Single-hop vs multi-hop swaps

A single-hop swap uses one market to exchange the input token for the output token. A multi-hop swap passes through at least one intermediate token. Splitting is a separate idea: portions of the input can follow different paths, each of which may contain one or several hops.

Three hypothetical route shapes

A diagram containing three pools does not, by itself, establish whether the trade is split. Read the connections and amount allocations. Sequential pools consume the preceding pool's output; parallel branches consume separate portions.

Why introduce another token?

A direct market can be weak while two indirect markets together offer stronger execution. For instance, a widely used intermediary may connect two assets that rarely trade directly. The router needs to compare the complete resulting output, not assume fewer hops are always better.

Uniswap's multi-hop guide describes sequential token paths. Its routing repository separately discusses splitting across routes.

Do you keep the intermediate asset?

In a standard atomic route, the intermediate token is used during execution to obtain the final token. It need not remain as a separate holding in the user's wallet. The requested recipient and final-delivery operations determine what the user receives.

A manual sequence is different. If you independently complete A-to-C and later decide whether to execute C-to-B, you hold C between those actions and the two swaps do not share one atomic boundary.

What additional hops imply

Every stage has its own pricing and settlement rules. More hops can create a useful connection, but they also introduce more state to model and operations to execute. The route must account for each stage's actual input, including any applicable fee effect.

Count hops to understand the sequence; inspect final output and constraints to evaluate the trade. Hop count alone does not rank route quality.