In comments and other answers it was mentioned that there are some other search engines which could be better when searching for mathematical expressions. But I think that as nowadays several pages uses LaTeX syntax (Wikipedia, this site, to mention just two important examples). Additionally, some people have also TeX-sources of their documents online. So it is not entirely hopeless to google simply for the LaTeX version of the formula. However, there is problem that many things can be typeset in LaTeX in many different ways. Another problem is choice of variables. It is often useful to restrict search to some site(s), for example, "x^2+y^2=z^2"+site:math.stackexchange.com+OR+site:wikipedia.org.
Searching for TeX using Google
Let us try some concrete examples.
Continuum hypothesis
For example, let us assume that I know that there was something like $\aleph_1=2^{\aleph_0}$ was mentioned in the class, but I do not know the name of this formula. (But, luckily enough, I know to typeset it in TeX.)
If I try some reasonable search queries, for example:
in all of them we can find at least some things related to continuum hypothesis.
Euler's formula
You mentioned Euler's formula. Simply searching for "e^{i\pi}" returns Wikipedia article about Euler's identity among the top results. Searching for "e^{\pi i}" also returns some relevant hits.
An integral
Let us say that I am looking, for some reason, for the integral $\int \frac{\sin(x)}{\sin(x)+\cos(x)}\,\mathrm dx$.
Here are some examples of searches, which seem reasonable when searching for this particular integral:
Triangular numbers
In this post, I have mentioned some examples of the search for the formula $\frac{n(n+1)}2$ for the $n$-th triangular numbers, although it was in a somehow different context.
Searching in Google Books
Google Books contain a lot of data. Google usually do a good job in OCR-ing these books. However, they don't OCR Greek letters, math formulas, etc. Occasionally it is possible to guess how some math symbol would be OCR-ed.
For example:
Of course, this trick only has a very limited usability.
e^(i*pi) + 1 = 0
. Doesn't work everytime, but.. $\endgroup$