Two builders control over 90% of Ethereum's block production. That single statistic captures one of the most urgent structural risks facing the $250 billion network: the infrastructure responsible for ordering every transaction on Ethereum has quietly consolidated into a duopoly. In a system desi...
"Every compromise of values that Ethereum has made up to this point — every moment where you might have been thinking, is it really worth diluting ourselves so much in the name of mainstream adoption — we are making that compromise no longer." — Vitalik Buterin, Co-founder of Ethereum
Two builders control over 90% of Ethereum's block production. That single statistic captures one of the most urgent structural risks facing the $250 billion network: the infrastructure responsible for ordering every transaction on Ethereum has quietly consolidated into a duopoly. In a system designed to be permissionless and censorship-resistant, the reality is that Titan and BuilderNet (Flashbots/Beaverbuild) decide what gets included in the vast majority of blocks — and in what order.
Ethereum's answer arrives this year in the form of Glamsterdam, the network's next major hard fork targeted for mid-2026. Its headline feature, EIP-7732 — Enshrined Proposer-Builder Separation (ePBS) — represents the most significant structural change to Ethereum's consensus layer since The Merge. By embedding the separation of block proposing and block building directly into the protocol, ePBS aims to dismantle the off-chain power structures that have made MEV extraction a centralized, opaque industry.
The stakes are enormous. Approximately $562 million in MEV was extracted on Ethereum in 2025 alone. Sandwich attacks, front-running, and back-running have become industrialized operations that extract value from ordinary users while concentrating power in the hands of sophisticated builders. Glamsterdam doesn't just patch the problem — it rewires how blocks are constructed at the deepest protocol level.
Ethereum's post-Merge block production was designed to be decentralized: over one million validators each have an equal shot at proposing blocks. But in practice, an off-chain middleware layer called MEV-Boost has created a shadow market where specialized builders assemble blocks and pay validators for the right to include them. The result is a three-tier hierarchy — searchers find MEV opportunities, builders assemble optimized blocks, and relays act as trusted intermediaries — none of which exist in the protocol itself.
As of January 2026, block builder market share has consolidated dramatically. Titan controls approximately 50.4% of Ethereum blocks, while BuilderNet (the Flashbots-led coalition that absorbed Beaverbuild) holds 27.1%. Quasar captures 14%, with all remaining builders splitting the scraps. This means two entities produce over 77% of all Ethereum blocks, and three produce over 91%.
This concentration has direct consequences for users. In 2025, sandwich attacks constituted $289.76 million — 51.6% of total MEV transaction volume of $562 million. While the average profit per sandwich attack has fallen to just $3 due to intense competition, the aggregate extraction remains substantial. Monthly MEV extraction from sandwich attacks alone ran at approximately $2.5 million by late 2025, down from $10 million in late 2024 — a decline driven by competition, not by protocol-level protection.
The deeper concern is censorship. When two builders produce the vast majority of blocks, they hold implicit veto power over which transactions reach the chain. This creates a single point of failure that contradicts Ethereum's core value proposition. A builder that complies with a government sanctions list, for instance, can effectively censor transactions at the infrastructure level — and users would have no protocol-level recourse.
EIP-7732 fundamentally restructures how Ethereum validates blocks by decoupling execution validation from consensus validation — both logically and temporally. Under the current MEV-Boost system, validators outsource block building to off-chain builders through trusted relays. Under ePBS, this relationship is enshrined directly in the protocol, eliminating the need for trusted intermediaries.
The mechanics work as follows:
Builders submit sealed commitments. Rather than a full execution payload, builders submit a SignedExecutionPayloadHeader — a cryptographic commitment that specifies the blockhash and a payment to the proposer. The contents remain hidden until after the proposer commits to a block.
Proposers select blindly. The block proposer selects the highest-paying bid without seeing or being able to tamper with the transaction ordering inside. This prevents proposers from extracting MEV themselves or colluding with specific builders.
Payload Timeliness Committee (PTC) validates. A subset of validators, assigned to the PTC, attests to whether the builder revealed the committed execution payload with the correct blockhash in a timely fashion and whether blob data was available. Critically, PTC members do not need to validate the full execution payload — they perform lightweight checks on signature and blockhash correctness.
Execution validation moves off the hot path. This is perhaps the most technically significant change: full execution payload validation is removed from the critical path of block validation. The next proposer gets 6 seconds, and every other validator gets 9 seconds, to validate the payload. This creates headroom for larger blocks and more complex transactions without slowing down consensus.
The net effect: builders compete in a protocol-native marketplace, proposers cannot see or manipulate transaction ordering, and the relay layer — currently a centralized bottleneck — becomes unnecessary. Trust assumptions shift from off-chain reputation to on-chain cryptographic guarantees.
While ePBS is the consensus-layer headline, Glamsterdam includes a second major feature: EIP-7928 — Block-Level Access Lists (BALs). This execution-layer upgrade addresses a different but related problem: state bloat and gas mispricing.
BALs introduce deterministic access lists that allow the protocol to more accurately price storage operations based on actual resource consumption. Currently, Ethereum's gas model imperfectly captures the cost of state access, leading to mispriced operations that can be exploited. EIP-7904, a complementary proposal for benchmarked gas repricing, aligns resource costs more precisely with computational reality.
Together, these changes target Ethereum's two oldest infrastructure problems: centralization in block production (via ePBS) and inefficiency in state management (via BALs). The Ethereum Foundation's January 2026 Checkpoint noted that BALs are further along in development, with devnets already running, while ePBS remains a more complex implementation with devnets still in development.
For Layer 2 networks like Base (Coinbase's L2), the implications are significant. Base has publicly endorsed both headliners, noting that ePBS supports increasing blob targets — critical as blob usage is consistently saturated and L1 data availability has become a bottleneck for L2 scaling. Post-Fusaka, Ethereum targets 14 blobs per block (maximum 21), a 2.3x increase in L2 data space. Glamsterdam is expected to push these numbers higher.
Not everyone believes ePBS should ship in Glamsterdam. Sigma Prime researcher Dapplion published a detailed technical critique arguing against inclusion on several grounds:
Lack of urgency. MEV-Boost has been performing well operationally, with infrequent incidents and minimal user impact. The current system works — it just works through off-chain infrastructure. There is no imminent crisis forcing a protocol change.
Premature commitment. Multiple design questions remain unresolved, including slot auctions, execution-layer payment paths, and MEV Burn integration. Committing to one ePBS flavor now sacrifices design flexibility for better implementations that may emerge from active research.
Capital barrier increase. Under ePBS, builders must stake capital to participate. This staked-builder requirement raises entry costs for an already highly centralized market, potentially adding friction to new players rather than reducing concentration.
Sufficient headroom. The chain currently has enough capacity for the next fork cycle. The scaling benefits of ePBS — payload splitting and delayed execution — can theoretically be delivered without the full complexity of trustless builder payments.
The counterargument, advanced by Prysm developer potuz, is that the core technical changes have already been implemented in the Prysm and Teku consensus clients, and given consensus on All Core Developers calls, ePBS could ship by Q1 2026. If ePBS isn't ready for interoperability by the end of February, however, it may be deferred to the Hegota fork later in the year.
Glamsterdam addresses who builds blocks. The follow-up upgrade, Hegota (targeted for late 2026), addresses what must be included in them.
The headline candidate for Hegota is EIP-7805: Fork-Choice Enforced Inclusion Lists (FOCIL), proposed by Thomas Thiery of the Ethereum Foundation's Robust Incentives Group. FOCIL was originally considered for Glamsterdam but was moved out to reduce fork complexity — a decision Base publicly supported, noting that testing FOCIL alongside ePBS could delay the upgrade beyond 2026.
FOCIL works by allowing multiple validators — rather than a single block builder — to enforce transaction inclusion through Ethereum's fork-choice rule. In each slot, a committee of validators builds and gossips their own inclusion lists based on their view of the mempool. Attesters then only vote for a proposer's block if it includes transactions from all stored inclusion lists. This creates a protocol-level guarantee that valid transactions will be included within a bounded timeframe, regardless of builder preferences.
Buterin has described distributed block building as the "long-term ideal holy grail" — where no single entity ever constructs a complete block. The ePBS + FOCIL combination represents the two-step path toward that vision: first separate the roles (Glamsterdam), then enforce inclusion (Hegota).
For validators: ePBS makes block proposing simpler and more profitable. Validators no longer need to trust off-chain relays or run MEV-Boost infrastructure. The protocol-native auction ensures they receive the highest bid transparently. Currently, validators using MEV-Boost earn approximately 5.69% APY — ePBS is expected to maintain or improve these returns by removing relay fees from the value chain.
For builders: The competitive dynamics shift significantly. The staked-builder requirement introduces capital costs but also levels the playing field by moving the marketplace on-chain. Builders can no longer gain advantages through exclusive relay relationships or proprietary infrastructure. Competition will be purely on execution quality and MEV strategy.
For L2s: Both ePBS and BALs directly improve L2 economics. ePBS supports higher blob throughput, reducing data availability costs for rollups. BALs improve gas pricing accuracy, which reduces the overhead L2 sequencers pay for L1 settlement. Base estimates that these improvements collectively make Ethereum L1 "a more reliable and efficient settlement layer for the long term."
For users: The most immediate benefit is reduced MEV extraction. While ePBS alone doesn't eliminate sandwich attacks, it creates the structural foundation for future improvements (like MEV Burn, which would redistribute extracted value to validators rather than builders). FOCIL, when it ships in Hegota, adds explicit censorship resistance — ensuring no builder can indefinitely delay or exclude a valid transaction.
Ethereum's block production has quietly become one of the most centralized layers in all of crypto. Two entities control over three-quarters of the blocks, a trusted relay layer operates entirely off-chain, and hundreds of millions of dollars in MEV flow through opaque infrastructure that exists outside the protocol. For a network that positions itself as the credibly neutral settlement layer for the global economy, this is an existential contradiction.
Glamsterdam doesn't solve everything. The debate between shipping ePBS now versus waiting for better designs is legitimate, and the staked-builder requirement introduces genuine concerns about further entrenching capital advantages. But the direction is clear: Ethereum is moving to enshrine the separation of powers that currently operates on handshake agreements and reputation into cryptographic protocol-level guarantees.
Combined with Hegota's FOCIL later this year, the 2026 roadmap represents Ethereum's most ambitious attempt to deliver on its original promise: a permissionless, censorship-resistant, decentralized world computer. The question is no longer whether Ethereum will address its centralization problem, but whether the implementation can match the ambition — and whether the market will wait.