We propose and implement a four-phase handshake protocol for bundled-data asynchronous circuits with consideration given to power consumption and area. A key aspect is that our protocol uses three phases for generating the matched delay to signal the completion of the data-path stage operation whereas conventional methods use only one phase. A comparison with other protocols at 0.18 µm process showed that our protocol realized lower power consumption than any other protocol at cycle times of 1.2 ns or more. The area of the delay generator required for a given data-path delay was less than half that of other protocols. The overhead of the timing generator was the same as or less than that of other protocols.