Wireless power transfer (WPT) is expected to be a technology reshaping the landscape of low-power applications such as the Internet of Things, machine-to-machine communications and radio frequency identification networks. Although there has been some progress towards multi-antenna multi-sine WPT design, the large-scale design of WPT, reminiscent of massive multiple-input multiple-output (MIMO) in communications, remains an open problem. Considering the nonlinear rectifier model, a multiuser waveform optimization algorithm is derived based on successive convex approximation (SCA). A lower-complexity algorithm is derived based on asymptotic analysis and sequential approximation (SA). It is shown that the difference between the average output voltage achieved by the two algorithms can be negligible provided the number of antennas is large enough. The performance gain of the nonlinear model based design over the linear model based design can be large, in the presence of a large number of tones.