> For the complete documentation index, see [llms.txt](https://docs.lumia.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.lumia.org/ecosystem/lumia/architecture/data-availability/validium.md).

# Validium

**Who this is for:** developers and issuers assessing the security/cost trade-off underneath Lumia. **You should already know:** [Data Availability](https://app.gitbook.com/s/chmhPiVNAafGw48c4oU2/ecosystem/lumia/architecture/data-availability).

### Definition

A **validium** is a chain that proves its state transitions correct with a validity proof (the same cryptographic guarantee a rollup uses), but keeps the underlying transaction data **off** the base chain, holding it instead with a separate data-availability mechanism (a committee of signers, or a dedicated DA network). A **rollup**, by contrast, posts that data directly onto the base chain (as calldata or in a dedicated data blob), so anyone can reconstruct the L2's full state from the base chain alone.

### Trade-offs vs. a rollup

The trade is cost against a specific, narrow trust assumption. Posting full transaction data to Ethereum is the single largest cost driver for a rollup at scale: a validium avoids that cost by keeping data elsewhere. What it gives up is the rollup's strongest guarantee: with a rollup, correctness *and* data are both fully backed by the base chain, so even if every off-chain party disappeared, the data to reconstruct state would still be recoverable from the base chain itself. A validium instead depends on its data-availability layer actually holding up its end: if that layer failed to make data available, state could become unrecoverable even though the validity proof itself was never wrong. This is why a validium's data-availability design (see [Where Lumia's transaction data lives](https://app.gitbook.com/s/chmhPiVNAafGw48c4oU2/ecosystem/lumia/architecture/data-availability/where-lumia-transaction-data-lives)) is worth understanding on its own, not waved away as an implementation detail.