Layer 1 vs Layer 2: Where a Stablecoin Transfer Settles

The question this page answers

Readers arrive asking which is layer 1 and which is layer 2. They also ask whether the difference matters for a stablecoin transfer. This page answers the second half on documents. It does not attempt a full layer 2 taxonomy. General explainers cover that ground. What this page holds is narrower and more checkable. Where a stablecoin transfer settles. What the fee legs are. What an exit costs in time.

The classification words come from the chains themselves. Base and Arbitrum describe rollups posting to Ethereum ^{[5] [3]}. Polygon’s own documentation describes sidechains with checkpoint anchoring ^[7]. Hyperliquid describes a layer 1 with one unified state ^[9]. Ethereum needs no introduction as the settlement layer the others point at.

Where transfers settle

Settlement is the structural question. A rollup executes independently and settles upstream. The Arbitrum docs state it in their own words. Arbitrum runs on top of Ethereum, which settles the results and keeps the data available ^[5]. The OP Stack model behind Base states the same. Hard finality is guaranteed by Ethereum once the batch lands in a finalized Ethereum block. That typically takes 15 to 30 minutes after submission ^[3].

Polygon settles differently. Milestones finalize a transaction on the sidechain in 2 to 5 seconds, without waiting for Ethereum ^[7]. Checkpoints anchor ranges of blocks to Ethereum afterward, on a governance-set cadence ^[7]. That is anchoring, not per-transaction settlement, and the chain’s own docs frame it as a partial mitigation ^[7].

Hyperliquid removes the boundary entirely within its own system. HyperCore and the HyperEVM share one unified state under one consensus ^[9]. Moving between them is an intra-L1 move with no bridging between the two ^[9]. That makes it the useful baseline. Whatever bridge risk means below, an L1 with one state has none of it internally.

Ethereum finalizes through its own checkpoints. A block finalizes in a chain with a supermajority link between two epoch checkpoints. The Ethereum page holds that mechanism in full.

What an L2 changes in the fee

The fee structure splits at the boundary. L2 fees divide into an L2 execution leg and an L1 data leg. The OP Stack docs state the L1 data fee follows the Ethereum gas price tracked by the GasPriceOracle contract ^[3]. Arbitrum names two payees. The poster covers parent-chain resources, the calldata needed to post the transaction ^[5]. The network fee account covers child-chain resources ^[5].

The L1 leg is now measurable, and small. Arbitrum receipts expose the L1 component directly. On 11 September 2026 pulls, the median L1 component was 187 gas-equivalent, about 0.1 percent of a USDC transfer’s fee ^[6]. Base receipts do not expose the L1 component. Base figures here exclude that leg. The exclusion is named, not silent ^[4]. Polygon has no per-transaction calldata leg at all. Checkpoints are validator-submitted on a governance cadence, and fees are plain EIP-1559 on the sidechain ^[7].

What the fees measured

The conditions metric, one date, one method across all chains. On 11 September 2026, a native USDC transfer at current conditions cost $0.0056 on Ethereum, softening to $0.0051 on a second pull ^[2]. The same transfer cost $0.0006 on Base, at its documented fee floor ^[4]. It cost $0.0020 on Arbitrum, L1 leg included ^[6]. It cost $0.00135 on Polygon, at $0.0013 on the second pull ^[8]. It cost $0.00056 on Hyperliquid, at its base-fee floor ^[10].

Those are protocol-conditions figures, not traffic. The observed medians, batch-inclusive, ran higher on every chain, and Ethereum’s observed median of $0.21 reflected exchange traffic pricing at 2 gwei ^[2]. The chains hub holds each chain’s full two-metric record.

No ranking follows from these numbers. Lower fee with an external settlement dependency is a trade, not a verdict. Higher fee with self-settlement is the same trade seen from the other side.

The bridge and exit question

Settlement structure shows up most plainly in exits. An Arbitrum withdrawal to Ethereum involves inclusion in a rollup assertion, a 6.4-day challenge period, and manual layer 1 execution by any party ^[5]. A Polygon exit requires the checkpoint path. Checkpoint verification is required for bridging assets from Polygon to Ethereum, with the exit queue enforced by Ethereum-side contracts ^[7]. A Base transaction is Ethereum-guaranteed only at hard finality. Before that, it is sequencer-confirmed, and the conservative send-and-settle bar is to wait ^[3].

The stablecoin angle on all three. The moment of guarantee differs by design. A sender who needs settlement certainty reads these numbers before the fee column.

Native and bridged across the boundary

One distinction sharpens exactly at this boundary. The same asset can exist as an issuer-native token on one side. On the other side it can exist as a bridged representation. The canonical current example is X Layer, where Circle publishes a native USDC and a bridged USDC.e side by side, with the issuer’s own footnote stating the bridged form is not issued or backed by Circle ^[1]. This library rosters issuer-native tokens only, and every fee figure above measured a native contract.

What would change the answer

The comparison moves on named events, not arguments. A rollup changing its settlement path would change the settlement row. Its own docs would show it. A fee-floor parameter change would move the conditions column. Base’s minimum base fee change of 19 February 2026 already shows that ^[4]. A checkpoint cadence change on Polygon would move the anchoring row, and it is governance-set by design ^[7]. Nothing here changes by preference.

Where this page sits

The Ethereum page, the Base page, the Arbitrum page, the Polygon page, and the Hyperliquid page hold the full chain records this comparison draws from. The chains hub indexes all of them, with the cost of sending pages carrying the route-level fee tables. Every figure above carries its date.