Ethereum’s Sepolia testnet is now running Glamsterdam upgrade rehearsal blocks with a gas budget pushed toward 200 million per block, more than three times the mainnet limit, according to CoinDesk coverage of the test runs on Wednesday. The first two upgraded epochs also finalized without incident.
The rehearsal comes after a late software fix. Developers released version 7.2.1 of Prysm, a widely used consensus client, late Monday to make sure Tuesday’s trial properly tested the network’s new capacity ceiling. The earlier release capped processing at 60 million gas; the updated client scales the maximum up to 200 million. Without the fix, the exercise would have measured nothing: the client would have refused blocks the test was designed to produce.
There is a gap between what the testnet allows and what it uses. Sampled blocks during the rehearsal consumed less than half of the new allowance, CoinDesk reported. That is deliberate. Spiking the ceiling far above actual load lets client teams see how the network behaves under headroom it has never run before, and exposes problems that only appear when a single block can carry an unusually large batch of transactions. Propagation delays, client memory growth and reorg risk all scale with block size, and none of them can be measured by arithmetic alone.
Client diversity is what makes the trial meaningful rather than a single piece of software’s self-assessment. Sepolia runs several execution and consensus clients against each other, including Geth, Besu, Lighthouse and Prysm. If a limit that one client tolerates breaks another, the mismatch shows up during a rehearsal instead of during mainnet activation, where the response to a client stall is reverting a hard fork, an event Ethereum has so far avoided entirely in its history.
What Glamsterdam changes
Glamsterdam bundles several changes to how Ethereum produces and prices blocks. Among them is enshrined proposer-builder separation, the plan to move the block-building market into the protocol itself rather than leaving it to outside relay software. Today, most blocks are assembled by specialized builders chosen through out-of-protocol auctions, a design that works commercially but concentrates builder power in a handful of firms. Enshrining it pulls that auction into the consensus layer and reduces the trust validators must place in intermediaries.
The bundle also covers gas repricing work, adjusting individual operation costs so resource-heavy instructions pay rates that match what they actually cost the network to process. This is not window dressing: mispriced opcodes have historically been the route for both network attacks and stuck blocks, and repricing rounds have followed each major capacity expansion since EIP-1559.
| Ethereum gas limit, current vs Glamsterdam target | Per block |
|---|---|
| Previous testnet ceiling | 60 million gas |
| Glamsterdam rehearsal ceiling on Sepolia | 200 million gas |
| Highest sampled usage in rehearsal | Under half of the allowance |
| Approximate multiple of current headroom | More than 3x |
Why bigger L1 blocks matter for L2 users
The capacity jump matters for users on layer-2 rollups most of all, whether or not they ever pay attention to the base chain. Bigger L1 blocks mean cheaper data posting for rollups: blob and call-data capacity stretches further, so the fixed cost of publishing a batch covers more transactions. Cheaper data posting flows straight into fee estimates for millions of Arbitrum, Base and Optimism users. Capacity on the base layer and the economics of every rollup on top of it are linked that way.
It also changes what rollups can offer. Tokenized stock platforms and prediction markets have all pushed activity to Ethereum and its rollups this month. BNB Chain alone crossed $1 billion in tokenized equities and exchange-traded funds in October, and platforms such as Ondo now support tokenized transfers across Ethereum, Solana and BNB Chain. Demand like that is what core developers are sizing capacity against, not an abstract transaction counter.
Cheap data posting also affects stablecoin and real-world asset settlement volumes, which have grown faster than retail activity since the 2025 cycle. Every basis point saved on publishing proofs to L1 accrues to platforms running dollar settlement for institutional users, a segment that pushed USDC adoption through consumer apps this month when Samsung added cross-border transfers to its wallet.
Timeline pressure
Ethereum has balanced two goals in this cycle: making blocks heavier, and keeping validator hardware requirements reasonable. Bigger blocks raise the cost of running a full node, and client developers have spent months arguing over how far to push the limit without centralizing node operation. The Sepolia trial gives them empirical numbers to argue with instead of projections, which is worth more than any roadmap document.
Testnet rehearsals tend to precede mainnet activation by weeks or a few months, depending on how many failures surface. Validators, staking services and node operators will be watching client releases over the coming weeks, since a hard-fork upgrade means updating every machine that participates in consensus. The reward mechanics around staking are also under review through proposals such as EIP-8148, which streamlines reward sweeps for validators, and the validator population itself has dropped to about 863,000 units after Pectra let operators combine balances and shut down redundant validators.
For traders the rehearsal is background noise. Ether traded near $2,570 on Wednesday, down on the session as Bitmine’s announcement from Token2049 redirected attention to treasury flows. But for the chain’s longer-term ability to carry real transaction volume, a successfully finalized epoch sequence under a tripled block budget is one of the clearer data points core developers have collected all year. The next gate is a longer soak period on Sepolia, then a mainnet shadow-fork before any activation date gets set in stone.