Ethereum and Solana, the two largest programmable blockchains by total value locked ($38.9 billion and $5.1 billion respectively as of July 2026), are both entering final testing for their most significant protocol upgrades in years. Ethereum's Glamsterdam hard fork — shipping ten EIPs including ...
"We're working on devnets with all the EIPs in them right now. This is the last phase before we work on hardening and then shipping the testnets." — Parithosh Jayanthi, Core Developer & DevOps Engineer, Ethereum Foundation
Ethereum and Solana, the two largest programmable blockchains by total value locked ($38.9 billion and $5.1 billion respectively as of July 2026), are both entering final testing for their most significant protocol upgrades in years. Ethereum's Glamsterdam hard fork — shipping ten EIPs including enshrined Proposer-Builder Separation (ePBS) and Block-Level Access Lists — targets a 200-million gas limit floor and mainnet activation between September and December 2026. Solana's Alpenglow consensus rewrite, which passed governance with 98.27% validator approval, replaces the five-year-old Tower BFT mechanism with Votor/Rotor and targets 100–150 millisecond finality, down from 12.8 seconds.
The upgrades address fundamentally different bottlenecks — Ethereum is rearchitecting block construction and enabling parallel execution on L1, while Solana is rewriting consensus to close a finality gap that has limited institutional adoption. Both face multi-client parity challenges, and both carry execution risk that could push mainnet activation into Q4 2026. The combined TVL exposed to these protocol transitions exceeds $44 billion.
Glamsterdam is Ethereum's next hard fork after Fusaka, tracked under Meta EIP-7773. The upgrade ships ten Ethereum Improvement Proposals and targets mainnet activation in H2 2026, with current estimates placing the window between September and December based on prior fork cadences of two to four months of public testnet seasoning.
The upgrade's scope was outlined by Vitalik Buterin in late February 2026 across eight core proposals, later expanded to ten. Two headline EIPs do the structural work: EIP-7732 (enshrined Proposer-Builder Separation) and EIP-7928 (Block-Level Access Lists). The remaining eight handle gas repricing, state growth controls, and ancillary protocol changes.
As of early July 2026, devnets include all planned EIPs running across all client implementations. The Ethereum Foundation describes this as the final phase before code hardening and public testnet deployment.
Ethereum L1 currently processes 15–30 transactions per second. The Glamsterdam bundle is designed to raise that ceiling toward 10,000 TPS by tripling the gas limit and enabling parallel transaction processing — a fundamental shift from Ethereum's sequential execution model.
The headline feature separates block building from block proposing at the protocol level. Today, this separation exists through external relay infrastructure (primarily Flashbots' MEV-Boost), which approximately 90% of Ethereum blocks flow through. EIP-7732 moves this logic into the consensus layer itself.
The practical effect: the data propagation window expands from approximately 2 seconds to roughly 9 seconds. This gives the network more time to process larger blocks safely, which is a prerequisite for the gas limit increase.
The implementation is proving complex. ePBS splits block production into two parties acting in sequence inside consensus. The protocol must now handle disagreement or failure between proposers and builders — a coordination problem that touches practically every layer of the stack. Ethereum core developers have identified ePBS complexity as the primary bottleneck to testnet readiness, with three risks that could push activation into Q4 or later: ePBS implementation itself, cross-client parity, and gas repricing interdependencies.
Centralization risk is a live concern. By enshrining the builder role, the protocol formalizes a function currently dominated by a small number of block builders. Whether this reduces or entrenches concentration remains an open debate among researchers.
EIP-7928 introduces Block-Level Access Lists (BALs) — a per-block record mapping every account and storage slot the block touches, along with post-execution state values. In effect, BALs function as a pre-computed map of the block's data dependencies.
The payoff is parallelization. Because BALs provide instant visibility into which transactions do not overlap, nodes can safely group unrelated transactions and process them simultaneously rather than sequentially. This enables parallel disk reads, parallel transaction validation, and parallel state root computation.
According to Toni Wahrstätter, an Ethereum core developer, the average block size under BALs is approximately 35 KiB at 36 million gas, with worst-case sizes remaining below calldata worst-case sizes. The proposal has received broad support from core developers.
For Ethereum, this is an architectural pivot. The chain has processed transactions sequentially since genesis. Parallel execution is the mechanism that bridges the current 15–30 TPS to the 10,000 TPS target — without it, the gas limit increase alone would be insufficient.
Glamsterdam establishes a 200-million gas limit floor, up from the current ~60 million — a 3.3x increase. This is not automatic; it requires the ePBS propagation window and BAL parallelization to be working correctly.
State growth is the primary constraint. Without safeguards, a 3.3x throughput increase would bloat Ethereum's state database past sustainable levels. EIP-8037, finalized in May 2026, addresses this by setting a fixed cost per byte of new state and creating a separate gas reservoir for state growth. The design gives client teams a sustainability ceiling — the gas limit can rise to 200 million without exceeding a manageable yearly state growth budget.
The gas limit change directly affects L2 economics. Ethereum's L2 ecosystem — Arbitrum, Optimism, Base, and others — posts transaction data to L1. A higher gas limit means more L2 data can fit per block, potentially reducing L2 operating costs. This has downstream implications for L2 fee structures and competitive positioning.
Alpenglow is the largest consensus change in Solana's history, replacing Tower BFT — the protocol's consensus mechanism since mainnet launch in 2020 — with a new two-component system. The governance proposal (SIMD-0326) passed with 98.27% validator approval, and a community test cluster went live on May 11, 2026, confirmed by development firm Anza.
The core promise: transaction finality drops from approximately 12.8 seconds under Tower BFT to 100–150 milliseconds, a reduction of roughly 100x. Solana co-founder Anatoly Yakovenko confirmed at Consensus Miami on May 7, 2026 that mainnet activation could arrive as soon as Q3 2026, with Q4 as fallback if testnet issues emerge.
Three gates must clear before mainnet activation: SIMD-0387 (BLS key registration) landing on public testnet, security audits completing without major findings, and the Agave 4.1 client release reaching stable.
The 12.8-second finality has been a persistent friction point for institutional adoption. Traditional finance settlement systems operate on sub-second confirmation. Closing this gap is a prerequisite for Solana's ambitions in tokenized securities and high-frequency DeFi applications.
Alpenglow consists of two protocol components that replace separate functions of Tower BFT:
Votor replaces Tower BFT's incremental voting rounds. Under the old system, validators had to wait through multiple confirmation rounds, each adding latency. Votor uses a lightweight voting aggregation model: validators aggregate votes off-chain before submitting final confirmation, allowing blocks to finalize within one to two rounds. The protocol supports both optimistic single-round finality (when supermajority is reached quickly) and a two-round fallback.
Rotor replaces the multi-hop relay system for block propagation. Instead of variable-latency relays, Rotor introduces deterministic relay assignments based on validator stake. Validators with high stake and reliable bandwidth serve as core relay points. Simulation data shows block propagation completing in as little as 18 milliseconds under typical bandwidth conditions.
The combination addresses both consensus latency (Votor) and network latency (Rotor). The 100–150 millisecond finality target assumes both components operating under normal conditions.
A structural change to validator economics accompanies the upgrade: Alpenglow introduces a Validator Admission Ticket (VAT) — a fixed fee of 1.6 SOL per epoch replacing on-chain vote fees. This has generated debate about whether the fixed cost creates barriers for smaller validators, potentially concentrating stake among larger operators.
Both upgrades face a common challenge: ensuring multiple independent client implementations reach parity before mainnet activation.
Ethereum runs five consensus clients and five execution clients. Glamsterdam's ten EIPs must be implemented consistently across all of them. Cross-client parity is one of the three identified risks that could delay the upgrade. Historical precedent suggests this is manageable — Ethereum has shipped multiple hard forks across its multi-client architecture — but ePBS adds novel complexity because it fundamentally changes how clients coordinate block production.
Solana is earlier in its multi-client journey. Firedancer, Jump Crypto's independent validator client written in C/C++, went live on mainnet in December 2025 and now runs on over 20% of active validators. By Q2–Q3 2026, Solana targets 50% Firedancer stake, at which point the network becomes resilient against single-implementation failures. However, Firedancer has endorsed Alpenglow's design but has no production Votor implementation on any cluster. Multi-client parity for the new consensus protocol is a key variable for the mainnet timeline.
Bitcoin developer Jeff Garzik told The Defiant in May 2026 that achieving sub-second finality "is impossible without security compromises," highlighting skepticism about whether Alpenglow can deliver its latency targets without tradeoffs. Concerns about Rotor's reliance on Web2 network connections — where connection uncertainty "can seriously impact the fast finality sought by Alpenglow" — have also been raised by protocol researchers.
The combined TVL exposed to these transitions — $38.9 billion on Ethereum and $5.1 billion on Solana — represents real capital at risk during the upgrade process. Historical data from prior Ethereum hard forks shows temporary TVL dislocation around activation dates, though no major post-fork failures have occurred.
For Ethereum: The 200-million gas limit and parallel execution change the L1 value proposition. If Ethereum L1 can process 10,000 TPS, it partially erodes the case for certain L2 solutions, while simultaneously reducing their operating costs. DeFi protocols with high-frequency operations — lending liquidations, DEX arbitrage, oracle updates — could migrate execution back to L1 if costs drop sufficiently.
For Solana: Sub-200-millisecond finality positions the network for institutional use cases currently dominated by centralized infrastructure. Solana's RWA TVL crossed $3 billion for the first time in Q2 2026, with tokenized asset volume hitting a record $5.77 billion. Faster finality directly addresses a primary objection from traditional finance participants.
Fee economics shift under both upgrades. Ethereum transactions currently cost $0.10–$0.30; higher throughput should compress fees. Solana transactions average $0.00025; Alpenglow's primary impact is on settlement assurance rather than cost. The VAT system changes validator revenue composition from variable vote fees to fixed epoch costs.
The two largest programmable blockchains are simultaneously rearchitecting their core infrastructure on overlapping timelines. Ethereum is making its L1 faster and more parallel — a concession that pure modular scaling has limits. Solana is rewriting consensus from scratch — an acknowledgment that 12.8-second finality is untenable for the institutional market it targets.
Both upgrades carry material execution risk. Ethereum's ePBS introduces two-party coordination into consensus for the first time; Solana is deploying a consensus protocol that has never operated at mainnet scale. The testing phases will determine whether Q3 or Q4 2026 becomes the activation window.
For the $44 billion in on-chain capital secured by these networks, the next three to six months represent the highest-stakes protocol engineering period since Ethereum's merge to proof of stake in September 2022.