Yes, this is like Tarski's result that if $\mathfrak{c}^2=\mathfrak{c}$ for every infinite cardinal $\mathfrak{c}$, then AC holds.
In fact, it suffices to suppose that $\mathfrak{c}\mathfrak{d}=\mathfrak{c}$ whenever $\mathfrak{c},\mathfrak{d}$ are cardinals and $\mathfrak{c}$ is infinite and $\mathfrak{d}<\mathfrak{c}$. (Note this follows from your hypothesis.)
For this, fix an infinite set $y$. It suffices to see that $y$ is wellorderable. We may assume that $y$ does not contain any ordinals. Let $\kappa$ be the least ordinal such that $\kappa\not\leq\mathrm{card}(y)$. Let $\mathfrak{c}=\mathrm{card}(y\bigcup\kappa)$. Then $\mathrm{card}(y)<\mathfrak{c}$, so by the hypothesis, $y\times(y\cup\kappa)$ has cardinality $y\cup\kappa$. So $y\times\kappa\leq y\cup\kappa$. But then the usual argument shows that $y$ is wellorderable, which suffices. That is, fix an injection $\pi:y\times\kappa\to y\cup\kappa$. Since $\kappa\not\leq\mathrm{card}(y)$, for each $x\in y$, there is some $\alpha<\kappa$ such that $\pi(x,\alpha)\notin y$, so $\pi(x,\alpha)\in\kappa$. Let $\alpha_x$ be the least such. Then define $f:y\to\kappa$ by setting $f(x)=\pi(x,\alpha_x)$, and note that $f$ is injective, so $y$ is wellorderable, as desired.