PCA on Three Centered Points

Consider the centered design matrix

representing three training points in $\mathbb{R}^2$. What percentage of the total variability in the data is captured by the first principal component?