Clicking "show" next to "Demonstration of the conservation form" reveals more context: the matrix
$$\begin{pmatrix}{\mathbf {u}}\otimes {\mathbf {u}}+w{\mathbf {I}}\\{\mathbf {u}}\end{pmatrix}$$
is written out in components as
$$
{\begin{pmatrix}u_{1}^{2}+w&u_{1}u_{2}&u_{1}u_{3}\\u_{2}u_{1}&u_{2}^{2}+w&u_{2}u_{3}\\u_{3}u_{1}&u_{3}u_{2}&u_{3}^{2}+w\\u_{1}&u_{2}&u_{3}\end{pmatrix}}$$
Other formulas clarify that $\nabla$ is applied to this matrix by differentiating the first column with respect to $x_1$, the second with respect to $x_2$, the third with respect to $x_3$, and then adding the results. The result being
$$
\nabla\cdot {\begin{pmatrix}u_{1}^{2}+w&u_{1}u_{2}&u_{1}u_{3}\\u_{2}u_{1}&u_{2}^{2}+w&u_{2}u_{3}\\u_{3}u_{1}&u_{3}u_{2}&u_{3}^{2}+w\\u_{1}&u_{2}&u_{3}\end{pmatrix}}
=\begin{pmatrix}
(\operatorname{div} \mathbf u)\mathbf u + \operatorname{grad} \mathbf w
\\
\operatorname{div} \mathbf u
\end{pmatrix}
$$
where I used explicit names to avoid any further $\nabla$-confusion.
If we think of $\nabla$ as a symbolic vector of partial derivatives, $\nabla = \begin{pmatrix} \partial /\partial x_1 \\ \partial /\partial x_2 \\ \partial /\partial x_3 \end{pmatrix}$, then the above is more properly
$$
\begin{pmatrix}{\mathbf {u}}\otimes {\mathbf {u}}+w{\mathbf {I}}\\{\mathbf {u}}\end{pmatrix} \nabla =
{\begin{pmatrix}u_{1}^{2}+w&u_{1}u_{2}&u_{1}u_{3}\\u_{2}u_{1}&u_{2}^{2}+w&u_{2}u_{3}\\u_{3}u_{1}&u_{3}u_{2}&u_{3}^{2}+w\\u_{1}&u_{2}&u_{3}\end{pmatrix}} \begin{pmatrix} \partial /\partial x_1 \\ \partial /\partial x_2 \\ \partial /\partial x_3 \end{pmatrix}
$$
This makes sense for any matrix where the number of columns matches the dimension.