Ethereum Roadmap Explained: Fusaka, Glamsterdam & the Road to Danksharding

Introduction

Ethereum has shipped five major upgrades since 2022 (ethereum.org, 2026), yet the next one still has no mainnet date (EIP-7773). Glamsterdam passed the Sepolia testnet on 6 October 2026 (Ethereum Magicians, 2026), but on 22 October developers postponed its 26 October fork on the Hoodi testnet to investigate how EIP-8037 interacts with existing mainnet contracts (Ethereum Foundation, 2026). No new Hoodi date has been set, and ethereum.org lists only "Q4 2026, date not yet confirmed." Behind that gap sits a steady thread: each fork since Dencun has widened the data path that rollups use, from three blobs per block to a target of 14. This article follows that thread from the Merge to Hegotá (ethereum.org, 2026), separates what is live from what is merely proposed, and maps every upgrade to the roadmap track it advances and to what it changed for users. Every date and status below was checked against the Ethereum Foundation, ethereum.org or client configuration on 16 October 2026, and the Glamsterdam schedule again on 5 November 2026.

What Is the Ethereum Roadmap and How Do Its Tracks Fit Together?

The Ethereum roadmap is six research tracks that run in parallel, not a queue of upgrades. Each named hard fork pulls work from several tracks at once. Danksharding, the plan for cheap rollup data, sits in the Surge and is the thread this article follows.

Six tracks, not six steps

Vitalik Buterin named the Merge, Surge, Scourge, Verge, Purge and Splurge in 2022 and revisited each in a six-part series in October 2024 (Vitalik Buterin, 2024). The Surge targets more than 100,000 transactions per second across layer 1 (L1) and layer 2 (L2) chains ↗. The Scourge limits centralization in block building and staking. The Verge makes verification cheap, the Purge expires old history, and the Splurge covers the Ethereum Virtual Machine (EVM) and account abstraction. ethereum.org says plainly that the roadmap will change as research moves (ethereum.org, 2026). Status in the table below is as of 5 November 2026.

Why "Ethereum 2.0" is a retired name

Before 2022, plans for proof of stake and sharding were bundled as "Ethereum 2.0." The ethereum.org team phased out the Eth1 and Eth2 labels in January 2022 in favor of execution layer and consensus layer, because the names implied a new chain replacing the old one and scammers used "ETH2" to sell fake migrations (Ethereum Foundation, 2022). No ETH2 token exists. The Merge joined two layers of one network. Older guides that describe Ethereum 2.0 phases describe a plan that changed when rollups replaced shard chains.

Data current as of November 2026.

Every track points at the same constraint: Ethereum scales by moving execution to rollups and keeping the base layer cheap to verify.

Why Is Danksharding the Through-Line of Ethereum's Scaling Plan?

Danksharding is Ethereum's plan to give rollups cheap, plentiful data, and each upgrade from Dencun to Fusaka delivered one step of it. Full danksharding, the last step, remains a later phase (ethereum.org, 2026). The roadmap since 2020 serves one division of labor: L1 secures and stores data, rollups execute transactions ↗.

From shard chains to a rollup-centric plan

Early plans split Ethereum into shard chains that each processed transactions. In October 2020 Ethereum's developers adopted a rollup-centric roadmap instead, because rollups, which run transactions off-chain and post the data to Ethereum, were developing faster than expected (Vitalik Buterin, 2020). ethereum.org states that shard chains "are no longer needed and have been dropped from the roadmap" (ethereum.org, 2026). Vitalik Buterin calls the result a pluralistic implementation of sharding in which each L2 acts as a shard with its own rules (Vitalik Buterin, 2024). Sharding survived in one place: data. Danksharding shards the verification of rollup data, not the execution of transactions. Each stage raises capacity without raising the hardware a home node needs ↗, which is the constraint the whole design is built around.

Proto-danksharding, PeerDAS and full danksharding

Three stages carry the plan. Proto-danksharding, specified in Ethereum Improvement Proposal (EIP) 4844 and named for researchers Protolambda and Dankrad Feist, added blobs: data attached to blocks that nodes delete after about 18 days (ethereum.org, 2026). PeerDAS, EIP-7594, let nodes confirm blob availability by sampling instead of downloading everything. Full danksharding would spread responsibility for verifying blobs across subsets of nodes. ethereum.org lists it as a later phase and the Ethereum Foundation (EF) groups the follow-on designs as FullDAS research (ethereum.org, 2026; Ethereum Foundation, 2025).

Two of those five steps are history that predates the current fork queue, so the sequence starts with the Merge and Dencun.

What Did the Merge and Dencun Change for Ethereum Users?

The Merge changed who secures Ethereum, and Dencun changed what rollups pay to store data. The Merge ended mining on 15 September 2022 and Dencun began danksharding on 13 March 2024 (ethereum.org, 2026). Both are complete, and both sit on today's roadmap as foundations rather than open work.

The Merge, September 2022

Paris, the Merge upgrade, activated on 15 September 2022 and moved Ethereum from proof-of-work mining to proof-of-stake validation. Energy use fell by about 99.95% (ethereum.org, 2026). It joined the existing execution chain with the Beacon Chain, the consensus layer that had run proof of stake since 1 December 2020. The proof-of-work chain and the Beacon Chain ran side by side for almost two years before the join. Shapella followed on 12 April 2023 and let validators withdraw staked ETH for the first time. The Merge changed who secures the network, not the interface: addresses, balances and applications carried over unchanged.

Dencun and EIP-4844, March 2024

Dencun activated on 13 March 2024 and introduced proto-danksharding through EIP-4844, a transaction type that carries blobs, 128 KiB data containers priced in their own fee market (ethereum.org, 2026; Etherscan, 2024). The blob target started at three per block with a maximum of six. Dencun also shipped transient storage opcodes (EIP-1153) and exposed beacon block roots to the EVM (EIP-4788) (ethereum.org, 2026). Blob data expires after about 18 days, long enough for anyone to check a rollup and far cheaper than permanent storage. A Rutgers preprint associates Dencun with a 95.5% fall in the average fee on Arbitrum, Base and OP Mainnet (arXiv, 2026; preprint). The fee arithmetic belongs to a companion article on L2 gas fees ↗.

Dencun opened the data path, and Pectra then widened it while changing what a wallet can do.

What Did Pectra Change for Wallets, Stakers and Rollups?

Pectra, live since 7 May 2025, is finished work: it changed what an ordinary wallet can do, how much ETH one validator can hold and how many blobs a block carries. It is history on this roadmap, but its wallet and blob changes set up later forks.

What Pectra changed for wallets and stakers

The Ethereum Pectra upgrade shipped EIP-7702, which lets an externally owned account, the key-controlled wallet type, adopt the code of an existing smart contract. That brings transaction batching, gas sponsorship and recovery options without moving funds to a new address, a design known as account abstraction ↗ (ethereum.org, 2026). The same fork raised a validator's maximum effective balance from 32 ETH to 2,048 ETH, so large stakers can run fewer validators and earn rewards on every ETH above the minimum (ethereum.org, 2026).

What Pectra changed for blobs

Pectra raised the blob target from three to six and the maximum from six to nine per block through EIP-7691 (ethereum.org, 2026). Demand justified it: before Fusaka, layer 2 usage often reached the nine-blob limit (Ethereum Foundation, 2025). Pectra opened the twice-a-year cadence that core developers now aim for, and Fusaka followed about seven months later (Consensys, 2025). It was also the last blob increase tied to a full named fork so far. Fusaka's blob-parameter-only forks now handle later steps. The EF's 2026 priorities call EIP-7702 "an important step" and name smart contract wallets by default as the end state (Ethereum Foundation, 2026).

Pectra doubled the blob target 14 months after Dencun, and Fusaka made the next increases safe.

How Did Fusaka Deliver PeerDAS and Blob-Parameter-Only Forks?

Fusaka went live on 3 December 2025 at 21:49 UTC and put danksharding's sampling on mainnet. Its headliner, PeerDAS, lets nodes verify blob data without storing all of it. Two blob-parameter-only forks followed within five weeks and lifted capacity to 14 blobs per block (Ethereum Foundation, 2025).

PeerDAS, the Fusaka headliner

PeerDAS, defined in EIP-7594, splits blob data into columns spread across 128 subnets; each node custodies some columns and samples others to confirm the rest exist (ethereum.org, 2026). The EF put the theoretical gain at eight times the blob throughput, from about 64 KB/s to about 512 KB/s (Ethereum Foundation, 2025). Prysm developer Preston Van Loon estimated an immediate cut of almost 90% in blob bandwidth for each node at the old six-blob target (ETHDenver talk, 2025). Erasure coding means any 50% of the columns can rebuild a whole blob, and a validator custodies about one-eighth of the data, so withholding a minority of data cannot hide it (ETHDenver talk, 2025). Smaller nodes, including home stakers, can therefore keep verifying as capacity grows.

BPO forks and the 14/21 blob schedule

A blob-parameter-only (BPO) fork changes the blob target, maximum and fee-update fraction and nothing else; it needs a client release but adds no features (Ethereum Foundation, 2025). BPO1 on 9 December 2025 raised the target and maximum from 6 and 9 to 10 and 15, and BPO2 on 7 January 2026 raised them to 14 and 21 (Ethereum Foundation, 2025; EIP-8138). On 5 November 2026 live client configuration reports 14 and 21 with no later blob fork scheduled, and BPO3 exists only as a draft proposal (Ethereum RPC eth_config, 2026). Core developers chose gradual BPO steps over jumping to PeerDAS's theoretical maximum so that they can watch mainnet behavior between steps (Ethereum Foundation, 2025).

L1 changes in Fusaka and what each upgrade changed for users

The gas limit had climbed from 30 million to 45 million during 2025 through validator signaling (Ethereum Foundation, 2025). Fusaka, which carried 13 EIPs, set the default at 60 million through EIP-7935 and capped a single transaction at 16,777,216 gas through EIP-7825, a limit that protects nodes and prepares parallel processing (Ethereum Foundation, 2025). The live block gas limit on 5 November 2026 was still 60 million (Ethereum RPC, 2026). EIP-7951 added a secp256r1 precompile, so the secure hardware in phones and passkeys can sign transactions directly. The table summarizes what each shipped upgrade changed for users.

Data current as of November 2026.

With blob capacity at 14 and 21, the next fork targets block production rather than data.

What Is Glamsterdam, and What Is Actually Scheduled?

Glamsterdam is the upgrade after Fusaka, named for the Amsterdam execution layer and the Gloas consensus layer. It has two headliners, ePBS and block-level access lists. As of 5 November 2026 it has run on one testnet, its fork on a second was postponed with no new date, and it has no mainnet date (Ethereum Foundation, 2026).

The two Glamsterdam headliners

Enshrined proposer-builder separation (ePBS, EIP-7732) writes the split between block proposers and builders into the protocol, which reduces reliance on trusted middleware, and widens the propagation window from 2 seconds to about 9 through a Payload Timeliness Committee that attests to the block and its payload separately (Ethereum Foundation, 2026; ethereum.org, 2026). Block-level access lists (EIP-7928) record the accounts and storage slots a block touches, so clients can read state and validate transactions in parallel. Developers chose both in 2025 (Ethereum Foundation, 2025). ePBS belongs to the Scourge and block-level access lists to L1 scaling. A repricing of state creation (EIP-8037) targets state growth of 120 GiB a year (ethereum.org, 2026).

What is scheduled and what is not

Glamsterdam was scheduled for the Sepolia testnet at 13:53 UTC on 6 October 2026 and activated there (Ethereum Foundation, 2026; Ethereum Magicians, 2026). Developers agreed on All Core Devs execution call 247 on 8 October to run it on Hoodi on 26 October 2026 (Ethereum Magicians, 2026), but on 22 October the Ethereum Foundation announced that call 248 had postponed the Hoodi fork to investigate how EIP-8037's higher gas cost for creating state interacts with existing mainnet smart contracts; no new Hoodi date had been set by 5 November, and the next execution-layer call, ACDE 249, is scheduled for 19 November 2026 (Ethereum Foundation, 2026; Ethereum PM, 2026). The postponement notice "does not set a Glamsterdam mainnet activation date," the meta EIP-7773 leaves its mainnet row blank, and ethereum.org says "Q4 2026 · Date not yet confirmed" (Ethereum Foundation, 2026; EIP-7773; ethereum.org, 2026). Scope is frozen, but the meta EIP is still in Review status, so it can change before mainnet (ethereum.org, 2026). Candidate December mainnet dates floated on 15 October depended on the Hoodi fork and lapsed with its postponement, so nothing is scheduled (Ethereum PM, 2026; Ethereum Foundation, 2026).

Data current as of November 2026.

Proposed, scheduled and live are three different states, and Glamsterdam's mainnet step is none of them yet (EIP-7773; Ethereum Foundation, 2026).

What Is Hegotá, and Which Changes Are Chosen So Far?

Hegotá follows Glamsterdam, named for the Bogotá execution layer and the Heze consensus layer. Its consensus-layer headliner, FOCIL, was chosen in the first half of 2026 (Ethereum Foundation, 2026). ethereum.org expects the fork in Q2 2027 with no confirmed date, and its scope beyond two changes is undecided.

FOCIL, the chosen headliner

Fork-choice enforced inclusion lists (FOCIL, EIP-7805) let a committee of validators each propose transactions that the block builder must include, so no single builder can censor a valid transaction alone (ethereum.org, 2026). Developers moved FOCIL out of Glamsterdam in January 2026 to cut that fork's scope, and the EF's April 2026 checkpoint reports it selected as Hegotá's consensus-layer headliner (Ethereum Foundation, 2026). The EF's September 2026 tier list pairs it with EIP-8369, which defines which transactions qualify for inclusion lists. ethereum.org adds that FOCIL can strengthen layer 2 settlement and shorten exit windows for optimistic rollups ↗. FOCIL also belongs to the Scourge.

Frame transactions and the unsettled scope

All Core Devs execution call 233 on 26 March 2026 rejected frame transactions, EIP-8141, as a headliner for lack of client support and kept them as a placeholder under Considered for Inclusion (Ethereum Magicians, 2026; Ethereum Foundation, 2026). By 27 August 2026 execution clients backed moving EIP-8141 to Scheduled for Inclusion, and ethereum.org now lists FOCIL and frame transactions as the two scheduled changes (EtherWorld, 2026; ethereum.org, 2026). Frame transactions let the account define what makes a transaction valid, which supports social recovery, spending limits, sponsored gas and quantum-resistant signatures natively (ethereum.org, 2026). On 8 October call 247 moved several more proposals to Considered for Inclusion, so Hegotá's scope is not final (Ethereum Magicians, 2026).

Hegotá's two named changes cover the Scourge and the Splurge, while the Surge's next step is slower to define.

How Close Is Full Danksharding, and What Comes After Hegotá?

Full danksharding is not scheduled for any fork. Fusaka and the BPO forks delivered its sampling layer, and the Ethereum Foundation describes the next steps as networking work plus research under the FullDAS name. Past Hegotá, the roadmap shifts toward verification and post-quantum security.

What full danksharding still needs

The EF's blob plan lists three open items: networking refinements that cut redundant sampling traffic, a next generation of PeerDAS that builds on the pipelining ePBS provides, and FullDAS research into designs where nodes sample more of the data while staying fault-tolerant (Ethereum Foundation, 2025). None carries a mainnet date, and client configuration on 5 November 2026 shows no fork after BPO2 (Ethereum RPC, 2026). The EF's 2026 priorities add continued gas limit increases toward and beyond 100 million (Ethereum Foundation, 2026), and Sepolia's gas limit ramped to 200 million after Glamsterdam activated there (Ethereum Magicians, 2026).

Binary trees, proofs and post-quantum targets

The Verge changed course. Verkle trees were the original plan for stateless clients, but the EF's 2026 priorities describe "a move to binary trees and statelessness in the long term" and a zero-knowledge Ethereum Virtual Machine (zkEVM) ↗ attester client moving from prototype toward production (Ethereum Foundation, 2026). A September 2026 EF post sets a goal of quantum-resistant Ethereum L1 across execution, consensus and data by December 2029 (Ethereum Foundation, 2026). That goal is a target, not a scheduled upgrade. The post calls the cadence it requires, about 7.2 months per fork from a December 2026 Glamsterdam, "quite aggressive" (Ethereum Foundation, 2026). That December start predates the Hoodi postponement, and no testnet date now supports it.

The long-range plan is clear, so the practical question is what to watch before the next fork lands.

What Should Readers Watch Before Glamsterdam Reaches Mainnet?

Three signposts show whether Glamsterdam is moving toward mainnet: a new Hoodi date to replace the 26 October fork postponed on 22 October, the client releases that follow it, and a mainnet date in an Ethereum Foundation announcement. None of them forecasts the ETH price.

Signposts between now and a mainnet date

Hoodi will be the first public test with a broad set of staking operators; Sepolia ran with only four or five active builders (Ethereum Magicians, 2026). A mainnet date then arrives in an EF announcement with an activation slot and a client release table, the form taken by Fusaka's 6 November 2025 announcement for its 3 December activation (Ethereum Foundation, 2025). Forkcast, the All Core Devs notes and ethereum.org's Glamsterdam page carry live status. Client teams had agreed to release Hoodi-compatible versions by 16 October 2026 (Ethereum Magicians, 2026); after the postponement they shipped new releases, such as Geth v1.17.10 and Lighthouse v8.3.2, that remove the fork, and the schedule that would have raised Hoodi's gas limit to 200 million at the fork epoch was dropped too (Ethereum Foundation, 2026).

What the upgrades do not tell holders about ETH

ETH was about $2,420 on 5 November 2026, roughly 51% below its all-time high of about $4,946 on 24 August 2025 (CoinPaprika API, 2026-11-05). Fusaka shipped about three months after that August 2025 high. Upgrades change how many blobs L2s can post, how much state nodes must carry and how block building is policed; they leave demand for blockspace and the price of ETH open, and none of them is a price forecast. Following the schedule is a way to understand the network, not to time an entry.

FAQ

?Do ordinary wallet holders need to do anything when a fork activates?

No. ethereum.org says regular users do not need to take action or move funds to stay secure when a network upgrade activates. Node operators and validators must update their clients before the fork, and some applications may change how they fetch data. Wallets and exchanges usually announce their own handling of a fork, so reading their notices is the practical step.

?What is the difference between a hard fork, a testnet fork and a blob-parameter-only fork?

A hard fork is a coordinated protocol change that every client must adopt on the same block. A testnet fork is the same change run first on test networks, Sepolia then Hoodi in Glamsterdam's case, so bugs surface before mainnet (Ethereum Foundation, 2026). A blob-parameter-only fork changes only blob target, maximum and fee-update settings, with no new features (Ethereum Foundation, 2025). Fusaka's two blob-parameter-only (BPO) forks used that route.

?Why was Glamsterdam not ready in the first half of 2026?

The Ethereum Foundation's February 2026 priorities post targeted the first half of 2026. ethereum.org now shows Q4 2026. The Foundation's October 2025 checkpoint (Ethereum Foundation, 2025) had already flagged 23 small proposed features plus the winter holidays, and said whether the fork landed in mid- or late-2026 was still open (Ethereum Foundation, 2025). Scope, testing and benchmarking all took longer than the early target allowed.

?Why do Ethereum upgrades have two-part names such as Glamsterdam?

Ethereum has two layers that upgrade together. The execution layer takes the name of a city, Amsterdam for Glamsterdam, and the consensus layer takes the name of a star, Gloas. The names are then merged. Fusaka is Osaka plus Fulu, and Hegotá is Bogotá plus Heze, named after the Devcon location and a star respectively (ethereum.org, 2026).

?What happened to Verkle trees on the roadmap?

Verkle trees were the original plan for stateless clients, which verify blocks without storing the full state. The Ethereum Foundation's 2026 priorities now describe a longer-term move to binary trees and statelessness, plus a zero-knowledge Ethereum Virtual Machine (zkEVM) attester client, instead. Vitalik Buterin's 2024 Verge post had already listed Verkle trees, STARK-friendly hash trees and conservative hashes as options to choose between.

?Will Glamsterdam make Ethereum mainnet transactions cheaper?

ethereum.org says block-level access lists will likely reduce per-transaction gas costs as capacity grows, because parallel processing lets the network safely raise the gas limit. No fee figure is promised. Sepolia's gas limit ramped to 200 million after the fork, but mainnet limits are set by validators and depend on testing, so any change in fees will follow later, not at activation.

?Is quantum resistance part of the Ethereum roadmap?

Yes. ethereum.org lists post-quantum security as a roadmap item, and the Ethereum Foundation's September 2026 priorities set a target of quantum-resistant L1 across execution, consensus and data by December 2029. Frame transactions in Hegotá support the execution-layer part, because accounts can adopt new signature schemes without a hard fork for each one.

?Does the roadmap still depend on layer 2s, or is Ethereum scaling the base layer instead?

Both. The Ethereum Foundation's March 2026 platform post says L1 should scale by orders of magnitude as a settlement and shared-state hub, while layer 2 value lies in differentiation and customization. Blob capacity keeps growing for rollups, and the gas limit keeps rising for L1. The 2020 rollup-centric plan has been widened, not abandoned.

References / Sources