Christiaan Hogendorn, Brett Frischmann
Studies of economic growth often refer to general purpose technology (GPT), infrastructure, and openness as keys to improving productivity. Some GPTs, like railroads and the Internet, fit common notions of infrastructure and spawn debates about openness, such as network neutrality. Other GPTs, like the steam engine and the computer, seem to be in a different group that is more modular and open by nature. Big data, artificial intelligence, and various emerging smart technological assemblages have been described both as GPTs and infrastructure. We present a technology flow framework that clarifies when a GPT is implemented through infrastructure, provides a basis for policy analysis, and defines empirical research questions. On the demand side, all GPTswhether implemented through infrastructure or notenable a wide variety of productive uses and generate substantial spillovers to the rest of the economy. On the supply side, infrastructure is different from many other implementations of GPTs; infrastructure is partially nonrival, which may complicate appropriation problems and raise congestion issues. It also exhibits tethering, meaning that different users must be physically or virtually connected for the infrastructure to function, and this makes control of its uses more feasible and more salient to policy.