In this paper, a novel LPWAN technology, ZCNET, is proposed, which achieves significantly higher network capacity than existing solutions, such as LoRa, Sigfox, and RPMA. The capacity boost of ZCNET is mainly due to two reasons. First, a ZCNET node transmits signals that occupy a small fraction of the signal space, resulting in a low collision probability. Second, ZCNET supports 8 parallel root channels within a single frequency channel by using 8 Zadoff-Chu (ZC) root sequences. The root channels do not severely interfere with each other, mainly because the interference power is spread evenly over the entire signal space. A simple ALOHA-style protocol is used for medium access, with which a node randomly chooses the root channel and the range it occupies within the root channel. ZCNET has been extensively tested with both real-world experiments on the USRP and simulations, and the results confirm that ZCNET achieves significant gains over LoRa, Sigfox, and RPMA. ZCNET will likely better accommodate the explosive growth of IoT network sizes and meet the demand of IoT applications.