Ethereum's roadmap is a sequence of planned upgrades aimed at cheaper transactions, stronger security and better wallets. The last three upgrades focused on rollups and staking: Dencun (March 2024) added cheap "blob" data, Pectra (May 2025) changed staking limits and wallets, and Fusaka (December 2025) let nodes check blob data by sampling. Next is Glamsterdam, which was in testing with no mainnet date as of mid-November 2026.
How does Ethereum upgrade itself?
Ethereum changes through coordinated hard forks: every node updates its software to follow new rules from an agreed block. A hard fork that most nodes skip would split the chain, so upgrades are planned months ahead.
Each change starts as an Ethereum Improvement Proposal (EIP), a public design document with a number. Client developers discuss EIPs on open "All Core Devs" calls and agree on a bundle for each upgrade.
Since the 2022 Merge, an Ethereum node runs two programs, an execution client and a consensus client, and each layer names its own upgrade, and the two names get combined. Prague plus Electra became Pectra. Fulu plus Osaka became Fusaka.
Before mainnet, an upgrade runs on test networks. Ethereum's main public testnets today are Sepolia and Hoodi. Problems found there can, and do, delay mainnet.
What did recent upgrades change?
The three most recent major upgrades all activated on these dates:
| Upgrade | Mainnet date | Headline change | Who noticed |
|---|---|---|---|
| Dencun | March 13, 2024 | Blobs: a cheap, temporary data space for rollups | Layer 2 users, through lower fees |
| Pectra | May 7, 2025 | Larger validators, smart account features for ordinary addresses | Stakers and wallet users |
| Fusaka | December 3, 2025 | PeerDAS: nodes sample blob data instead of storing all of it | Rollups, through more blob space |
Dencun: blobs for rollups
Layer 2 rollups post their transaction data to Ethereum so anyone can check it. Before Dencun, they paid for permanent calldata, the same space every node keeps forever.
Dencun's main change, EIP-4844, or proto-danksharding, added blobs. A blob holds 4,096 pieces of 32 bytes each, 128 KiB in total. Blobs have their own fee market, separate from regular gas fees, and the EVM can't read them. Nodes delete them after about 18 days, which is long enough for anyone to check a rollup's data. The starting target was 3 blobs per block, with a maximum of 6.
Rollups such as Base could now post data far more cheaply, savings that ethereum.org says can be passed on to users as lower fees.
Pectra: staking and wallets
Pectra bundled about a dozen EIPs. The ones that matter most to users:
- EIP-7251 raised a validator's maximum effective balance from 32 ETH to 2,048 ETH. Large operators can run fewer, bigger validators, and staking rewards now compound on balances above 32 ETH. The minimum to run a validator is still 32 ETH.
- EIP-7702 lets an ordinary address sign an authorization to behave like a chosen smart contract. That brings account abstraction features such as batched transactions and sponsored gas to existing wallets. The same power cuts the other way: a malicious authorization can hand an address to an attacker's code, so it belongs on the list of things wallet drainers ask you to sign.
- EIP-7691 doubled blob capacity to a target of 6 and a maximum of 9.
- EIP-7002 lets stakers trigger a validator exit from their withdrawal address.
- EIP-7623 raised calldata costs for data-heavy transactions. Ethereum.org says over 99% of transactions were unaffected.
Fusaka: sampling instead of storing
Before Fusaka, every full node downloaded every blob. That capped blob space at what a modest home node could handle.
Fusaka's PeerDAS (EIP-7594) changed this. Blob data is extended with redundancy, so any half of it can rebuild the whole. Each regular full node then holds only about one-eighth of the data, while validators with large balances hold more. It is the data availability sampling idea that modular blockchains rely on, applied to Ethereum itself.
Fusaka also introduced blob-parameter-only (BPO) forks, small preset upgrades that only raise blob limits. The Ethereum Foundation scheduled two: BPO1 on December 9, 2025, raised the target to 10 and maximum to 15, and BPO2 on January 7, 2026, raised them to 14 and 21.
Other Fusaka changes include a default block gas limit of 60 million, up from 45 million, and a cap of about 16.8 million gas on any single transaction. It also added a precompile for the P-256 signatures that phones and passkeys use.
How much more data can Ethereum carry now?
A worked example makes the blob changes concrete. Ethereum has a slot every 12 seconds, which is 7,200 slots a day. Assume a block in every slot and blocks at the target.
| Stage | Target blobs per block | Data per block | Data per day |
|---|---|---|---|
| Dencun (March 2024) | 3 | 384 KiB | About 2.6 GiB |
| Pectra (May 2025) | 6 | 768 KiB | About 5.3 GiB |
| After BPO2 (January 2026) | 14 | 1,792 KiB | About 12.3 GiB |
The arithmetic for the last row: 14 blobs times 128 KiB is 1,792 KiB per block, and 1,792 KiB times 7,200 blocks is about 12.3 GiB a day. That is over four times Dencun's target capacity in under two years. Because of PeerDAS, a regular node stores only a fraction of that, and deletes it after about 18 days.
The target is not a ceiling. Blocks can carry up to the maximum, and blob fees rise when demand stays above target, much as base fees do for gas.
What is Glamsterdam, and when does it ship?
Glamsterdam pairs the Amsterdam execution upgrade with the Gloas consensus upgrade. Its official EIP list included 18 scheduled changes as of November 2026. The two headline ones, according to the Ethereum Foundation's testnet announcement, are:
- Enshrined proposer-builder separation (ePBS, EIP-7732). Today, most validators outsource block building to specialist builders through outside software, a market shaped by MEV. ePBS moves that exchange into the protocol, reducing reliance on trusted middleware and giving validators more time to check blocks.
- Block-level access lists (EIP-7928). Each block records which accounts and storage slots its transactions touch. Clients can then load data and check transactions in parallel, a foundation for raising capacity.
It also reprices gas for several operations, so costs better match the work they cause.
The schedule has slipped. As of mid-November 2026, Glamsterdam activated on Sepolia on October 6. Its Hoodi testnet fork, planned for October 26, was postponed. An Ethereum Foundation update said audits found that two gas changes, EIP-8037 and EIP-8038, would break assumptions some deployed contracts rely on. Developers decided not to fork Hoodi until those issues were resolved, and no mainnet date had been set.
What comes after Glamsterdam?
The upgrade after it is named Hegotá. Ethereum.org's roadmap page lists it for 2027, with fork-choice enforced inclusion lists (FOCIL) and frame transactions, and notes that its scope is not final. Inclusion lists aim to make it harder for block builders to censor transactions.
Further out, ethereum.org groups the work into four goals: cheaper transactions, extra security, better user experience and privacy by default. Longer-term items include:
- Statelessness, so nodes can verify blocks without storing large amounts of state.
- Single-slot finality, so blocks finalize within one slot instead of about 15 minutes.
- Post-quantum cryptography, to replace signatures that a large quantum computer could break. Ethereum.org says some upgrades like this are likely 5 to 10 years away.
Treat dates beyond the next upgrade as goals, not commitments. The Glamsterdam delay shows how one finding in testing can push a schedule back by months.
Do you need to do anything?
Usually not. For the Sepolia launch, the Ethereum Foundation said ETH holders and users need take no action, and that has been true of every recent mainnet upgrade. Your ETH doesn't move or change during a fork.
That makes any message telling you to "upgrade," "migrate" or "swap" your ETH for an upgrade a red flag. Node operators and stakers are different: they must update both clients before each fork. Developers should test contracts against new gas rules, especially for Glamsterdam.
For the bigger picture, see how Ethereum differs from Bitcoin, our Ethereum hub and our upgrades coverage.
Key takeaways
- Ethereum upgrades through scheduled hard forks built from numbered EIPs, tested on Sepolia and Hoodi before mainnet.
- Dencun added blobs, a cheap temporary data space that cut costs for layer 2 rollups.
- Pectra raised the maximum validator balance to 2,048 ETH and let ordinary addresses act like smart contracts through EIP-7702.
- Fusaka's PeerDAS lets nodes sample blob data, and two follow-up forks raised the target from 6 to 14 blobs per block.
- Glamsterdam brings ePBS and block-level access lists, but as of mid-November 2026 its Hoodi fork was postponed and it had no mainnet date.
- Ordinary ETH holders don't need to do anything for an upgrade, so "upgrade your ETH" messages are scams.
Sources
Get the weekly commit
New blockchain deep dives every week.

