Ethereum's?

Two design choices dominate. Solana executes non-conflicting transactions in parallel across many CPU cores, while Ethereum's virtual machine executes sequentially, so Solana's blockspace supply is vastly larger — and fees are fundamentally a price on scarce blockspace. Second, Solana's local fee markets confine congestion pricing to the specific accounts under contention, whereas Ethereum's gas market historically repriced the entire chain when any application got busy. Ethereum has moved activity to cheaper Layer 2 rollups, which narrows the gap but adds bridging steps and their own trust assumptions.

Can Solana fees ever become expensive?

The base fee — 5,000 lamports, well under a tenth of a euro cent — has been stable for years. Priority fees, however, float with demand for specific accounts. During extreme events, users bidding for contested state (a token launch, a liquidation) have paid several euros to win placement. A plain payment between uncontested accounts stays cheap even then, because local fee markets isolate the congestion. The realistic worst case for a normal transfer is a few cents; the routine case is under a tenth of a cent.

What happens to my payment if Solana has an outage?

During a full outage, no transactions settle until validators restart the network — historical outages lasted from a few hours to nearly a day. A transaction submitted during downtime simply fails; funds do not vanish in transit, because a transfer either executes atomically or not at all. Regulated institutions mitigate this by keeping alternative rails available. At NGIBANK, classical SEPA and SWIFT connectivity exists alongside Solana settlement, so a network incident delays convenience rather than freezing access to money.

Do I need SOL to pay fees when my bank uses Solana?

If you hold your own wallet, yes — fees are paid in SOL, so you need a small balance even to move USDC or other tokens, a classic beginner stumbling block. Through a bank or regulated custodian, no: the institution executes transactions on your behalf, pays lamport-denominated fees itself and absorbs or prices them into its service. Given that a transfer costs a fraction of a cent, that cost is negligible for the institution — one reason on-chain settlement supports business models that card interchange never could.