Like in any language, a FLOAT will eventually corrupt data due to rounding errors.
FLOAT
Please use DECIMAL, which has well-defined behavior for rounding and range overflows.
DECIMAL