Silo V2

Silo V2 runs permissionless lending markets where each asset pair is isolated, so a bad asset in one market cannot drain the others. The August 15, 2026 survey reported about $5.71M of supplied assets across Avalanche, Sonic, Ethereum and Arbitrum, plus about $2.47M separately tagged as borrowed. The survey covers permissionless isolated markets, not a vault wrapper run by a curator, so no more basic existing review applies. The live protocol is below our size floor, and we do not open an individual review until it clears the floor. The v1 size rule therefore decides the result.

Review triggers: events that reopen the verdict
• TVL sustained above the retired TVL threshold for 30 days

Watched nightly: a warning on its venues or files, or a cited document that changes, reopens the memo. The first confirmation is due 2026-11-15.

The research file

Applicability to the surveyed record

Silo V2 consists of permissionless, risk-isolated lending pairs with two fixed asset silos per market. A direct supplier chooses the named market and earns borrower interest while taking that market’s collateral, oracle, interest-rate and liquidation risks. DefiLlama builds this slug from Silo factory contracts and adds up deposited assets. That supports treating it as direct lending, not a delegated vault.

Current observation and perimeter

The DefiLlama API read on 2026-08-15 classified Silo V2 as Lending and reported approximately $5.71M TVL across Avalanche, Sonic, Ethereum and Arbitrum, with roughly $2.47M separately tagged as borrowed. Official deployment materials still list V2 markets alongside newer V3, so the record is live rather than archived. It remains far below the shared v1 size floor, so we do not open an individual review until it clears that floor.

Control, loss and exit applicability

Anyone can deploy a V2 market. Isolation limits contagion but does not prove that a chosen pair is safe. Suppliers take the risks of borrower use, collateral, oracle and liquidation failure. A Silo risk report warns that bad pricing can cause losses despite isolation. A withdrawal burns the lending receipt for the underlying and depends on available liquidity when borrowers have taken supplied assets.

Why the class rule decides

The shared v1 size rule decides this factory-level direct-lending record. We do not open an individual review until it clears the size floor. Reopen after reproducible TVL remains at least the size floor for 30 days. Then review a named market rather than the aggregate slug: collateral and debt assets, oracle and fixed parameters, deployer and hooks, use by borrowers, bad debt, audit scope, incident history, and enough liquidity for a withdrawal of the proposed size. A separate managed-vault product would need the delegated-allocation test.

Research status

This is a capacity-unproven record for Silo V2, not a quality rejection or approval. Reported TVL says how big the venue is, not what the client owns, who can change the rules, or how a position exits at a proposed size. The individual review opens when the protocol clears the size floor.

Sources

The claims above trace to these. Where a number could not be independently verified, the thesis says so.

• Silo — V2 security and mechanism review · primary · accessed 2026-08-15
  Supports: V2 isolated pairs, permissionless deployment, immutable silos, security scope
• Silo — V2 market risk example · primary · accessed 2026-08-15
  Supports: V2 isolation, permissionless markets, oracle loss, liquidation, market-specific risk
• Silo — deployments and live V2 perimeter · primary · accessed 2026-08-15
  Supports: V2 markets, contract deployments, lifecycle
• DefiLlama — Silo V2 survey record · secondary · accessed 2026-08-15
  Supports: current TVL, borrowed value, four-chain perimeter, factory methodology, Lending category

Inherited controls

The research above describes the protocol layer. Every position also inherits the asset it holds and the chain it settles on. The layer with the most administrative power sets the position’s effective control; that describes control, not quality or suitability.