A typical animal cell has 1000-2000 mitochondria. From a statistical point of view, assuming a random distribution of the mitochondria and that the cell splits in half, the probability of having 0 mitochondria is (1/2)^1000 or 9e-302. This makes it an impossibility for all practical purposes.
With enough mitochondria, a process to ensure the cell splits roughly in half and a somewhat random distribution of mitochondria would be sufficient to get at least one mitochondria in each daughter cell.
To address the assumptions:
- Random distribution of mitochondria - assumed in the question
- Cells split roughly in half - source on dividing assymetries "In somatic divisions, however, cell size asymmetry is mild and, only rarely, one daughter cell is more than double the size of the other."