In Kenneth Kunen's The Foundations of Mathematics, he presents the following exercise and hint.
Exercise I.11.34 Ordinal arithmetic doesn't raise cardinality. That is, assume that $\alpha$, $\beta$ are ordinals with $2 \leq \text{min}(\alpha,\beta)$ and $\omega \leq \text{max}(\alpha,\beta)$. Then prove that $|\alpha + \beta| = |\alpha \cdot \beta| = |\alpha^\beta| = \text{max}(|\alpha|,|\beta|)$.
Hint. For $\alpha + \beta$ and $\alpha \cdot \beta$ this is easy from Theorem I.11.32 and the definition of $+$ and $\cdot$. The proof is slightly tricky for exponentiation, since the natural proof by induction on $\beta$ seems to use the Axiom of Choice (AC). Here, we take the recursive computation in Table I.1 to be the definition of $\alpha^\beta$. Now say we want to prove by induction on $\beta$ that $\alpha^\beta$ is countable whenever $\alpha, \beta$ are countable. If $\beta$ is a limit, we have $\alpha^\beta = \text{sup}_{\xi < \beta}(\alpha^\xi) = \bigcup_{\xi < \beta} (\alpha^\xi)$ which (applying induction) is a countable union of countable sets. But the well-known fact that a countable union of countable sets is countable uses AC. To avoid AC first fix a $\delta < \omega_1$ and then fix an $f: \delta \xrightarrow{1-1} \omega$. Now, we may now define, for $\alpha,\beta<\delta$, an injection from $\alpha^\beta$ into $\omega$; the definition uses $f$, and is done by recursion on $\beta$; the definition also uses a (fixed) injection from $\omega \times \omega$ into $\omega$.
I finished proving that $|\alpha + \beta| = |\alpha \cdot \beta| = \text{max}(|\alpha|, |\beta|)$ without issue. I'm stuck on following the hint for exponentiation. Usually I find the hints clear, but this time I'm confused. We are defining a function based off an injection from $\alpha^\beta$ into $\omega$ by recursion on $\beta$. I don't exactly follow the construction. I'm also wondering, this hint only gives the case that we are assuming that $\alpha$ and $\beta$ are countable. What about ordinals of arbitrary cardinality that satisfy the above conditions? Is it just a small modification of the argument we use for the countable case?
Prove this by induction on $\beta$.
$\alpha^2$ is $\alpha\cdot\alpha$, so you are back to multiplication.
$\alpha^{\gamma+1}=\alpha^\gamma\cdot\alpha$, by the induction hypothesis that's the same cardinality as $\alpha\cdot\alpha$, so we are back to multiplication again.
If $\beta$ is limit, then $\alpha^\beta=\sup\{\alpha^\gamma\mid \gamma<\beta\}$. Then either $\beta>\alpha$ is a cardinal, in which case $|\alpha^\gamma|<\beta$, and so $\alpha^\beta=\beta$. Or else we have a uniform way of enumerating $\alpha^\gamma$ for all $\gamma<\beta$. This means that we get an injection into $\max(\alpha,\beta)\times\beta$, so we're done.
Or, if you don't mind using AC, just apply it judiciously to get $\max\{\alpha,\beta\}$ again.