Understanding the noise affecting a quantum device is of fundamental importance for scaling quantum technologies. A particularly important class of noise models is that of Pauli channels, as randomized compiling techniques can effectively bring any quantum channel to this form and are significantly more structured than general quantum channels. In this paper, we show fundamental lower bounds on the sample complexity for learning Pauli channels in diamond norm with unentangled measurements. We consider both adaptive and non-adaptive strategies. In the non-adaptive setting, we show a lower bound of $\Omega(2^{3n}\epsilon^{-2})$ to learn an $n$-qubit Pauli channel. In particular, this shows that the recently introduced learning procedure by Flammia and Wallman is essentially optimal. In the adaptive setting, we show a lower bound of $\Omega(2^{2.5n}\epsilon^{-2})$ for $\epsilon=\mathcal{O}(2^{-n})$, and a lower bound of $\Omega(2^{2n}\epsilon^{-2} )$ for any $\epsilon > 0$. This last lower bound even applies for arbitrarily many sequential uses of the channel, as long as they are only interspersed with other unital operations.