Suppose $X$ is a $n \times m$ matrix with elements $x_{ij}$ and $$\lim_{n \to \infty} \frac{X'X}{n}=Q,$$ where $Q$ is positive definite. Let $x_i$ be the $i$-th row vector of $X$. For fixed $m$, a conjecture is $$\max_{i=1, ..., n} x_i'(X'X)^{-1}x_i =0, \text{ as } n \to \infty.$$
For $m=1$, it can be proved. But how to prove or disprove in a general case? This might be related here
Upon reconsideration of your question, I believe that you are trying to ask the following.
If this is what you meant, please say so.
For this version of the question, the conjecture is correct. In particular, $$ \lim_{n \to \infty} x_i'(X_n'X_n)^{-1}x_i = \lim_{n \to \infty} \frac 1n x_i'\left(\frac{X_n'X_n}{n}\right)^{-1}x_i= \lim_{n \to \infty} \frac 1n x_i'Qx_i = 0. $$
I assume that what you're trying to say is something like this:
The answer to this question is no. For example, take $X(k) = \sqrt{k}I$. Then $\max_{i=1,\dots,n}x_i'(X'(k)X(k))^{-1}x_i = 1$ for all $k$.
For an more general example (with a non-square $X$), take $X(k) = \pmatrix{\sqrt{k}I_n & 0_{n \times (m-n)}}$. Again, we find that $\max_{i=1,\dots,n}x_i'(X'(k)X(k))^{-1}x_i = 1$ for all $k$.