Ethereum and Solana are both executing protocol-level consensus rewrites in October 2026. Ethereum's Glamsterdam upgrade hits its Sepolia testnet fork on October 6, introducing enshrined Proposer-Builder Separation (ePBS) and Block-Level Access Lists (BALs) ahead of a Q4 mainnet target. Solana's ...
"What Glamsterdam does is remove the technical reasons [the gas limit] couldn't safely go higher." — Ethereum Foundation, Glamsterdam Testnet Announcement, September 17, 2026
Ethereum and Solana are both executing protocol-level consensus rewrites in October 2026. Ethereum's Glamsterdam upgrade hits its Sepolia testnet fork on October 6, introducing enshrined Proposer-Builder Separation (ePBS) and Block-Level Access Lists (BALs) ahead of a Q4 mainnet target. Solana's Alpenglow upgrade, which activated on testnet September 24 and devnet September 25, replaces TowerBFT with the Votor/Rotor consensus stack and targets Agave 4.3 mainnet activation in October.
The two upgrades attack different bottlenecks. Glamsterdam restructures how blocks are built and executed, aiming to triple Ethereum L1 capacity from roughly 60 million to 200 million gas per block while enshrining MEV protections at the protocol level. Alpenglow restructures how blocks are finalized and propagated, targeting a reduction in finality from 12.8 seconds to approximately 150 milliseconds while eliminating on-chain vote transactions that consume roughly 75% of Solana's current block space. Neither upgrade changes each chain's fundamental architecture — Ethereum remains modular, Solana remains monolithic — but both represent the most consequential protocol changes either network has shipped since the Merge (September 2022) and the original TowerBFT deployment, respectively.
Glamsterdam merges two upgrade scopes — the Amsterdam execution-layer fork and the Gloas consensus-layer fork — into a single coordinated activation. The Ethereum Foundation confirmed the Sepolia testnet fork at epoch 353,024, slot 11,296,768 (October 6, 2026, 13:53:36 UTC), with a provisional Hoodi testnet fork on October 27 and mainnet activation targeted for Q4 2026.
Two EIPs define the upgrade:
EIP-7732 (Enshrined Proposer-Builder Separation). Currently, approximately 90% of Ethereum blocks route through MEV-Boost, an off-protocol relay system. According to Relayscan data from 2026, the top two block builders produce over 80% of Ethereum blocks, and OFAC-compliant relays handle 42.9% of MEV-Boost payloads as of August 2026. EIP-7732 moves builder-proposer coordination into the consensus protocol itself. Builders assemble and cryptographically seal block contents; proposers select the highest-paying block without seeing its contents until after finalization. This reduces reliance on external relays — several of which have shut down due to lack of economic incentive — and aims to mitigate censorship and manipulation vectors inherent in the current off-protocol MEV supply chain.
EIP-7928 (Block-Level Access Lists). BALs declare which accounts and storage locations a block will access before execution begins. This enables clients to read state from disk in parallel and validate transactions concurrently rather than sequentially. The practical effect: node operators can process blocks faster, supporting the upgrade's gas limit increase from roughly 60 million to a 200 million floor — a 3.3x capacity expansion. To manage the state growth implications of higher throughput, EIP-8037 introduces a fixed cost per state byte, targeting a state growth rate of approximately 120 GiB per year to keep consumer hardware viable for full nodes.
Node operators must update both execution and consensus layer clients before activation. Confirmed compatible client versions include Lodestar 1.49.0, Prysm 7.2.0, Teku 26.9.1 on consensus, and Besu 26.9.0, Erigon 3.7.0, go-ethereum 1.17.6, Nethermind 2.0.0, and Reth 2.7.0 on execution.
Alpenglow replaces Solana's TowerBFT consensus mechanism — the system that has governed block finality since mainnet launch — with a two-component architecture:
Votor replaces TowerBFT's on-chain vote transaction model with off-chain BLS-aggregated voting. Under TowerBFT, vote transactions account for approximately 75% of all on-chain transactions. On September 24, 2026, Solana mainnet recorded 217.1 million vote transactions versus 168.8 million non-vote transactions. Each slot currently carries roughly 500KB of vote data. Under Votor, only the aggregated certificate — approximately 1,000 bytes — lands on-chain. When 80% or more of validator stake participates, finality is achieved in a single round (~100ms). At 60% stake participation, two rounds are needed, targeting ~150ms. The upgrade received 98.27% validator approval in September 2025.
Rotor will eventually replace Turbine as the block propagation layer, using stake-weighted relays and erasure coding to eliminate bandwidth bottlenecks. Rotor is not included in the initial Alpenglow activation and will follow in a subsequent release.
The upgrade activated on Solana's public testnet at slot 444,625,255 on September 24 and on devnet at slot 504,148,999 on September 25. Early results exceeded targets: testnet recorded a 54ms median finalization time and demonstrated more than 25,000 transactions per second with 150ms finality. Devnet finality registered at 322ms on September 26, with vote transactions absent from blocks — confirming the vote elimination mechanism is functional.
Mainnet activation targets Agave 4.3, expected in October 2026. Anza's release note stated explicitly that mainnet-beta activation would occur only after an observation period on test clusters, with no firm date committed.
| Dimension | Ethereum (Glamsterdam) | Solana (Alpenglow) | |---|---|---| | Primary target | Block construction and execution | Block finalization and propagation | | Current bottleneck addressed | Off-protocol MEV relay centralization; sequential transaction execution | 12.8-second finality; vote transaction overhead (75% of blocks) | | Finality change | No direct change (remains ~13 minutes for full finality) | 12.8 seconds → ~150ms (target); 54ms observed on testnet | | Throughput change | Gas limit floor: 60M → 200M (3.3x); targets ~10,000 TPS | Frees ~75% of block space currently used by votes | | Consensus mechanism | ePBS enshrined at protocol level | TowerBFT replaced by Votor | | Execution model | Parallel state reads via BALs | No change to Sealevel runtime | | Architecture | Modular (L1 + L2 ecosystem) | Monolithic (single-layer) | | Sepolia/testnet date | October 6, 2026 | September 24, 2026 (activated) | | Mainnet target | Q4 2026 (no specific date) | October 2026 via Agave 4.3 (no specific date) |
The upgrades are complementary in what they reveal about each chain's priorities. Ethereum is solving for decentralized block production at scale — the ePBS mechanism directly addresses the concentration among builders and the relay dependency that has produced censorship risks. Solana is solving for settlement speed — a finality reduction of nearly two orders of magnitude serves payment, exchange, and institutional settlement use cases where seconds matter.
Neither upgrade changes the fundamental architectural divide. Ethereum's throughput improvements apply to L1 but the network's scaling strategy remains dependent on Layer 2 rollups for consumer-facing volume. Solana's finality improvements reinforce its position as a single-layer execution environment where applications interact directly with the base chain.
Ethereum's Glamsterdam carries the heavier testing burden. The Ethereum Foundation's blog post on September 17 noted "critical bugs" discovered during devnet testing that required patching before the Sepolia fork could be scheduled. The upgrade was originally positioned for H1 2026, then pushed to Q3, and now targets Q4. This delay pattern is consistent with previous Ethereum upgrades — the Merge was delayed multiple times across 2021-2022 before shipping in September 2022.
Solana's testnet performance has exceeded stated targets. The 54ms median finalization time on testnet outperformed the 100-150ms specification. However, testnet conditions differ materially from mainnet: Solana mainnet runs 906 active validators handling real economic value, and the transition from TowerBFT to Votor on a live network with $5.7-12 billion in TVL carries execution risk that testnet results cannot fully capture.
Both upgrades face the same fundamental risk: coordinated client updates across decentralized validator sets. Ethereum's multi-client architecture requires simultaneous updates across five execution clients and three consensus clients. Solana's single-client dependency (Agave, formerly the Agave fork of the original Labs validator) simplifies coordination but concentrates implementation risk.
Ethereum validators stand to see MEV dynamics shift. Currently, MEV-Boost adds 10-30% to validator base rewards. ePBS changes the settlement mechanism — builders still compete, but proposers no longer need external relays to mediate. The elimination of relay dependency removes an unpaid infrastructure layer that has seen multiple operators (including Blocknative) shut down, concentrating the market among Ultra Sound (26.8% market share), Titan (22.3%), and bloXroute (20.7% and 20.3% across two products).
Solana validators see a direct cost reduction. Vote transaction elimination saves approximately 1 SOL per day in fees, replaced by a Validator Admission Ticket burn of approximately 0.8 SOL per day — a net savings of $35,000-$50,000 per year per validator at current SOL prices. The freed block space from eliminating vote transactions represents a capacity increase for revenue-generating user transactions.
The upgrades land against contrasting network fundamentals:
Market capitalization: Ethereum $228.5 billion; Solana $44.5 billion (as of Q3 2026).
Total value locked: Ethereum $54 billion on L1, approximately $123 billion including L2s; Solana $5.7-12 billion depending on measurement methodology.
Stablecoin value on-chain: Ethereum $165.3 billion; Solana $15.0 billion.
Developer activity: Ethereum maintains 37% of professional blockchain developer share (31,869 active developers as of June 2026, down from 82% in 2020). Solana has grown to 23% (17,708 active developers), and attracted more new developers than Ethereum in 2025 (~4,100 vs. ~3,700).
Transaction throughput: Ethereum L1 processes 15-30 TPS; Solana processes 5,000-5,500 TPS (non-vote). With Glamsterdam, Ethereum targets ~10,000 TPS on L1. With Alpenglow, Solana's effective non-vote capacity increases as 75% of current block space is freed.
These numbers frame the stakes differently for each chain. Ethereum's upgrade is about maintaining relevance of L1 as a settlement layer — without higher throughput and enshrined MEV protections, value increasingly migrates to L2s where Ethereum captures less direct fee revenue. Solana's upgrade is about meeting institutional-grade settlement requirements — sub-second finality is a prerequisite for exchange settlement, payment finality, and the high-frequency DeFi applications that constitute Solana's competitive differentiation.
October 2026 marks the first time both leading smart contract platforms are simultaneously rewriting core protocol infrastructure. The parallel timing is coincidental but the strategic logic is not: both chains face scaling constraints that incremental parameter changes cannot resolve, and both are choosing protocol-level surgery over patch-level fixes.
The upgrades reflect divergent theories of what limits blockchain adoption. Ethereum's Glamsterdam argues the bottleneck is block construction and MEV governance — that the chain's throughput ceiling is artificially low because off-protocol infrastructure has captured too much of the block production pipeline. Solana's Alpenglow argues the bottleneck is settlement finality — that 12.8 seconds is too slow for the payment, exchange, and institutional use cases that would drive the next wave of on-chain economic activity.
Both may be correct about their own chains. The data will follow on the other side of mainnet activation.