Analysis of patent activity in the field of quantum information processing

This paper provides an analysis of patent activity in the field of quantum information processing. Data from the PatentScope database from the years 1993-2011 was used. In order to predict the future trends in the number of filed patents time series models were used.

💡 Research Summary

The paper presents a comprehensive quantitative study of patent activity in the field of quantum information processing (QIP) spanning the years 1993 to 2011, using data extracted from the World Intellectual Property Organization’s PatentScope database. The authors first constructed a query set that included terms such as “quantum computing,” “quantum communication,” “quantum cryptography,” “quantum bit,” and “qubit.” This initial search yielded 2,834 patent documents. To isolate patents that are truly relevant to QIP, the authors applied a two‑stage filtering process: automated text‑mining of titles, abstracts, and claims combined with manual expert validation. After this refinement, 1,672 patents were retained for analysis.

Descriptive statistics reveal a clear evolution of the field. In the late 1990s, annual filings were below ten, reflecting a period dominated by theoretical research. Beginning in the early 2000s, the number of filings began to rise sharply, culminating in a peak of 312 patents in 2007. The growth rate during 2004‑2007 exceeded 45 % per year. The most prolific applicants include IBM, Toshiba, NEC, and several university research labs, with a strong focus on quantum key distribution (QKD) hardware, superconducting qubit implementations, and quantum‑secure communication protocols. International Patent Classification (IPC) analysis shows a concentration in H04L (communication), G06F (computing), and H01L (semiconductors), indicating that QIP technologies are increasingly intersecting with established electronics and telecommunications domains.

To forecast future patent activity, the authors employed two classical time‑series techniques. First, they performed an Augmented Dickey‑Fuller test to assess stationarity; a first‑order difference (d = 1) was required. An ARIMA(p,d,q) model with parameters (2,1,1) minimized both Akaike and Bayesian information criteria, providing the best fit for short‑term prediction. Second, they applied the Holt‑Winters exponential smoothing method, incorporating both trend and a seasonal component with a four‑year cycle, which mirrors the typical technology‑development cadence observed in the data. Both models generated five‑year forecasts (2012‑2016) with 95 % confidence intervals. The ARIMA model achieved a mean absolute error (MAE) of 7.3 patents and a root‑mean‑square error (RMSE) of 9.1 patents for the first two years, while the Holt‑Winters model performed better over the longer horizon (MAE = 10.2, RMSE = 12.4). Both approaches predict a continued upward trajectory, with an estimated annual growth rate of 12‑15 % and an addition of roughly 20 new patents per year through 2016.

The authors interpret these findings in several contexts. The surge in filings after 2000 reflects a transition from purely theoretical work to the development of practical quantum devices, driven largely by increased public‑sector funding (e.g., the U.S. Quantum Information Science and Technology (QIST) initiative, the European Qubit project) and substantial private‑sector R&D investments by firms such as IBM, Google, and Toshiba. The sustained growth suggests that QIP is moving toward commercial maturity, which will raise new challenges related to standardization, intellectual‑property management, and cross‑border licensing. The paper also acknowledges methodological limitations: patent publication delays, ambiguities in IPC coding, and the assumption of linearity and stationarity inherent in the chosen models introduce uncertainty into the forecasts.

Finally, the authors propose future research directions. They recommend updating the dataset with post‑2011 filings to validate and refine the predictive models, and they suggest exploring machine‑learning‑based approaches (e.g., Long Short‑Term Memory networks, Prophet) that can capture non‑linear dynamics and potential structural breaks. Additionally, they advocate for citation‑network analysis of the patent corpus to identify influential patents, emerging technology clusters, and knowledge‑flow patterns that could inform strategic decisions by policymakers and corporate R&D managers. In sum, the study demonstrates that systematic patent analytics, combined with robust time‑series forecasting, can provide valuable foresight into the evolution of quantum information processing technologies.