The novel coronavirus COVID-19 is wreaking havoc on the world, both in terms of lives as well as in economic terms, especially in China of course. Understanding the dynamics of its spread is crucial to monitoring it and predicting its eventual outcome. Standard epidemiological models assume a fixed reproduction rate which leads to exponential growth for short times, then attenuation due to immunity, such as is predicted by the SEIR (Susceptible-Exposed-Infected-Recovered) model of Kermack and McKendrick from 1927. Similar methods are being used now to model this epidemic. We have found, concentrating on the data for deaths which is perhaps less open to ambiguities than confirmed infected people, that the behavior follows a power law with deaths going as time^x where the exponent x approx. 2.27 indicates a kind of fractal dynamics of the system. Here time is the number of days since the World Health Organization started publishing their reports (Jan. 21, 2020), where there were just three deaths. As of this writing (February 18), the data for the epidemic continue to follow this curve except for one day where there were some counting discrepancies. Some possible models that would predict power law growth will be discussed, as well as a general introduction of viruses and epidemiology.