Why are grouped aggregate functions on a columnstore index fast?

Suppose that a table only contains integers and has a clustered columnstore index.

CREATE TABLE cci_account
(  
    AccountKey int NOT NULL,  
    UnitSold int,
    INDEX ColStore CLUSTERED COLUMNSTORE
);

then it appears that summing one column grouped by another is always fast on such a table, e.g.

SELECT
    AccountKey,
    SUM(UnitSold)
FROM
    cci_account
GROUP BY
    AccountKey;

Why is this so fast? My understanding of the columnstore architecture is that the segments containing AccountKey have no idea what UnitSold value is on their rows. How, then, is SQL Server able to so quickly group by AccountKey? I cannot imagine any algorithm that would achieve this.