In a simple linear regression model $\hat{y} = \beta_0 + \beta_1 x$, assume the error terms $\epsilon_i$ satisfy the Gauss-Markov theorem assumptions, except for homoscedasticity. Specifically, $Var(\epsilon_i) = \sigma^2 x_i^2$. What is the primary implication for the Ordinary Least Squares (OLS) estimator $\hat{\beta}_1$?