---
title: "Rollups"
url: "https://radix-wiki.miow.me/contents/tech/core-concepts/rollups"
updated: 2026-10-03
description: "A rollup is a layer-2 blockchain scaling construction in which transactions are executed outside the base blockchain (the layer 1) while the data required to…"
---

# Rollups

|  |  |
| --- | --- |
| **Topic** | Layer-2 blockchain scaling |
| **Main variants** | Optimistic rollups; zero-knowledge (validity) rollups |
| **Dominant cost** | Layer-1 data publication |
| **Related** | [Sharding](/contents/tech/core-concepts/sharding) · [Atomic composability](/contents/tech/core-concepts/atomic-composability) · [Blockchain trilemma](/contents/tech/core-concepts/blockchain-trilemma) |

# Rollups

A **rollup** is a layer-2 blockchain scaling construction in which transactions are executed outside the base blockchain (the layer 1) while the data required to reconstruct them is published to it. By moving computation off-chain and posting compressed transaction data on-chain, rollups increase throughput while continuing to derive their security from the underlying chain rather than from an independent validator set.

Two designs dominate: _optimistic rollups_, which presume submitted batches are valid and rely on fraud proofs to catch invalid ones, and _zero-knowledge (zk) rollups_, which accompany every batch with a cryptographic validity proof. Beyond the proof mechanism, rollups are distinguished by an unusual fee structure: because a batch's fixed costs are shared by all transactions inside it, fees can fall as usage grows – within limits.

## Optimistic and zero-knowledge rollups

An optimistic rollup executes transactions off-chain and posts the resulting batch data to the base layer – on Ethereum [as calldata or as blobs](https://ethereum.org/en/developers/docs/scaling/optimistic-rollups/), a dedicated data format that nodes prune after roughly eighteen days. Batches are presumed valid when submitted; correctness is enforced after the fact through [multi-round interactive fraud proofs](https://ethereum.org/en/developers/docs/scaling/optimistic-rollups/), under which an invalid state commitment can be disputed. A zk-rollup instead submits a validity proof alongside each state commitment. In cost terms the two differ in where their overheads sit: for a zk-rollup the [validity proof is a fixed cost incurred even when a batch contains no transactions](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/), while an optimistic rollup carries per-transaction signature data as an additional variable cost.

### Data availability

Both designs depend on the base layer to store the data needed to reconstruct the rollup's state, and this transaction data is [the largest variable cost of operating a rollup](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/) and the primary contributor to rising rollup costs. That expense has motivated hybrid designs – _validiums_ and _volitions_ – that keep transaction data off-chain in search of cheaper average costs. The analytics site L2BEAT accordingly [classifies live systems by proof type and data-availability mode](https://l2beat.com/scaling/summary), tracking validiums and optimiums separately from rollups proper.

## Fee economics

What a rollup user pays bundles two distinct services. Ethereum's developer documentation describes the fee on an optimistic rollup as [the cost of writing data to layer 1 plus an operator fee for layer-2 computation](https://ethereum.org/en/developers/docs/scaling/optimistic-rollups/). Barnabé Monnot's first-principles analysis draws the same division on the cost side: operators bear layer-2 operating costs, layer-1 data-publication costs – [the dominant new cost of the rollup model](https://barnabe.substack.com/p/understanding-rollup-economics-from) – and congestion costs on the rollup itself, and must recover them from user fees, extractable value, and issuance, since [an operator cannot sustainably run at a loss](https://barnabe.substack.com/p/understanding-rollup-economics-from).

Alex Beckett's analysis of rollup fee economics decomposes the cost of a batch into [fixed costs – the state commitment, plus the validity proof for a zk-rollup – and variable costs, chiefly transaction data plus signatures for optimistic rollups](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/). The fee an individual user pays corresponds to the batch's average cost: total cost divided by the number of transactions in the batch. Because fixed costs are amortized over the whole batch, the average falls as the batch grows – so long as the marginal cost of one more transaction sits below the average. In Beckett's worked example, [a 500-transaction batch at an average cost of $1 totals $500; adding a transaction with a $0.70 marginal cost lowers the average to $0.9994](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/). On this basis he describes rollups as ["the first type of blockchain that can incur positive network effects with regards to transaction fees"](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/), in contrast to monolithic chains, where each additional user raises fees for everyone.

The effect has limits. Once marginal cost reaches average cost, the rollup's costs follow a standard short-run cost curve, and [past the minimum of that curve each additional transaction pushes fees upward again](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/) – the negative network effects of a monolithic chain reappear. And because transaction-data costs are bound to the base layer's finite block space, rollups remain [exposed to fee spikes originating on layer 1](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/), especially when the base layer also hosts user-facing applications. Beckett concludes that scalable data-availability layers are fundamental to keeping rollups inexpensive without sacrificing security or decentralization.

## Composability across rollups

Each rollup maintains its own state and executes transactions in its own environment. A single transaction therefore cannot atomically span two rollups: moving assets or invoking contracts across them requires bridging or settlement through the base layer. On 27 September 2026, [L2BEAT tracked 22 layer-2 rollups on Ethereum](https://l2beat.com/scaling/summary), 11 optimistic and 11 zero-knowledge, alongside validiums and optimiums, with Base the largest at about 16.5 billion US dollars in total value secured and Arbitrum One second at about 11.7 billion; it also grades systems on a [Stage 0–2 decentralization framework](https://l2beat.com/scaling/summary) reflecting how far each has reduced reliance on trusted operators. A consequence of this proliferation is that liquidity and application state fragment across many execution environments, and the synchronous, all-or-nothing composability available within a single chain does not extend across the ecosystem as a whole.

## Rollups and Radix

Radix addresses the same scaling constraint – often framed as the [blockchain trilemma](/contents/tech/core-concepts/blockchain-trilemma) – without a rollup layer, instead scaling the base layer itself through integrated [sharding](/contents/tech/core-concepts/sharding). The published design uses a "pre-sharded" model with a fixed, pre-allocated maximum shard space rather than adding shards dynamically, and its [Cerberus](/contents/tech/core-protocols/cerberus-consensus-protocol) consensus "braids" validation across shards to enforce system-wide transaction ordering; deterministic shard indexing means transactions from separate accounts involve separate shards and can be processed asynchronously in parallel. Neither property is live today: [Babylon](/contents/tech/releases/radix-mainnet-babylon) mainnet runs Cerberus unsharded, and the [Xi’an production candidate](/contents/tech/research/hyperscale-rs) neither braids nor pre-allocates a shard space, splitting shards on demand under a beacon-maintained trie instead. On the composability question, radix.wiki's [atomic composability](/contents/tech/core-concepts/atomic-composability) page states that a single [transaction manifest](/contents/tech/core-protocols/transaction-manifests) can chain any number of operations – swaps, loans, deposits, withdrawals – into one all-or-nothing sequence, and that Cerberus braids consensus across every shard a transaction touches, maintaining atomicity across shards, in contrast to sharded systems whose cross-shard operations depend on asynchronous bridges. Rollups and integrated sharding thus represent two responses to the same problem: rollups amortize base-layer costs by moving execution off-chain at the price of fragmenting state across separate environments, while an integrated sharded layer 1 aims to expand execution capacity while preserving a single atomically composable state. For a broader comparison of the two ecosystems' approaches, see [Radix vs Ethereum](/contents/tech/comparisons/radix-vs-ethereum).

## References

1. Beckett, Alex (2022). [The economics of rollup fees](https://web.archive.org/web/20260119081538/https://www.alexbeckett.xyz/the-economics-for-rollup-fees/). alexbeckett.xyz (archived via the Internet Archive).
2. Ethereum Foundation. [Optimistic Rollups](https://ethereum.org/en/developers/docs/scaling/optimistic-rollups/). ethereum.org developer documentation.
3. Monnot, Barnabé (2022). [Understanding rollup economics from first principles](https://barnabe.substack.com/p/understanding-rollup-economics-from). The Price of Agency (Substack).
4. L2BEAT. [Scaling Summary](https://l2beat.com/scaling/summary). l2beat.com, accessed 27 September 2026.