We propose a preferential attachment model for network growth where new entering nodes have a partial information about the state of the network. Our main result is that the presence of bounded information modifies the degree distribution by introducing an exponential tail, while it preserves a power law behaviour over a finite small range of degrees. On the other hand, unbounded information is sufficient to let the network grow as in the standard Barabási-Albert model. Surprisingly, the latter feature holds true also when the fraction of known nodes goes asymptotically to zero. Analytical results are compared to direct simulations.