Wi-Fi is the dominant wireless indoor broadband solution and thus key for meeting the exponential traffic growth. The recent IEEE 802.11ac amendment enables PHY data rates exceeding 1 Gbps. However, it is not clear how this increased per-link performance, achieved especially via wider channels, translates to network-level performance. The latter is crucial for understanding the true potential of emerging Wi-Fi, as massive densification of network infrastructure is needed for keeping up with capacity demands. In this paper we present results from an extensive measurement study of the performance of IEEE 802.11ac Wi-Fi in a large 24-node indoor testbed. We investigate in detail the impact of channel width, network deployment density, and type/volume of traffic on the achieved network performance in dense indoor deployments. Our results show that using wide 80 MHz channels is beneficial only in dense networks with extremely high traffic loads, owing to strong adjacent channel interference (ACI) effects with narrower channels. We show that ACI not only reduces aggregate network throughput but causes severe unfairness among nodes, where up to half the nodes may experience starvation. Starvation occurs due to frequency flow-in-the-middle effects and the heterogeneous interference coupling among pairs of nodes typical of indoor deployments. However, our results also demonstrate that for bursty TCP traffic with loads typical of modern residential Wi-Fi deployments, there is no harm from using narrower 20 or 40 MHz channels. Finally, we study the scaling of IEEE 802.11ac deployments with increasing traffic load, which is highly relevant for the dimensioning of future networks. We show that ACI effects only become evident at very high loads which are beyond the per-node traffic demand expected in the foreseeable future. Therefore, in practice there is no network-level benefit from employing the wider channels enabled by IEEE 802.11ac, even in emerging highly dense indoor deployments.