A hypervisor determines a target CPU usage of a virtual machine in responding to a request to allocate guest memory. Then the hypervisor receives a first balloon request for a first quantity of guest memory to be ballooned. The hypervisor transmits a second balloon request for guest memory to be allocated to a balloon of the virtual machine, where the second quantity of guest memory is a first fraction of the first quantity of guest memory. The hypervisor then determines a first quantity of time that the virtual machine took to execute the second balloon request and waits a second quantity of time. The hypervisor then transmits a third balloon request for a third quantity of guest memory to be allocated to the balloon of the virtual machine, where the third quantity of guest memory is a second fraction of the first quantity of guest memory.