The Internet of Things (IoT) promises new opportunities for better monitoring and control of thousands of objects and sensors in households and industrial applications. The viability of large-scale IoT is, however, still a challenge given that the most widely known options for connecting everyday objects, i.e. duty-cycled active radios such as WiFi, Bluetooth and Zigbee, are power-hungry and increase the cost of deployment and maintenance of the connected devices. The main argument of this thesis is that passive radios that use backscatter communication, which has been used primarily for RFIDs, can fill this gap as an ultra-low power replacement for active radios to enable truly large-scale IoT deployments. However, passive radios offer insufficient performance today in terms of bandwidth and range that makes them unattractive for integration in IoT applications. The main contributions of this thesis are: (1) xShift, enabling battery-free backscatter tags that can directly communicate with commodity radios without any additional infrastructure. (2) MIXIQ, a new ultra low power receiver design that leverages the available devices nearby for converting a simple envelope detector to a high-range, high-throughput receiver. (3) Polymorphic Radio (Morpho), a novel approach for ultra low power radio design based on combining active and passive radios, in order to enable robust and pervasive streaming and cloud offloading. The proposed contributions have all been successfully prototyped using off-the-shelf components and show promising performance.