Ethereum and Solana shipped competing Layer 1 consensus upgrades in October 2026. The results diverge sharply. Solana activated Alpenglow on mainnet October 21, replacing TowerBFT with Votor and cutting observed finality from 12.8 seconds to a median of 404 milliseconds in the first 48-hour sampl...
"We already found 13 real-world examples." — ChainSecurity, Smart Contract Security Auditor (October 13, 2026 disclosure)
Ethereum and Solana shipped competing Layer 1 consensus upgrades in October 2026. The results diverge sharply. Solana activated Alpenglow on mainnet October 21, replacing TowerBFT with Votor and cutting observed finality from 12.8 seconds to a median of 404 milliseconds in the first 48-hour sample — a 97% reduction. Ethereum cancelled Glamsterdam's Hoodi testnet fork on October 22, three days before the scheduled October 26 activation, after ChainSecurity flagged 13 deployed contracts vulnerable to gas-accounting changes in EIP-8037 and EIP-8038. No replacement testnet date has been set. The December mainnet target is now uncertain.
The contrast is structural, not accidental. Solana's monolithic architecture allowed a single consensus swap across one execution environment. Ethereum's modular design, with 19 EIPs bundled into one fork, exposed cross-EIP interaction risks that required halting the pipeline. Both chains are pursuing the same goal — sub-second finality and higher throughput — through fundamentally different engineering strategies. The data from October suggests those strategies carry different execution risk profiles.
Solana activated Alpenglow at slot 459,653,000 on October 21, 2026, during epoch 1064. The upgrade replaced TowerBFT — the chain's consensus mechanism since its 2020 genesis — with two new subsystems: Votor (voting and finalization) and Rotor (block propagation).
Votor runs a dual-path finalization model. When a block collects votes from validators representing at least 80% of staked SOL in the first round, it finalizes via the "fast path" at approximately 100 milliseconds. When participation lands between 60% and 80%, a second voting round triggers, targeting finality at approximately 150 milliseconds. Test cluster data collected through August 2026 showed 96% of blocks finalizing via the fast path at a median of 54 milliseconds.
Rotor replaced Turbine, the previous block-propagation layer. It uses a one-hop dissemination strategy that reaches validators representing 80% of stake in roughly 150 relay transmissions and 2 milliseconds of dispersal time, with full data delivery completing in approximately 18 milliseconds.
The activation proceeded without a cluster outage. A total of 446 validators participated in the genesis vote. One immediate side effect: vote transactions, previously recorded on-chain, moved off-chain under Votor. In the last TowerBFT block, approximately three in five transactions were validator votes. In the first Alpenglow block, there were none. This eliminated roughly 75% of reported on-chain transactions, causing headline TPS figures on block explorers to drop from approximately 5,165 tx/s to approximately 2,083 tx/s — entirely an accounting change, with no decrease in user activity.
Early mainnet performance: The Solana Foundation's Alpenglow dashboard recorded 202 milliseconds of observed finality on day one. A 24-hour sample ending October 23 at 12:30 UTC showed median finality of 404 milliseconds — approximately 2.7x above the stated 150ms target. Geographic dispersion of validators appears to account for much of the gap between test-cluster results and mainnet observations.
Ethereum's Glamsterdam fork bundles two headline EIPs — EIP-7732 (Enshrined Proposer-Builder Separation, or ePBS) and EIP-7928 (Block-Level Access Lists, or BALs) — alongside 17 additional EIPs, for 19 total. The upgrade activated successfully on the Sepolia testnet on October 6, 2026, at epoch 353,024.
The next step was Hoodi testnet activation, scheduled for October 26 at 17:42:48 UTC, with the gas limit set to rise to 200 million. On October 22, four days before the scheduled fork, Ethereum core developers cancelled it.
The cause: a ChainSecurity disclosure dated October 13 identified "insufficient gas griefing" risks in EIP-8037 and EIP-8038. EIP-8037 introduces a "state-gas reservoir" — a separate gas budget for state-creation costs that operates independently of execution gas. ChainSecurity calculated that creating a new storage slot could require 110,020 gas when the separate allowance is exhausted, compared with 22,100 under current rules. Ethereum researcher Toni Wahrstätter confirmed that the change could cause transactions to revert unexpectedly or enable execution paths that contract developers had assumed were unreachable.
ChainSecurity identified 13 deployed contracts exhibiting changed behavior under the new gas-accounting rules and initiated responsible disclosure with affected projects. The Ethereum Foundation confirmed the decision to delay on October 22, stating developers needed more time to investigate interactions between EIP-8037 and contracts already on mainnet.
Timeline erosion: Glamsterdam's mainnet target has shifted three times in 2026 — from early 2026, to Q3, to Q4, to a tentative early December window. The Hoodi cancellation puts even December at risk. No replacement testnet date has been announced. Any engineering fix to EIP-8037 and EIP-8038 must complete before a new Hoodi date can be set; only after a successful Hoodi run can a mainnet date be selected.
The divergence in outcomes reflects a deeper architectural split.
Solana: Single-layer consensus swap. Alpenglow replaced one consensus mechanism (TowerBFT) with another (Votor/Rotor) across a single execution environment. The scope of the change was large — it is the most significant consensus overhaul since genesis — but the interaction surface was bounded. Vote transactions moved off-chain. Block propagation changed topology. Finality rules changed. All within one layer.
Ethereum: Multi-EIP bundled fork across a modular stack. Glamsterdam packages 19 EIPs into a single activation event. The two headline EIPs (ePBS and BALs) restructure block construction and execution. Additional EIPs, including EIP-8037 and EIP-8038, change gas accounting rules. The interaction surface between these EIPs and the existing contract ecosystem created an attack vector that testing on Sepolia did not surface. It was only caught through third-party security audit.
| Dimension | Solana (Alpenglow) | Ethereum (Glamsterdam) | |---|---|---| | Scope | 1 consensus mechanism replacement | 19 EIPs bundled | | Testnet phases | Devnet → Mainnet | Sepolia → Hoodi → Mainnet | | Testnet result | Clean activation | Sepolia passed; Hoodi cancelled | | Mainnet status | Live (Oct 21) | Not scheduled | | External audit halt | No | Yes (ChainSecurity) | | Breaking contract risk | Minimal (vote tx removal) | 13 contracts identified |
Finality:
| Metric | Solana (Pre-Alpenglow) | Solana (Post-Alpenglow) | Ethereum (Current) | Ethereum (Post-Glamsterdam Target) | |---|---|---|---|---| | L1 finality | ~12.8 sec | 404ms median observed | ~12 sec block time; ~15 min epoch finality | No change to slot time; ePBS restructures block pipeline | | Target finality | — | 150ms | — | Not directly addressed by Glamsterdam |
Glamsterdam does not directly target finality reduction on Ethereum L1. Its throughput gains come from parallel execution (BALs) and gas limit increases, not from consensus speed changes. Ethereum's finality roadmap is addressed by separate proposals (single-slot finality) not included in Glamsterdam.
Throughput:
| Metric | Solana (Current) | Ethereum (Current) | Ethereum (Post-Glamsterdam Target) | |---|---|---|---| | Gas limit | N/A | 60M | 200M (progressive increase) | | Effective TPS | ~2,083 user tx/s | ~30-100 tx/s (L1) | ~10,000 tx/s (L1 target) | | Architecture | Monolithic | Modular (L1 + L2s) |
The 10,000 TPS target for Ethereum L1 post-Glamsterdam would represent a 100-330x increase from current throughput. This is contingent on BALs enabling parallel execution and the gas limit reaching 200 million — neither of which can activate until Glamsterdam ships.
Alpenglow materially changes Solana's validator cost structure. By moving vote transactions off-chain, validators no longer pay transaction fees to record votes. According to estimates cited by Solana infrastructure providers, validator break-even stake requirements fall from approximately 4,850 SOL (~$800,000) to roughly 450 SOL (~$75,000) — a 90% reduction. This could broaden the validator set from its current approximately 773 active validators.
On the Ethereum side, Glamsterdam's ePBS shifts the block-production pipeline. Currently, approximately 90% of Ethereum blocks are built through third-party relays (MEV-Boost). ePBS moves the proposer-builder relationship into the consensus layer, eliminating relay dependency. Estimates suggest MEV extraction could decline by up to 70% once ePBS activates. However, these economics remain theoretical until Glamsterdam ships.
For Solana developers: The vote-transaction removal changes how TPS is reported on explorers and analytics dashboards. Applications that tracked or referenced on-chain vote transactions need to update their integrations. The finality improvement from 12.8 seconds to sub-second has direct implications for DeFi protocols, payment applications, and any system that depends on settlement confirmation timing.
For Ethereum developers: The 13 contracts flagged by ChainSecurity represent direct risk. Any contract that makes assumptions about gas remaining during execution — particularly around state-creation costs — may behave differently under EIP-8037's state-gas reservoir model. Developers should audit contracts for gas-sensitive logic before Glamsterdam activates. The delay provides additional time for this review but also extends the period during which Ethereum L1 operates at current throughput levels (30-100 TPS at 60M gas limit).
For users of both chains: Solana users benefit immediately from faster finality. Ethereum users see no change until Glamsterdam ships. The Ethereum L2 ecosystem continues to provide soft finality in under one second on leading rollups (Base, Arbitrum, Optimism), partially offsetting L1 limitations — though hard finality (settlement to L1) still requires minutes.
October 2026 produced a natural experiment in Layer 1 upgrade strategy. Solana shipped a single, large consensus change across a monolithic architecture and activated it on mainnet within one epoch. Ethereum attempted to bundle 19 improvements into one fork across a modular stack and was stopped by a third-party audit finding that two of those improvements interacted dangerously with deployed contracts.
Neither outcome proves one architecture superior. Solana's monolithic approach carries concentration risk — a single-layer failure has no fallback. Ethereum's modular approach distributes risk across layers but creates a larger interaction surface for upgrades. The data from October shows only that Solana's approach produced a faster deployment cycle for this particular upgrade, while Ethereum's approach caught a contract-compatibility risk before it reached mainnet — which is, by design, what testnets and audits are supposed to do.
The economic implications depend on timeline. Every month Glamsterdam is delayed, Ethereum L1 continues operating at 30-100 TPS with a 60M gas limit while competing chains scale their base layers. Solana's sub-second finality is live now. Whether the 404ms observed median converges toward the 150ms target over the coming weeks will determine whether Alpenglow delivers on its design specification or stabilizes at a different performance tier.