As supplemental to the other answers, and as I feel like that real powers should be defined before we get into Taylor series or differential equations, here is a direct approach using just the binomial theorem.
The usual definition of $e$ is $e=\lim_{n\rightarrow\infty}\left(1+\frac{1}{n}\right)^n$. Through this we can see that $$e^x=\lim_{n\rightarrow\infty}\left(1+\frac{1}{n}\right)^{xn}=\lim_{n\rightarrow\infty}\left(1+\frac{x}{n}\right)^n,\quad x\in\mathbb{R}$$
So this is how we define $e^x$.
Using the binomial theorem, we may write $$\left(1+\frac{x}{n}\right)^n=\sum_{k=0}^n\frac{\binom{n}{k}}{n^k}x^k=\sum_{k=0}^n\frac{n(n-1)..(n-k+1)}{n^k}\frac{x^k}{k!}\leq\sum_{k=0}^n\frac{x^k}{k!}$$
as $(n-j)/n\leq 1$. Hence $$e^x\leq
\sum_{k=0}^\infty\frac{x^k}{k!}$$
On the other hand, assuming that $x\geq 0$ we have for $n\geq m\geq 2$
$$\sum_{k=0}^m\frac{n(n-1)..(n-k+1)}{n^k}\frac{x^k}{k!}\leq\sum_{k=0}^n\frac{n(n-1)..(n-k+1)}{n^k}\frac{x^k}{k!}$$
Actually, we do not really have to assume positivity of $x$, just that $m$ is large enough and that $n,m$ are even, and then we can estimate the difference of the RHS and the LHS to be positive, and everything will still work for negative $x$ too.
Taking limits for $n\rightarrow\infty$ this gives
$$\sum_{k=0}^m\frac{x^k}{k!}\leq e^x$$ for any $m>2$, hence
$$\sum_{k=0}^\infty\frac{x^k}{k!}\leq e^x$$
Thus, we have proved that indeed the power series above is equal to $e^x$
$$\sum_{k=0}^\infty\frac{x^k}{k!}=e^x$$