We were told, that under the assumption that $y=\mathbf{x}'\beta_\circ+u$ in the representation of the ordinary least squares estimator (with $N$ being the sample number) $$\hat{\beta}=\beta_\circ + (1/N \cdot\mathbf{X'X})^{-1} (1/N \cdot\mathbf{X'u})$$ the $(1/N \cdot\mathbf{X'X})^{-1}$ converges to the expectation $\mathbb{E}(\mathbf{xx}')$ (and the $ (1/N \cdot\mathbf{X'u})$ to $\mathbb{E}(\mathbf{x}u)$), because of the LLN.
Why is that like that, how can $\mathbb{E}(\mathbf{xx}')$ be seen to be the expected value?