An industry is dynamic in nature, and, over its evolution from emergence to growth and maturity, associated with various aspects of structures and actors involved. Creative destruction is typical, and there is an emerging view on the significance of spontaneous self-organization, unpredictability, and underlying order on industrial activities. In line with this stream of research, many scholars have considered chaos theory which is the study of nonlinear dynamic systems as a useful framework for explaining the dynamics of industries and their distinctive growth patterns. Accordingly, the purpose of the paper is to draw on chaos theory to study the nonlinear nature of industrial development. Important aspects of chaos theory include non-linear equation, parameters, the initial conditions, and in particular the character of strange attractors. The emphasis is to examine dynamic feedback systems as geometrical forms and consider creative and continuous change behavior. The advantage is centered on the theory's ability to demonstrate how a simple set of deterministic relationships can result in patterning yet unpredictable outcomes. As the outcomes of the chaotic systems are bounded and embody mathematic constants, it is likely to detect a fundamental order behind complex events and actions. Interests on the application of chaos theory to understanding industry activity and firm behavior arise greatly thus, and Brown and Eisenhardt (1998), Levy (1994), Radzicki (1990), Stacey (1995), Radzicki (1990), Young (1997), and Wolfgang (2003), for example, have provided important implications of chaos theory both for research and practice. Following this emerging literature, in this paper, we use chaos theory to examine how an industry evolves over time. Using the number of patents in USPTO (United States Patent and Trademark Office) as a proxy of industry outcomes, we focus on the analyses of three industries-software, semiconductor and biotechnology-with the application of Lyapunov exponent (λ) as a measurement of chaos. Our search strategies in the USPTO database are specified as below: ACLM/software and APD/1/1/1997->1/31/1997 ACLM/semiconductor and APD/1/1/1997->1/31/1997 ACLM/DNA and APD/1/1/1997->1/31/1997 The calculation of λ values of the software industry is 0.1722 during 1992-1996, of semiconductor industry is 0.1434 during 1988-1997, and of the biotechnology industry is 0.2924 during 1993-1996. These results show that all of the industries had evolved from the period of λ<0 through the bifurcation point to the period of λ>0 during the time observed. Comparatively, the biotechnology had the highest value of λ, implying that the industry embodied a higher competitive condition and shorter life cycle. We also show that the three industries were all structured as non-linear, open-ended systems and as far-from-equilibrium paradoxes where the dynamics of the industries were both stable and unstable at the same time. All had affected by positive and negative forces that may co-exist during the evolving process. Path dependence was observable, and industry development was socially sensitive to technological innovation associated with self-organizing forces and endogenous change. Implications of our study are many, but the most critical is that we extend the use of chaos theory to the study of industry development which has been so far dominated by the industry organization school (e.g., five forces analysis) emphasizing on competition and stability. Our analysis shows that an industry, although containing the character of unpredictability and complex interactions among a variety of intended and unintended actions, is likely to be evolved in different yet distinctive growth patterns. Forecast and thus firm strategy are thus possible, as organizations retain the capacity of knowledgeability to adapt to the changing environment.