Blockchains are incredible at making a shared record nobody can quietly rewrite. They’re terrible at being a free, eternal database. That tension is finally catching up. If every token, allow
Blockchains are incredible at making a shared record nobody can quietly rewrite. They’re terrible at being a free, eternal database. That tension is finally catching up.
If every token, allowance, airdrop, NFT, and abandoned contract lives forever, state balloons. Nodes get heavier. Fewer people can run them. Fees reflect that pressure. And at some point, the network has to say no.
That’s where “state expiry” comes in. Not a silver bullet, not an overnight switch. Just a set of strategies that say: the chain keeps the most important stuff accessible, and the rest needs proofs, snapshots, or a fresh touch to stick around.
Point Details State growth is compounding Measured weekly new Ethereum state rose from roughly 105 MiB to ~326 MiB after gas limit increases, or about 116 GiB per year of extra state pressure (HackMD — Week 2). Expiry vs history pruning History expiry drops old block bodies and receipts. State expiry targets long-lived account and storage entries. Different goals, different risks. Tiering helps today A prototype showed about 58% of trie nodes moved cold and an observed ~21.6% total storage reduction, shrinking the hot trie near 60% (HackMD — EPF notes). UTXO-style payments on Ethereum Research suggests native UTXOs for one-shot payments could cut permanent state for those flows by about 99.8% (Ethereum Research (ethresear.ch)). Builders have levers now Transient storage, logs over storage, batch patterns, and state-light designs reduce footprint even before protocol-level expiry lands. User experience must not break Expiry needs renewals or proofs so balances, allowances, and NFTs are still usable without scaring normal users.
Every extra byte the chain promises to remember forever has a cost. It lives somewhere on disk. It moves across the network. It gets hashed into tries and proofs. And every node has to carry it, which slowly raises the minimum hardware bar for participation.
Cheap-to-store-forever is a broken price signal. If it costs almost nothing to plant an entry that never expires, actors will do it. A faucet here, airdrop dust there, a vault nobody closes. Times ten years, across millions of users. That creep doesn’t show up in one scary chart on day one. It shows up when sync goes from hours to days, and when a home node starts to feel like a small data center.
That’s why most serious roadmaps now include ideas that either shift work off state, time-box what lingers, or make you pay more if you want your data to stick around.
Two words that get mixed up all the time: history and state.
History can be safely trimmed once we agree on a cutoff and a way to fetch old data when needed. State is trickier. Cut the wrong piece and a normal payment might fail because a balance or allowance went missing. So state expiry has to be careful. Either make it cheap to refresh what you still use or make it easy to resurrect entries with a compact proof.
We aren’t hand-waving here. We have measurements. When Ethereum’s gas limit roughly doubled to the 60 million range, weekly new state didn’t just scale linearly with everyone’s optimism. It spiked to around 326 MiB per week, up from something like 105 MiB per week. That pencils out to roughly 116 GiB more state per year that nodes have to absorb (HackMD — Week 2).
Even if your node hardware can take the hit, think about decentralization pressure. More disk. More IOPS. Heavier tries. Longer reorg processing. Slow snapshots. You end up selecting for operators with better budgets and bigger rigs. The do-it-at-home crowd quietly drops off.
And that’s before you factor in new L2 traffic settling on L1, NFTs with sprawling metadata pointers, and DeFi contracts that persist a lot of per-user data. The slope is up and to the right.
These words get thrown around interchangeably. They’re not the same thing.
This removes the obligation for nodes to serve old block bodies and receipts beyond a certain horizon. It doesn’t touch the live state. Think of it as deleting your camera roll backups from 2018 while keeping your current photo album intact. Clients already ship practical forms of this in various modes.
This is the big one. You place a time or activity window around which state entries are considered live. If an account or storage slot hasn’t been touched within that window, it is considered expired. To use it again, you either refresh it before expiry or present a proof to resurrect it. Getting the UX right here is the hard part.
Not deletion. Just moving cold data off the hot path. An EPF prototype showed that around 58% of trie nodes could live in a cold store with an observed ~21.6% reduction in total node storage, and an effective shrink of the hot trie to about 60% of its prior size (HackMD — EPF notes).
Bottom line: tiering is a near-term win for node ops. Expiry changes the economics of what sticks around. History expiry cleans up dead weight that consensus no longer needs to serve by default.
If you want long-lived state, post collateral or pay rent. If you stop paying, your data is parked or expired until renewed. It is blunt, but it solves the “free forever” bug.
Move toward a world where block producers and verifiers can operate with small witnesses instead of a full copy of state. Verkle trees are often discussed as the commitment structure that makes these proofs small enough to be practical. Not a flip-the-switch upgrade, but a direction that pairs well with expiry.
Account-based chains are great for expressive smart contracts, but they tend to accumulate per-account state. For simple payments, the UTXO model shines because spent outputs disappear. There is active research on adding native UTXOs to Ethereum for one-shot payments, and the estimate is eye-catching: roughly a 99.8% reduction in permanent state for those flows (Ethereum Research (ethresear.ch)).
Blobs helped rollups scale by giving them cheaper data availability lanes. But rollups still carry their own state growth. Many will also need expiry, rent, or markets for state to avoid centralizing their sequencers and provers.
Pro tip: if you’re building critical infra, test your flows on a node that simulates cold-tier latencies and expiry-like reactivation. You’ll surface UX bugs early, before policy makes them unavoidable.
It’s easy to talk ideas. 2026 has given us some numbers to chew on.
These are not hypotheticals. They are snapshots of a community that sees the wall and is experimenting with ways to avoid hitting it at speed.
Bitcoin’s UTXO model deletes spent outputs. That naturally limits long-lived clutter. You still get bloat from tiny “dust” outputs, but policy can discourage that. Account-based systems like Ethereum make for powerful smart contracts, but every long-lived mapping and slot is a barnacle on the hull. If it’s free to keep, it stays.
One way forward is hybridization. Keep accounts for complex apps. Use UTXO-like constructs for simple payments and one-shot entitlements. If those UTXOs vanish once spent and never touch account storage, permanent state drops for the most common on-chain action: paying someone. That’s the attractive part of the 99.8% figure in the research. It attacks the biggest workload with the simplest cleanup property.
One caution: policy changes take time. Expect long lead times, public testnets, and migration kits. The worst outcome would be surprising users. The best outcome is a slow, well-signaled shift where most people barely notice because their wallets and apps do the right thing by default.
If you want a steady pulse on these changes without living in client repos all week, we cover the practical bits and the data behind them at Crypto Daily.
No. Coins are part of the live state because they move often. Expiry targets long-idle accounts or storage. If an entry expires, designs typically include a way to revive it with a proof or a renewal. Wallets should abstract that step.
Pruning old blocks is history expiry. It drops the obligation to serve ancient block bodies and receipts. State expiry concerns the living snapshot that nodes need to process new blocks. It is riskier to touch and needs a careful UX.
If they sit idle past the expiry window, they could require a refresh. Good apps can auto-renew small allowances or prompt you in advance. The goal is that users rarely notice, but builders must plan for it.
Archive nodes help you serve history and old states, but they don’t change network-wide costs. If the protocol expires state, your app still needs to handle reactivation or proofs, even if your own infra is rich.
Not necessarily. The aim is better pricing for storage, not blanket hikes. Some workloads may get cheaper as nodes get lighter, while heavy storage users may pay more to keep data alive.
You can, but that shrinks the set of people who can validate or run a full node. That’s a decentralization hit. Expiry and tiering try to keep the bar low so more people can participate.
It is research for now, but it targets a clear pain point. If simple payments move to a native UTXO path and mostly avoid permanent state, that’s a big win. The trade-offs and engineering are active topics in the research forums.
Disclaimer: This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.
