Forecasting the Future Capacities of Superconducting Quantum Computers: Extending Moore's Law Through Machine Learning
Abstract
Quantum computing has emerged as a promising technology that can perform certain tasks exponentially faster than classical computers. Despite the potential for quantum computers to revolutionize the field of computing, the development of fault-tolerant quantum computers remains a critical challenge. Moore's Law has accurately predicted the exponential growth in the capacity of classical computers, with transistor capacity doubling roughly every year. This prediction, established in the 1960s, held true until the early 2010s. However, the emergence of quantum computers raises questions about how to predict the rate of development these technologies. This work presents a novel approach using machine learning to extend classical Moore's Law into a quantum Moore's Law. Unlike previous attempts, which relied on limited quantum computer data, this model incorporates historical classical transistor data to predict qubit capacities. This thesis proposes a novel approach to forecasting the future capacities of superconducting qubits and gate speeds using machine learning. The proposed model builds upon Moore's Law and its predictions for the transistor capacity of classical computers. First, it establishes a polynomial relationship between the number of qubits and the number of classical transistors. Then, it trains a machine learning model to predict the number of classical transistors for future years. This prediction is used in conjunction with the established relationship to estimate the number of qubits for a given year. The same methodology is applied on data on the best achieved classical computations per second values to predict the speed of execution of quantum gates in the future. The findings indicate that the proposed model outperforms previously proposed methods in predicting qubit capacities, suggesting an improved method for predicting the future capacities of superconducting qubits and gate speeds based on the relationship between qubit and classical transistor capacities. Using a data-driven approach, the model can incorporate new data as quantum milestones are achieved. In this study, we present a novel approach to predicting the growth of quantum computing by extending and evolving classical Moore’s Law using machine learning. Our proposed model makes use of historical information on classical transistors to estimate recent qubit capacities more accurately than earlier studies, showing improved prediction accuracy in comparison to previous work. The proposed model provides valuable insight into the potential trajectory of quantum computing technology if Moore’s Law continues to hold in this domain.