The weakness in most bridge designs is that the verifying parties have nothing at stake. Axelar's answer is to make verification a proof-of-stake consensus problem: validators bond tokens, sign for cross-chain messages, and can be slashed for misbehaviour.
Why bonding changes the calculus
An unbonded multisig signer who colludes loses nothing but reputation. A bonded validator loses capital. That converts bridge security from a trust question into an economic one, where the cost of attacking exceeds the gain as long as the bonded value is large enough relative to the value being secured — which is the same logic that secures proof-of-stake chains generally.
The size question
Axelar's validator set and total bonded value are much smaller than Ethereum's or Cosmos Hub's. For a transfer of moderate size that is ample; for very large transfers, the economic security should be compared against the amount being moved. This is the right analysis to perform and almost nobody performs it.
Cost and speed
A full consensus round-trip means transfers take longer than intent-based bridges and cost more than CCTP. General message passing is supported, so Axelar carries contract calls as well as tokens.
Who should use it
Users and applications wanting an economically secured cross-chain route with broad coverage, sized against the network's bonded value. For speed on EVM routes, Across is faster; for USDC specifically, CCTP is safer.