Reading time: 4 min
On Base and Optimism, automated arbitrage programmes now consume more than 50% of blockspace while paying less than 10% of fees. Flashbots calls it the dominant limit to scaling blockchains. The proposed fix: replace the gas auction with privacy-preserving order flow and explicit MEV bidding.
The Numbers
Research from Flashbots, the Ethereum-focused MEV research and development group, has documented a structural problem that threatens to neutralise the scaling progress Layer 2 networks were designed to deliver. On major OP-Stack rollups — Base, Optimism, Unichain and World — MEV-related search spam now consumes more than 50% of total gas while contributing less than 10% of transaction fees. An academic study from ETH Zurich, analysing on-chain data across Base, Optimism and Arbitrum in Q1 2025, confirmed the pattern: cyclic arbitrage contracts accounted for over 50% of gas on Base and Optimism, yet paid less than one-quarter of total fees. Arbitrum, with its different sequencer design, showed just 7% gas consumption from the same activity.
The scale of waste is striking. A single successful arbitrage on Base documented by Flashbots required approximately 132 million gas in failed attempts — equivalent to nearly four full Ethereum blocks — before producing a profit of $0.12 on a $0.02 fee. Between November 2024 and February 2025, Base added 11 million gas per second in throughput — roughly three Ethereum Mainnets’ worth of capacity — and nearly all of it was absorbed by spam bots. On Base specifically, just two bot operators account for more than 80% of all spam transactions. The problem extends beyond Ethereum: on Solana, MEV bots now occupy approximately 40% of total blockspace.
Why the Bots Win
The MEV spam problem is structural, not technical. Bots flood L2 blocks with speculative transactions because they cannot see the private mempool — so they probe on-chain, submitting hundreds of transactions for each low-margin arbitrage opportunity, hoping one will succeed. Because L2 gas is cheap relative to Ethereum mainnet, the cost of failure is negligible for the bot but cumulative for the network: every failed transaction consumes compute resources, crowds out legitimate users and pushes baseline fees higher. The result is what Flashbots describes as a “spam auction” market structure, where economic congestion — not bandwidth — becomes the binding constraint on scalability.
The financial extraction is substantial. Alchemy, citing EigenPhi data, reported nearly $24 million in MEV profit extracted on Ethereum over a single 30-day window from December 8, 2025 to January 6, 2026. Sandwich attacks — where a bot inserts transactions before and after a user’s trade to capture the price movement — victimised over 33,000 users in March 2025 alone, orchestrated by just 101 entities. When institutional-scale DEX trades are visible in the mempool before execution, slippage from sandwich attacks can exceed gas costs by orders of magnitude. For retail users, the average experience in 2026 remains unchanged: every swap, balance check and approval is broadcast in plaintext, creating extractable value that bots are structurally incentivised to capture.
The Fix: Privacy as Infrastructure
Flashbots has proposed replacing the gas-based spam auction with a two-part design: programmable privacy using Trusted Execution Environments (TEEs), which allow searchers to view live state and plan trades without front-running users, and explicit off-chain bidding for transaction ordering rights in a dedicated auction. The model builds on Flashbots’ existing MEV-Share system, which operates as an order-flow auction redistributing approximately 90% of extracted value back to users rather than to bots. The Ethereum Foundation’s Privacy and Scaling Explorations team has separately standardised a three-part framework covering private writes (hiding transaction intent before execution), private reads (concealing which users and apps query balances and positions), and private proving (making zero-knowledge proofs cheap enough to embed everywhere).
The technology largely exists. Zero-knowledge proving costs collapsed over 2025 — Ethproofs’ year-end review documented a roughly fivefold reduction in latency and fifteenfold reduction in cost. Shutter is developing threshold encryption and timed key release integrated with proposer-builder separation. The bottleneck, according to Andy Guzman of the Ethereum Foundation’s PSE team, is no longer cryptography but coordination, developer tooling and the work of making privacy the default rather than opt-in. The industry built privacy as an exception in the last cycle. Whether it becomes invisible infrastructure — or remains a niche feature — depends on whether wallet and protocol developers treat the current plaintext default as a bug worth fixing. The economists, at least, have finally agreed it is one.
