This paper presents an analytic model for computing the nonlinear power spectrum of dark matter, galaxies, and their cross-correlation. The model is based on the Press-Schechter halo model, which assumes that all matter is in virialized halos with a given mass and internal density profile. The total power spectrum is the sum of two contributions: one from correlations between halos and one from correlations within the same halo. The model shows that it can accurately reproduce results from N-body simulations, provided that the concentration parameter decreases with halo mass. Galaxy power spectra differ from dark matter power spectra because the number of galaxies per halo does not scale with halo mass, and most halos contain a central galaxy. If the number of galaxies per halo increases less rapidly than halo mass, the resulting power spectrum becomes a power law with a slope close to the observed value over several orders of magnitude in scale. The model predicts that galaxy clustering begins later than dark matter clustering, which is necessary to reconcile CDM models with observations. The model also predicts that bias is scale-dependent and nonmonotonic, which is particularly important for red or elliptical galaxies, which are found in larger halos. The model predicts that galaxy-dark matter correlations, observable through galaxy-galaxy lensing, cannot be interpreted as an average halo profile because different halo masses dominate at different scales and larger halos host more than one galaxy. The model computes predictions for the cross-correlation coefficient as a function of scale and discusses the potential of using cross-correlations with galaxy clustering to determine the dark matter power spectrum. The model is tested against semi-analytic models of galaxy formation and shows good agreement with simulations. The galaxy power spectrum is found to be a power law over several decades in scale, in agreement with observations. The model predicts that the slope of the galaxy power spectrum is close to the observed value and that the bias is scale-dependent and nonmonotonic. The model also predicts that red or elliptical galaxies, which are found in more massive halos, have a different bias dependence compared to normal galaxies. The model is applied to dark matter-galaxy cross-correlations, which are measured through galaxy-galaxy lensing and correlations between foreground and background galaxies. The cross-correlation power spectrum includes halo-halo and halo Poisson terms. The halo-halo term dominates on large scales and the halo Poisson term dominates on small scales. The model shows that the cross-correlation power spectrum can be used to determine the dark matter power spectrum from the galaxy power spectrum. The model is consistent with semi-analytic models and N-body simulations, indicating that it is a valid tool for predicting galaxy and dark matter power spectra.This paper presents an analytic model for computing the nonlinear power spectrum of dark matter, galaxies, and their cross-correlation. The model is based on the Press-Schechter halo model, which assumes that all matter is in virialized halos with a given mass and internal density profile. The total power spectrum is the sum of two contributions: one from correlations between halos and one from correlations within the same halo. The model shows that it can accurately reproduce results from N-body simulations, provided that the concentration parameter decreases with halo mass. Galaxy power spectra differ from dark matter power spectra because the number of galaxies per halo does not scale with halo mass, and most halos contain a central galaxy. If the number of galaxies per halo increases less rapidly than halo mass, the resulting power spectrum becomes a power law with a slope close to the observed value over several orders of magnitude in scale. The model predicts that galaxy clustering begins later than dark matter clustering, which is necessary to reconcile CDM models with observations. The model also predicts that bias is scale-dependent and nonmonotonic, which is particularly important for red or elliptical galaxies, which are found in larger halos. The model predicts that galaxy-dark matter correlations, observable through galaxy-galaxy lensing, cannot be interpreted as an average halo profile because different halo masses dominate at different scales and larger halos host more than one galaxy. The model computes predictions for the cross-correlation coefficient as a function of scale and discusses the potential of using cross-correlations with galaxy clustering to determine the dark matter power spectrum. The model is tested against semi-analytic models of galaxy formation and shows good agreement with simulations. The galaxy power spectrum is found to be a power law over several decades in scale, in agreement with observations. The model predicts that the slope of the galaxy power spectrum is close to the observed value and that the bias is scale-dependent and nonmonotonic. The model also predicts that red or elliptical galaxies, which are found in more massive halos, have a different bias dependence compared to normal galaxies. The model is applied to dark matter-galaxy cross-correlations, which are measured through galaxy-galaxy lensing and correlations between foreground and background galaxies. The cross-correlation power spectrum includes halo-halo and halo Poisson terms. The halo-halo term dominates on large scales and the halo Poisson term dominates on small scales. The model shows that the cross-correlation power spectrum can be used to determine the dark matter power spectrum from the galaxy power spectrum. The model is consistent with semi-analytic models and N-body simulations, indicating that it is a valid tool for predicting galaxy and dark matter power spectra.