Because:
Assuming that our model of the universe is correct and that it has been expanding at the rate we've measured for the time we've inferred...
1. We know what the maximum volume of the universe can be.
2. We know what the smallest possible volume is that an atom can be squeezed into.
3. Therefore, we know what the maximum number of atoms is that can fit into the universe, if they were squeezed as tightly as possible, given the observed conditions of the universe which do not have them packed like sardines.
4. That number is less than a googolplex.
5. Even if we hypothesized that we could pack them as tightly as sardines, we could write a much larger number that would do the job.
And yes, we do have a pretty good idea how many atoms there are in our planet. Certainly the order of magnitude.
When I say "we" I'm not speaking personally because I'm not a physicist. I don't keep that data on file because situations like this rarely come up where I wish I had it; and I don't have the skills to derive it myself. But the scientific community has it. I'm sure someone will pop up on this thread with those numbers, although it might be a year from now because this thread has a very strange habit of going dark for months at a time and then erupting into a flurry of activity.
The point that needs to be brought home here is this: We have a notational system that allows us to very easily and compactly express numbers that are SO LARGE that they have absolutely no practical application.
When you start using multiple levels of exponentiation, it's easy to get into inconceivably large numbers before you know it. If what you read was wrong and the universe really could hold one googolplex atoms, then we can just write a number so much larger that the concept of "orders of magnitude" isn't of any use in understanding it. How about googolplex^(googolplex^googolplex)?
The calculations I outlined above in the first three steps are trivial. I actually could track down the source data and do the arithmetic, without having to be a physicist. It's that easy. We will end up with a number that is expressed with multiple levels of exponentiation. It will be either greater than or less than a googolplex. If it's greater, we can write an adequately larger number with just a few keystrokes.
Our ability to express large numbers is expanding faster than the universe.