TL;DR
• MEV on Solana hit record volumes in early 2026. Jito-routed bundles are now responsible for 24%+ of validator non-base-fee revenue, with $480M+ in cumulative searcher profits tracked through Q2.
• Alpenglow consensus (150ms finality) and the Jito BAM (Block Assembly Marketplace) launch fundamentally changed the MEV landscape. Strategies that worked at 400ms slots no longer work the same way.
• Solana has no mempool. The contest is microsecond-level latency from searcher to slot leader. Strategy quality is now secondary to infrastructure quality for most active categories.
• Public RPC endpoints are not viable for production MEV. Median slot lag on public mainnet-beta runs 3–4 slots behind tip; competitive searchers need sub-1-slot lag and 20ms p99 read latency.
• RPC Fast benchmarks (Q2 2026): 0.3 median slot lag, 18ms getAccountInfo p99, 12ms Yellowstone gRPC arrival, 96.4% Jito bundle landing rate.
There are moments in blockchain where speed is more than an advantage—it is the difference between profit and loss. That is what Maximal Extractable Value (MEV) is about. The category started as a curiosity, became a serious revenue source for sophisticated traders, and in 2026 has matured into one of the largest non-base-fee revenue streams on Solana. The numbers behind it have grown to a point that makes the underlying mechanics worth understanding clearly, not just as a technical exercise but because the infrastructure choices around MEV now dominate how the rest of the on-chain trading ecosystem operates.
On Solana, the concept of MEV takes on a different shape than it does on Ethereum, driven by parallel execution, deterministic leader scheduling, and the absence of a public mempool. Where Ethereum MEV is largely a gas auction visible to anyone watching pending transactions, Solana MEV is a latency contest invisible to most observers. The race happens in microseconds between the moment a state change is published and the moment a competing transaction lands in the next slot.
MEV on Solana vs Ethereum: the architectural split
The two networks handle transaction ordering and inclusion fundamentally differently. Those differences shape what MEV looks like on each chain, what strategies work, and what infrastructure searchers need.
How Solana MEV actually works in 2026
Parallel execution under Sealevel
Sealevel processes transactions in parallel as long as they touch disjoint sets of accounts. Each transaction must declare upfront which accounts it will read or write, and any transactions with no overlap can run simultaneously across CPU cores. The practical implication for MEV is that a single transaction can contain multiple swap instructions across different DEXes, executed atomically—buy on Raydium, sell on Orca, with both legs settling or both reverting. This is fundamentally different from Ethereum, where executing two trades across two DEXes requires careful sequencing and exposes the searcher to state changes between steps.
With Firedancer now running across a meaningful share of mainnet validators, the effective throughput ceiling on parallel execution has expanded substantially. Solana's actual sustained throughput in mainstream conditions is now in the 3,000–5,000 TPS range for genuine user transactions (excluding vote and consensus overhead, which Alpenglow largely removed from the on-chain footprint).
Flash loans and atomic execution
Flash loans on Solana work the same way they do on Ethereum in principle—borrow within a transaction, use the funds, repay before the transaction ends, or the whole thing reverts. The difference is the unit economics. At sub-cent transaction costs and 400ms slot times, flash-loan-based arbitrage works at trade sizes that would not be profitable on Ethereum mainnet. The classic triangular pattern still applies:
- Borrow stablecoin liquidity from a flash loan provider
- Execute a price-correction swap on one DEX
- Execute the reverse swap on another DEX at the better price
- Repay the loan with interest and capture the spread
All four steps execute in a single atomic transaction. If any step fails (slippage, pool state change, insufficient liquidity), the whole transaction reverts and only the compute budget fee is lost.
Transaction ordering and Jito bundles
Without a public mempool, transaction ordering on Solana is determined by what reaches the slot leader first. This is where Jito's infrastructure became dominant. Jito-Solana, the modified validator client, exposes a Block Engine that accepts atomic bundles from searchers with explicit SOL tips attached. The block engine forwards prioritized bundles to leader validators, who include them in the order specified by the searcher.
In 2026, more than 78% of mainnet stake runs the Jito-Solana client, and Jito bundles account for the substantial majority of intentional MEV extraction on the network. The launch of Jito BAM (Block Assembly Marketplace) in late 2025 added a more sophisticated bidding layer, allowing searchers to compete for specific block positions with finer granularity than the original bundle auction.
Stop losing MEV to slow RPC
RPC Fast gives Solana searchers production-grade infrastructure on a SaaS plan—Yellowstone gRPC, Jito bundle routing, and sub-50ms slot lag. Start a free trial, no validator setup required.
The current MEV landscape on Solana
MEV activity on Solana has grown roughly in line with the network's overall expansion. Aggregate searcher profits captured through Jito bundles passed $480 million cumulative as of Q2 2026, with monthly run-rates routinely between $30M and $50M. The categories driving this volume have shifted somewhat from the early days when arbitrage dominated nearly all the value.
Where MEV value comes from in 2026
Five strategy categories produce the bulk of extracted value on Solana right now:
- DEX arbitrage—price discrepancies between Raydium, Orca, Meteora, Phoenix, and Lifinity pools. Still the largest category by transaction count, though margins have compressed significantly.
- Liquidations on lending markets—Kamino, MarginFi, Save, and other Solana lending protocols. Liquidations spike during volatility and represent some of the highest single-transaction value events.
- Backrunning large trades—capturing the price rebound after meaningful swaps move pool state. The shorter Alpenglow finality has compressed the window for this strategy but also made it more competitive.
- Just-in-time (JIT) liquidity—adding concentrated liquidity around predicted swaps and capturing fee revenue. Increasingly automated and infrastructure-intensive.
- Memecoin sniping and launchpad arbitrage—fast-execution strategies around pump.fun graduations and token launches. Volatile category that produced very large profits in 2024–2025 cycles.
The failure rate problem
Even with infrastructure investment, MEV attempts on Solana fail at high rates. Across the public bot population, more than 55% of submitted MEV transactions either revert on-chain (state changed between detection and execution) or fail to land in the targeted slot (lost the latency race). For high-frequency searchers running thousands of attempts per day, failure rates determine profitability more than strategy logic does.
Reducing this failure rate is largely an infrastructure problem rather than a strategy problem. The components that matter:
- Sub-slot state freshness from gRPC streams instead of polling-based WebSocket
- Direct Jito Block Engine submission with multi-region parallel routing
- Accurate priority fee and tip calibration based on real-time network conditions
- Pre-flight simulation with current account state before bundle submission
- Fast retry paths when a bundle is rejected, before the opportunity disappears
RPC Fast benchmarks: what the infrastructure tier actually delivers
Talking about latency in the abstract is easy. The numbers below are what RPC Fast measures internally on a rolling 30-day basis, compared against the default mainnet-beta public endpoint and a generic multi-chain RPC provider on a representative shared tier.
The gap is not subtle. The public endpoint trails the network tip by an average of 3.4 slots, which is enough to make most MEV strategies non-viable regardless of how good the logic is. A generic multi-chain provider on a shared tier does meaningfully better but still leaves substantial latency on the table. RPC Fast's dedicated Solana-only infrastructure runs an order of magnitude tighter on the metrics that determine whether MEV extraction succeeds.
The infrastructure tier comparison
In practice, the choice between infrastructure tiers maps onto specific operational outcomes:
Real-world deployment example
A recent infrastructure migration with Magneta Labs (a Solana arbitrage and JIT liquidity operation) illustrates what the benchmark gap translates to in production. Before moving onto RPC Fast's dedicated nodes, Magneta was running on a mid-tier shared RPC plan from a multi-chain provider. Their bundle landing rate sat around 84%, transaction submission p99 latency was around 95ms, and their backrun strategy on large Orca and Raydium swaps was missing roughly one in five opportunities to faster competitors.
After the migration to dedicated RPC Fast infrastructure with Yellowstone gRPC and parallel Jito BAM submission, the same trading logic—unchanged at the strategy layer—produced measurable improvements across the metrics that mattered:
- Bundle landing rate improved from 84% to 96.1%
- End-to-end submission latency dropped from 95ms p99 to 22ms p99
- Backrun capture rate on large Orca/Raydium swaps roughly doubled
- Failed transaction count dropped by 71% over the first month of operation
The change was not in the trading logic. It was in how fast the trading logic could see the market and how reliably it could land transactions when it acted. That distinction—strategy unchanged, infrastructure upgraded—is where most teams running mid-tier setups can find significant performance left on the table.
Where Solana MEV goes from here
Two structural trends are worth watching. The first is BAM maturity. Jito's Block Assembly Marketplace is still relatively new, and the bid mechanics around block position will likely keep evolving. Searchers who develop sophisticated tip calibration and BAM-aware bundle construction will pull ahead of those running older bundle submission patterns.
The second is the Alpenglow effect on MEV strategy design. 150ms finality compresses the window for many strategies and changes the risk profile of others. Strategies that depended on slot reorganization or longer confirmation windows have largely disappeared. New strategies that exploit the predictability of fast finality are emerging. The shape of MEV on Solana in late 2026 will likely look meaningfully different from what it looked like at the start of the year.
What does not change is the underlying dynamic: in a market where the contest is microsecond latency to slot leaders, the infrastructure stack between the searcher and the validator determines who captures value and who does not. Strategy advantages compress as more participants converge on similar logic. Infrastructure advantages compound because they affect every single transaction the system processes.
Questions about your setup?
Our team helps Solana MEV teams every day. Drop into the Telegram chat—we'll review your current stack and tell you where the latency is leaking.
.jpg)
.jpg){kind=tracking}
