Intracortical microelectrode recordings of neural activity show great promise as control signals for neuroprosthetic applications. However, faithful, consistent recording of single unit spiking activity with chronically implanted silicon-substrate microelectrode arrays has proven difficult. Many approaches seek to enhance the long-term performance of microelectrode arrays by, for example, increasing electrode biocompatibility, decreasing electrode impedance, or improving electrode interface properties through application of small voltage pulses. The purpose of this study was to use computational models to optimize the design of microelectrodes. We coupled detailed models of the neural source signal, silicon-substrate microelectrodes, and thermal noise to define the electrode contact size that maximized the signal-to-noise ratio (SNR). Model analysis combined a multi-compartment cable model of a layer V cortical pyramidal neuron with a 3D finite element model of the head and microelectrode to define the amplitude and time course of the recorded signal. A spatially-lumped impedance model was parameterized with in vitro and in vivo spectroscopy data and used to define thermal noise as a function of electrode contact size. Our results suggest that intracortical microelectrodes with a contact size of ∼380 μm2 will provide an increased SNR in vivo and improve the long-term recording capabilities of silicon-substrate microelectrode arrays.