As Stefan said, floating-point will also give the wrong result sometimes.
>>> cdiv = lambda x, y: -(-x // y)
>>> cdiv(2**60, 3)
384307168202282326
>>> import math
>>> math.ceil(2**60 / 3)
384307168202282304
But also it will confuse people reading the code, who will have to know the fiddly details of floating-point math to know under what circumstances it gives integer-exact answers.
To Neil’s comment about names, a longer name that would have precedent would be ceildiv, by analogy with operator.floordiv.