USDC balances without decimal bugs

About this guide

We checked the sources for this guide on 24 Sept 2026. Use the linked sources for current availability, terms and network details. See subsequent updates

An Arc wallet has one USDC balance exposed through two interfaces. Native EVM amounts use 18 decimals; ERC-20 USDC amounts use 6. Treating them as separate assets doubles the balance. Treating their raw integers as the same unit creates a factor-of-one-trillion error.

The stablecoin contract reference identifies the USDC ERC-20 interface. The EVM differences reference is the place to check Arc-specific runtime behavior.

Name the unit in your data model

Keep raw amounts as integer strings or big integers. A field called amount leaves too much unstated. Prefer names such as nativeAmountWei and usdcAmountMicros, with chain ID, token address and decimals attached to persisted transfers.

Human amountNative integer, 18 decimalsERC-20 integer, 6 decimals
1 USDC10000000000000000001000000
0.25 USDC250000000000000000250000

Format the amount only at the presentation boundary. Do not send a formatted currency string back through floating-point arithmetic to construct a transaction.

Make rounding a product decision

The integer relationship is:

native units = ERC-20 units × 1,000,000,000,000

The reverse conversion can leave a remainder. Define how your application handles that remainder rather than silently discarding it in every calculation. Display rounding, ledger precision and fee reimbursement are different decisions.

For example, a support screen can show a rounded USDC amount while preserving the original integer and decimals for reconciliation. A report should record both the display value and the unit used to calculate it.

Keep gas accounting separate from the transfer amount

An ordinary transfer reduces the sender’s balance by the transfer plus the actual fee. Use the receipt and documented fee fields for accounting; a preview is an estimate. If an application sponsors gas, record the sponsor’s payment separately from the recipient’s transfer.

Arc’s sponsored-transactions guide explains how relayers and paymasters cross the precision boundary. Check the current gas reference before relying on a fee-floor or ceiling value; parts of that page still identify their parameters as testnet configuration.

Test the failures that reveal unit mistakes

Use a whole-dollar amount, a fractional amount, the smallest ERC-20 unit and an amount large enough to expose floating-point precision loss. Assert the recipient’s delta as well as the sender’s delta. Run a sponsored transfer and a self-paid transfer separately.

Also test “send maximum.” The maximum transferable balance depends on who pays gas, the chosen operation and the current estimate. The screen should explain why its maximum differs from the displayed total instead of returning a generic insufficient-funds error.

Sources & references

Documents cited in this guide.

  1. Canonical contract addressesdocs.arc.io
  2. EVM differencesdocs.arc.io
  3. sponsored-transactions guidearc.io
  4. Gas and feesdocs.arc.io